CN104902246A - Video monitoring method and device - Google Patents

Abstract

The invention provides a video monitoring method and device and relates to monitoring field. The video monitoring method comprises: acquiring a video image; acquiring the three-dimensional coordinate information of a target according to the video image; extracting event generation according to a positional relation between the target and a virtual door, wherein the virtual door includes the three-dimensional coordinate information. According to such method, the three-dimensional coordinate information of the target is acquired according to the video image and the positional relation between the virtual door and the target is determined on the basis of the three-dimensional coordinate information of the virtual door and the target, and thus the event generation is extracted. As a result, event misjudgment caused by a perspective effect in a two-dimensional image is effectively prevented and event judgment accuracy is improved.

Description

Video monitoring method and device

Technical field

The present invention relates to supervision field, particularly a kind of video monitoring method and device.

Background technology

Intelligent video behavioural analysis system has very high using value in various monitoring place, its basic universal method is by carrying out background modeling to input video, the image of background image and present frame is utilized to detect moving target, follow-up moving target to be followed the tracks of, classify and behavioural analysis, or the mode adopting training to identify directly detects specified type target from video, the target detected is followed the tracks of and analyzed, and early warning judgement is carried out to behavior event, to reach the object of intelligent monitoring.

In behavioural analysis, it is basic measuring ability that line of stumbling detects with region intrusion detection.It is embodied as substantially: arrange at least one line segment or a region at video image, and whether the moving target detected in video is crossed over this line segment or enter/leave this region, if event occurs, then produces warning.Wherein, line of stumbling detects in video image, arranges the directive line segment of at least one band, detects moving target and whether moves to opposite side from the side of line, if line behavior of stumbling occurs, produces alert event; Whether region intrusion detection arranges at least one surveyed area in video image, detect moving target and enter in this region from a region, if there is region intrusion behavior to occur, produces alert event.

Existing line and the region Intrusion Detection Technique of stumbling, direct whether intersecting according to target and set line of stumbling, region judges whether to trigger respective rule on the image plane.Because video camera imaging exists transparent effect, when target in image with stumble line or region crossing time, in real world, the action might not mixed line or enter, therefore easily produces erroneous judgement, and make a mistake warning.

Summary of the invention

One object of the present invention is the problem solved because the transparent effect of video camera causes event to judge by accident.

According to an aspect of the present invention, a kind of video monitoring method is provided, comprises: obtain video image; The three-dimensional coordinate information of target is obtained according to video image; The position relationship of based target and virtual door extracts event and occurs, and wherein, virtual door comprises three-dimensional coordinate information.

Alternatively, virtual door is door region perpendicular to the ground, and the intersection on virtual door and ground is straight line, line segment or broken line.

Alternatively, the three-dimensional coordinate information obtaining target according to video image comprises: contrast successive frame video image, or video image and background image is contrasted, and obtains the change point in video image or point group; From change point or point group, extraction point or point group are as target; Determine the three-dimensional coordinate information of target.

Alternatively, video image is plane video image; The three-dimensional coordinate information obtaining target according to video image comprises: the plane coordinates information being obtained target by plane video image; Carry out 3D reconstruction according to plane coordinates information by 3D algorithm for reconstructing, obtain the three-dimensional coordinate information of target.

Alternatively, the equipment obtaining video image comprises a 2D video camera.

Alternatively, video image comprises the plane video image of multiple different camera site; The three-dimensional coordinate information obtaining target according to video image comprises: multiple plane video image is carried out 3D and rebuild acquisition 3D video image; The three-dimensional coordinate information of target is obtained according to 3D video image.

Alternatively, the equipment obtaining video image comprises two or more 2D video camera or the 3D video camera based on binocular vision.

Alternatively, video image is depth image; The three-dimensional coordinate information obtaining target according to video image is, obtains the three-dimensional coordinate information of target according to depth image.

Alternatively, the equipment obtaining depth image comprises distance sensitive device or 3D video camera.

Alternatively, the position relationship of based target and virtual door extracts event and occurs as: extract event according to the position relationship of the horizontal coordinate information in the three-dimensional coordinate information of target and virtual door and occur, wherein, virtual door comprises the horizontal coordinate information in three-dimensional coordinate.

Alternatively, also comprise: according to the movement locus of multi-frame video image determination target in video image; Determine the three-dimensional coordinate information of the movement locus of target; The movement locus of based target and the position relationship of virtual door extract event and occur.

Alternatively, event comprise be positioned at virtual door, be positioned at virtual outdoors, be positioned at virtual door region, from outside to inside through virtual door, from inside to outside through virtual door, from outside to inside motion and through virtual door and/or from inside to outside motion and not through virtual door.

Alternatively, the type of target comprises people, animal and/or car.

Alternatively, also comprise, if extract scheduled event, then send warning message, warning message includes invades positional information, intrusion directional information.

Alternatively, the position relationship of based target and virtual door extracts event and comprises, the continuous frame number of statistical phenomeon, and when frame number is greater than predetermined warning frame number, decision event occurs.

By such method, the three-dimensional coordinate information of target is obtained according to video image, position relationship both judging based on the three-dimensional coordinate information of virtual door and target, thus the event of extraction occurs, effectively prevent the event caused due to transparent effect in two dimensional image to judge by accident, improve the accuracy that event judges.

According to another aspect of the present invention, a kind of video monitoring apparatus is provided, comprises: video acquisition module, for obtaining video image; Three-dimensional coordinate determination module, for obtaining the three-dimensional coordinate information of target according to video image; Event Distillation module, the position relationship for based target and virtual door extracts event generation, and wherein, virtual door comprises three-dimensional coordinate information.

Alternatively, virtual door is door region perpendicular to the ground, and the intersection on virtual door and ground is straight line, line segment or broken line.

Alternatively, Target Acquisition module comprises: frame contrast unit, for contrasting successive frame video image, or contrasting video image and background image, obtaining the change point in video image or point group; Target determination unit, for extraction point or point group from change point or point group as target; Three-dimensional coordinate extraction unit, for determining the three-dimensional coordinate information of target.

Alternatively, video image is plane video image; Three-dimensional coordinate determination module comprises: plane coordinates acquiring unit, for being obtained the plane coordinates information of target by plane video image; Three-dimensional coordinate extraction unit, for carrying out 3D reconstruction according to plane coordinates information by 3D algorithm for reconstructing, obtains the three-dimensional coordinate information of target.

Alternatively, video acquisition module comprises a 2D video camera.

Alternatively, video image comprises the plane video image of multiple different camera site; Three-dimensional coordinate determination module comprises: 3D reconstruction unit, rebuilds acquisition 3D video image for multiple plane video image being carried out 3D; Three-dimensional coordinate extraction unit, for according to 3D video image, obtains the three-dimensional coordinate information of target.

Alternatively, video acquisition module comprises two or more 2D video camera or the 3D video camera based on binocular vision.

Alternatively, video image is depth image; Three-dimensional coordinate determination module, for obtaining the three-dimensional coordinate information of target according to depth image.

Alternatively, video acquisition module comprises distance sensitive device or 3D video camera.

Alternatively, Event Distillation module occurs for extracting event according to the position relationship of the horizontal coordinate information in the three-dimensional coordinate information of target and virtual door, and wherein, virtual door comprises the horizontal coordinate information in three-dimensional coordinate.

Alternatively, three-dimensional coordinate determination module also comprises: track determining unit, for determining the movement locus of target in video image; Three-dimensional coordinate extraction unit, also for determining the three-dimensional coordinate information of the movement locus of target; Event Distillation module, also extracts event for the movement locus of based target and the position relationship of virtual door and occurs.

Alternatively, event comprise be positioned at virtual door, be positioned at virtual outdoors, be positioned at virtual door region, from outside to inside through virtual door, from inside to outside through virtual door, from outside to inside motion and through virtual door and/or from inside to outside motion and not through virtual door.

Alternatively, also comprise target type analysis module, for evaluating objects type, the type of target comprises people, animal and/or car.

Alternatively, also comprise alarm module, for sending warning message according to the scheduled event extracted, warning message includes invades positional information and/or intrusion directional information.

Alternatively, Event Distillation module is also for the continuous frame number of statistical phenomeon, and when frame number is greater than predetermined warning frame number, decision event occurs.

By such device, the three-dimensional coordinate information of target is obtained according to video image, position relationship both judging based on the three-dimensional coordinate information of virtual door and target, thus the event of extraction occurs, effectively prevent the event caused due to transparent effect in two dimensional image to judge by accident, improve the accuracy that event judges.

Accompanying drawing explanation

Accompanying drawing described herein is used to provide a further understanding of the present invention, and form a application's part, schematic description and description of the present invention, for explaining the present invention, does not form inappropriate limitation of the present invention.In the accompanying drawings:

Fig. 1 is the flow chart of an embodiment of video monitoring method of the present invention.

Fig. 2 is the flow chart of another embodiment of video monitoring method of the present invention.

Fig. 3 is the flow chart determining a three-dimensional coordinate part embodiment of target of the present invention.

Fig. 4 is the flow chart of another embodiment of video monitoring method of the present invention.

Fig. 5 is the flow chart of another embodiment of video monitoring method of the present invention.

Fig. 6 is the schematic diagram of an embodiment of video monitoring devices of the present invention.

Fig. 7 is the schematic diagram of an embodiment of three-dimensional coordinate determination module in video monitoring devices of the present invention.

Fig. 8 is the schematic diagram of another embodiment of three-dimensional coordinate determination module in video monitoring devices of the present invention.

Fig. 9 is the schematic diagram of another embodiment of three-dimensional coordinate determination module in video monitoring devices of the present invention.

Figure 10 is the schematic diagram of another embodiment of video monitoring devices of the present invention.

Embodiment

Below by drawings and Examples, technical scheme of the present invention is described in further detail.

The flow chart of an embodiment of video monitoring method of the present invention as shown in Figure 1.

In a step 101, picture pick-up device obtains the video image of guarded region.

In a step 102, the three-dimensional coordinate information needing the target monitored in video image is determined.This target can be obtained by the contrast of current image and background image, or can be obtained by the contrast of a few two field picture in front and back.This target can be motion or static object, also can be the pixel changed in image or point group.The target three-dimensional coordinate information determined according to video image can under camera coordinates system, also can under earth axes.

In step 103, determine the position relationship of the two according to the three-dimensional coordinate information of target and virtual door, thus the generation of the event of extraction.

By such method, the three-dimensional coordinate information of target is obtained according to video image, position relationship both judging based on the three-dimensional coordinate information of virtual door and target, thus the event of extraction occurs, effectively prevent the event caused due to transparent effect in two dimensional image to judge by accident, improve the accuracy that event judges.

In one embodiment, the target that video monitoring obtains simultaneously can be multiple, thus the leakage of the event of minimizing is extracted.

Virtual door is door region perpendicular to the ground, and the intersection on virtual door and ground can be straight line, line segment or broken line.By such method, the border in the region of wanting monitoring and protection can be delimited as much as possible, and all monitor from ground to space, improve the comprehensive and accurate property of Event Distillation.

Virtual door upwards extends on the basis of this straight line, line segment or broken line, can be highly unlimited, also can predetermined altitude.Virtual door can be arranged by the mode of the boundary line arranging virtual door and ground; Also directly can set virtual door by the mode of delimiting convex polygon, this polygon is perpendicular to ground, and polygonal lower boundary is the intersection on virtual door and ground; Can also be set by the mode of the distance arranging virtual door and camera; Or first set the boundary line on virtual door extended surface and ground, then arrange virtual door region, virtual door up-and-down boundary can be specified by user images, or setting height.By such mode, the virtual door of free setting can be needed according to monitoring, have more flexibility, make video monitoring regional more targeted.

In one embodiment, the video image of successive frame can be contrasted, or video image and background image are contrasted, according to the difference of the value of same position pixel, obtain change point or point group.According to the difference of video image kind, the value of pixel can comprise color information, half-tone information or depth information.Can pass through denoising, the mode rejecting error point obtains target from change point or point group, or arranges thresholding, determines that change point or point group are target when changing value exceedes threshold value.

By such method, the point of change or point group can be captured as target, thus improve the sensitivity of monitoring, reduce the probability omitted.

The flow chart of another embodiment of video monitoring method of the present invention as shown in Figure 2.

In step 201, picture pick-up device obtains the video image of guarded region, and this video image is plane video image.

In step 202., the plane coordinates information of target is determined.This target can be positioned at that guarded region is static, the object of motion, can also be by frame contrast before and after plane video image or with the contrast of background image, the point that the value of pixel changes or point group.

In step 203, by 3D algorithm, plane video image is carried out 3D reconstruction, obtain the three-dimensional coordinate information of target.

In step 204, according to the three-dimensional coordinate information of target and the three-dimensional coordinate information of virtual door, both relative position relations are determined, the generation of decision event.

By such method, 3D reconstruction can be carried out according to the plane coordinates information of monoplane video image and target, obtain the three-dimensional coordinate information of target, according to the generation of three-dimensional coordinate information decision event, equipment is installed and is demarcated simply, and avoid the event erroneous judgement caused due to transparent effect in two dimensional image, improve the accuracy that event judges.

Fig. 3 is the flow chart of an embodiment of the three-dimensional coordinate of plane coordinates information determination target according to monoplane video image and target.

In step 301, utilize an object in the picture in plane the elevation information of three diverse locations obtain the line that goes out (vanishing line, being a concept the most basic in projective geometry, is exactly the straight line that the intersection point of all parallel lines after projection on image in real world is formed) equation.Need to know this object height in space, three diverse locations of acquisition need not point-blank, and elevation information can represent by pixel.

In step 302, the length information of straight line on ground level is utilized to represent the angle obtaining the camera intrinsic X-axis anglec of rotation and rotate around Y-axis.The straight line length in space obtained is known, and its length information in the picture can represent by pixel.

In step 303, the anglec of rotation of go out line equation and video camera is utilized to obtain the mapping matrix of 2D and 3D.

In step 304, according to the mapping matrix in step 303, and plane coordinates, obtain three-dimensional coordinate.

By such method, according to the plane coordinates of target, virtual door, three-dimensional coordinate can be obtained, thus carry out Event Distillation according to three-dimensional coordinate.

In one embodiment, the implementation method obtaining the corresponding relation between 2D plane video image and three-dimensional body is specific as follows:

First, the plane that video monitoring is paid close attention to is demarcated, after demarcating, the Euclidean distance that in this plane, any two points is corresponding in real world coordinates system can be obtained.

Corresponding relation between two dimensional image and three-dimensional body can be expressed as formula:

$\mathrm{\λ}\left[\begin{array}{c}u\\ v\\ 1\end{array}\right]=P\left[\begin{array}{c}X\\ Y\\ Z\\ 1\end{array}\right]---\left(1\right)$

Wherein λ represents the distortion factor of video camera, considers that general distortion of camera is all smaller, is reduced to λ=1 here, and therefore the emphasis of plane reference is and obtains projection matrix P.Known through deriving, this matrix can by α, (α represents the anglec of rotation (tilt angle) of camera around X-axis to β, β represents the anglec of rotation (pan angle) of camera around Y-axis) and line equation decision of going out, concrete derivation details can list of references: " Fengjun Lv Tao Zhao Ram Nevatia; Self Calibration of a camera from video of a walking human, ICPR, 2002 ".

Then utilize an object (highly known in space) in the picture on ground level the elevation information (pixel represents) of three diverse locations (not point-blank) obtain the line equation that goes out, utilize the length information (pixel represents) of straight line on ground level (in space, length is known), obtain α, β, thus obtain projection matrix P to demarcate a plane.Concrete grammar is as follows:

A, user find ground level on the video image of input, and taking up an official post at ground level means fixed two points, and its location of pixels is expressed as (u ₁, v ₁) and (u ₂, v ₂), and provide the Euclidean distance d between the coordinate of these two points in real world coordinates system.

B, use following method to calculate optimum α, β: first, by α, β respectively discretization between 0 degree to 360 degree, to the combination (α of each possible α, β _i, β _i) a structure mapping matrix P _i, by the location of pixels (u obtained in steps A ₁, v ₁) and (u ₂, v ₂) pass through P _ian Euclidean distance d is calculated after obtaining the coordinate of the three-dimensional real world of its correspondence _i, by the d minimum with d error _icorresponding (α _i, β _i) as camera or camera parameters.

Size due to α, β is all between 0 degree to 360 degree, can by α, β respectively discretization, and as α gets 1 degree, 2 degree ... .360 degree, β gets 1 degree, 2 degree ... .360 degree, for wherein each possible angle value as one group of candidate combinations (α _i, β _i).

Formula 1 is out of shape a little, obtains formula 2:

$\left[\begin{array}{c}X\\ Y\\ Z\\ 1\end{array}\right]={\mathrm{\λP}}^{-1}\left[\begin{array}{c}u\\ v\\ 1\end{array}\right]---\left(2\right)$

Wherein P ^-1the inverse matrix of representing matrix P, i.e. P ^-1p=I, I represent the unit matrix of 3 × 3.The dimension of matrix P is 3 × 4, but considers that the point of demarcation is arranged on the ground level of real world, and namely the coordinate of Z-direction is 0, and like this, matrix P deteriorates to the matrix of 3 × 3, so just can invert.

By (u ₁, v ₁), (u ₂, v ₂) substitute into above-mentioned formula, obtain this world coordinates (X of 2 ₁, Y ₁, Z ₁) and (X ₂, Y ₂, Z ₂), then, calculate its Euclidean distance with the error delta (α of d _i, β _i), it specifically defines can a variety of form of expression, only recommends relatively more conventional two kinds here or | (X ₁-X ₂) ²+ (Y ₁-Y ₂) ²+ (Z ₁-Z ₂) ²-d ²|.

For the value of all possible α, β, select one group of parameter alpha that error is minimum ^*, β ^*parameter as optimum:

${\mathrm{\α}}^{*},{\mathrm{\β}}^{*}=\underset{({\mathrm{\α}}_i,{\mathrm{\β}}_i)}{\arg \min }\mathrm{\Δ}({\mathrm{\α}}_i,{\mathrm{\β}}_i)---\left(3\right)$

C, calculate and to go out line equation.In prior art, any one method obtaining the line that goes out all is suitable for the present invention, wherein, the computational methods of line of going out can list of references " Single-View Metrology:Algorithms and Applications ", Antonio Criminisi, proceeding of 24DAGM symposium on Pattern Recognition.

D, calculating estimated matrix two-dimensional coordinate are to the projection matrix P of three-dimensional coordinate.

Obtain above-mentioned camera parameter α, after β, projection matrix can be obtained by following formula:

P＝K[R|t] (4)

Wherein, matrix P is the mapping matrix of 3 × 4, and K is the inner parameter matrix of 3 × 3, $K=\left[\begin{array}{ccc}f& 0& u_0\\ 0& f& v_0\\ 0& 0& 1\end{array}\right],$ (u ₀, v ₀) representing the intrinsic point of video image, can represent with the central point of video image, represent the focal length of camera or video camera, R is the spin matrix of 3 × 3, represented by formula (5), wherein α represents the anglec of rotation (tilt angle) of camera intrinsic X-axis, β is the anglec of rotation (pan angle) around Y-axis, γ is the anglec of rotation (yaw angle) around Z axis $\mathrm{\α}=\mathrm{arcc}\tan \left(\frac{(u_{V_x}-u_0)\cos \mathrm{\β}}{f}\right),\mathrm{\β}=\arctan \left(\frac{v_0-v_{V_x}}{f}\right),$ γ is similar to the inclination angle of line equation relative to horizontal direction of going out.

$R=\left[\begin{array}{ccc}\cos \mathrm{\γ}& -\sin \mathrm{\γ}& 0\\ \sin \mathrm{\γ}& \cos \mathrm{\γ}& 0\\ 0& 0& 1\end{array}\right]\left[\begin{array}{ccc}1& 0& 0\\ 0& \cos \mathrm{\β}& -\sin \mathrm{\β}\\ 0& \sin \mathrm{\β}& \cos \mathrm{\β}\end{array}\right]\left[\begin{array}{ccc}\cos \mathrm{\α}& 0& -\sin \mathrm{\α}\\ 0& 1& 0\\ \sin \mathrm{\α}& 0& \cos \mathrm{\α}\end{array}\right]---\left(5\right)$

T is the matrix of 3 × 1, can be expressed as: t=R [0, Hc, 0] ^t, Hc represents the height on viewing apparatus distance ground, and T represents and carries out transpose operation to [0, Hc, 0].

E, because virtual door and target are in same level, namely Z axis coordinate is identical, therefore in computational process, omits the calculating of Z axis, only needs to obtain the XY axial coordinate of target under three-dimensional coordinate.Utilize formula (6) to bring the arbitrary coordinate point on ground level in image into mapping matrix P, obtain the XY axial coordinate under the three-dimensional coordinate of its correspondence.P ^-1represent the inverse matrix of the mapping matrix of 3 × 3 after process of degenerating.

$\left[\begin{array}{c}X\\ Y\\ 1\end{array}\right]={\mathrm{\λP}}^{-1}\left[\begin{array}{c}u\\ v\\ 1\end{array}\right]---\left(6\right)$

Use as above method, the Matrix Formula (6) changed between 2D and 3D can be obtained, plane coordinates is brought into formula (6), the three-dimensional coordinate that this plane coordinates is corresponding can be obtained.

Before carrying out image calibration, first can carry out the preliminary treatment such as such as noise reduction filtering, image enhaucament and/or electronic steady image to image, thus improve the accuracy detected.

By such method, can only have a 2D camera, obtaining image is under the condition of plane video image, and the plane coordinates of target is converted into three-dimensional coordinate, thus with less cost, equipment, realize the object of carrying out Event Distillation based on the three-dimensional coordinate of object.

Fig. 4 is the flow chart of another embodiment of video frequency monitoring method of the present invention.

In step 401, multiple plane video image from diverse location shooting is obtained.Multiple plane video image is the video image of monitoring the same area.

In step 402, multiple plane video image 3D is rebuild, obtain 3D video image.

In step 403, the three-dimensional coordinate information of target is obtained according to 3D video image.Point or the point group of change can be obtained according to the change of pixel coordinate, color information in 3D video image, from the point or point group of change, extract target, obtain the three-dimensional coordinate information of target according to 3D video image.

In step 404, according to the three-dimensional coordinate information of target and the three-dimensional coordinate information of virtual door, the position relationship of both judgements, extracts the generation of event.

By such method, multiple plane video image that same guarded region is taken from diverse location can be obtained, multiple plane video image is carried out 3D and rebuild acquisition 3D video image, the determination of target three-dimensional coordinate information is carried out according to 3D video image, thus three-dimensional coordinate information can be used to carry out the extraction of event, when avoiding using two-dimensional coordinate to carry out Event Distillation, due to the event erroneous judgement that transparent effect causes.Compared with carrying out the method for 3D reconstruction with use monoplane video image, more accurately, thus reduce the error of Event Distillation.

In one embodiment, two 2D video cameras are used to obtain plane video image from two incomplete same position shootings.Can be specific as follows by the method that two plane video image carry out 3D reconstruction:

First, Binocular Stereo Vision System based on twin camera must be arranged on a stabilised platform, when carrying out the shooting of monitoring scene, guarantee that the internal reference (such as focal length) of video camera and the position relationship of two video cameras can not change, otherwise will again demarcate system.Obtain the imaging picture of two video cameras, and result is analyzed, extract depth information.

In order to obtain higher precision, the focal length of video camera and the length of base can be made to increase, make guarded region as far as possible near stereo visual system, enough large and two video cameras of the overlapping region guaranteeing guarded region roughly align, and namely each camera intrinsic optical axis anglec of rotation can not be too large simultaneously.

A, elimination distortion.Make mathematically to eliminate the lens distortion in radial and tangential direction,

Radial distortion makes light in the place away from lens centre than more bending by paracentral place.For radial distortion, image space is corrected according to formula (7).

x ₁＝x(1+k ₁r ²+k ₂r ⁴+k ₃r ⁶)

y ₁＝y(1+k ₁r ²+k ₂r ⁴+k ₃r ⁶) (7)

Here, x ₁, y ₁for the reposition after correction, x, y are home position, for cheap web camera, use first two; Distort very large as flake video camera, use Section 3.

Tangential distortion is because lens manufacturing defect makes lens itself cause with the plane of delineation is not parallel.For tangential distortion, image space is corrected according to formula (8).

x ₂＝x+[2p ₁y+p ₂(r ²+2x ²)]

y ₂＝y+[p ₁(r ²+2y ²)+2p ₂x] (8)

Here, x ₂, y ₂for the reposition after correction, x, y are home position.

B, Camera calibration.The angle and distance of adjustment video camera, exports the correcting image that row aims at (in one plane, the every a line on image is close alignment to image).

C, images match: search the same characteristic features in two camera field of view, export disparity map, difference refers to the difference x of same coordinate on x coordinate on two images ₁-x ₂.

D, re-projection.After being aware of the relative geometry position of two video cameras, disparity map is become distance by the method migration of triangulation.

As shown in Figure 3, similar triangles are utilized can to release depth information Z value.

$Z=\frac{\mathrm{fT}}{x_1-x_2}---\left(9\right)$

Wherein, x ₁-x ₂be parallax, dimension is pixel, and the dimension of f is also pixel, and T is the centre-to-centre spacing of two video cameras, is generally set as millimeter.

By such method, the distance between plane video image and camera can be obtained, thus obtain depth information on the basis of plane coordinates, realize the acquisition of three-dimensional coordinate.

In one embodiment, the 3D video camera based on binocular vision can be used to obtain video image.Such method, can simplify many planar pickup seat in the plane and put the process selected and install demarcation, use convenient.

Fig. 5 is the flow chart of another embodiment of video frequency monitoring method of the present invention.

In step 501, depth image is obtained by 3D video camera, distance sensitive device etc.As kinect transducer, three-dimensional laser scanner or photographic-type scanner.Carry out the environment before perception camera lens by transducer, use the mode of such as black and white spectrum to judge the physical distance of article and transducer, collect the depth information of the every bit in the camera lens visual field, thus obtain an amplitude deepness image.In depth image, the value of pixel is depth information, and namely the numerical value of each pixel expresses the range information of physical environment corresponding to this pixel and camera.In one embodiment, kinect transducer is used to obtain depth information, infrared ray launched by the infrared transmitter of kinect transducer, structured light is formed through grating, speckle pattern is projected at body surface, cmos camera shooting speckle image, the distance corresponding according to reference speckle pattern obtains the approximate depth distance of object under test, utilize triangulation, the speckle pattern of measuring targets carries out local equalize, obtain depth image, every 30ms integrates an amplitude deepness image, and uses 3D effect models show.In one embodiment, can three-dimensional laser scanner or photographic-type scanner be passed through, by laser measurement principle, obtain the distance of the object before camera lens and camera lens, form the depth image be made up of the point off density cloud comprising three-dimensional coordinate information.

In step 502, the three-dimensional coordinate information of target is obtained according to depth image.Target can be the object being positioned at guarded region, also can be to be compared by successive frame depth image, and the point that depth information changes or point group can also be compared with background depth image, the point that depth information changes or point group.

In step 503, the three-dimensional coordinate information of based target and the position relationship of virtual door extract event and occur, and this three-dimensional coordinate information is the three-dimensional coordinate information of target under true environment.Extractible event comprises, and target appears in virtual door, target from outside to inside through virtual door or target be positioned at virtual outdoors etc.Can judge whether report to the police and determine warning message according to the relative position relation of target and virtual door.

By such method, the three-dimensional coordinate information of target is obtained by depth image, position relationship both judging based on the three-dimensional coordinate information of virtual door and target, thus the event of extraction occurs, effectively prevent the event caused due to transparent effect in two dimensional image to judge by accident, improve the accuracy that event judges.

In one embodiment, 2D video camera can be used to obtain the plane video image of guarded region, plane video image and depth image monitor the same area, the point changed according to the change detection of color information in plane video image or point group as target, then obtain the three-dimensional coordinate information of target according to depth image.

By such method, higher according to definition, that there is color information plane video image can obtain target, prevent because depth information change not obvious causing is mistaken for noise, thus reduce the probability leaking captured target, make monitoring more reliable.

In one embodiment, because target may be kept in motion, the generation of event can be extracted according to the movement locus of target.

By multi-frame video image before and after comparison, extract moving target, obtain the movement locus of target.Can according to the kind of the video image obtained, the method in Fig. 2, Fig. 3 or Fig. 4 is adopted to obtain the three-dimensional coordinate information of movement locus, the three-dimensional coordinate information of the movement locus of evaluating objects and the three-dimensional coordinate information of virtual door, the relative position relation of both judgements, thus decision event occurs.

Extract the generation of event according to the three-dimensional coordinate information of the movement locus of target and virtual door, the event extracted can comprise: from outside to inside through virtual door, from inside to outside through virtual door, from outside to inside motion and through virtual door, from inside to outside motion and not through virtual door.By such method, can continuing to monitor target be realized, improve the accuracy of Event Distillation.

In one embodiment, specific as follows according to the method for the three-dimensional coordinate information of target and the position relationship extraction event of virtual door:

The three-dimensional coordinate information of A, acquisition target and virtual door.Determine a consult straight line, choose here via the center lowest point of the image straight line vertical with image lower boundary.

B, the line of each line segment end points to reference point coordinate calculating virtual door setting in current frame image respectively and the angle of consult straight line, be denoted as θ respectively ₁, θ ₂... θ _m, m is end points number, calculates the line of coordinates of targets point to reference point coordinate and the angle α of consult straight line in current frame image, by θ ₁, θ ₂... θ _msort according to the size of numerical value with α, select the θ minimum value being greater than α to be denoted as T ₁, select the θ maximum being less than α, be denoted as T ₂, record T ₁, T ₂three-dimensional coordinate (x after corresponding line segment end points conversion ₁, y ₁) and (x ₂, y ₂), record the three-dimensional coordinate (x, y) after now moving target conversion, the three-dimensional coordinate (X, Y) after the conversion of record reference point.

C, the line of each line segment end points to reference point coordinate calculating virtual door setting in previous frame image respectively and the angle of consult straight line, be denoted as θ respectively ₁', θ ₂' ... θ _m', m is end points number, calculates the line of coordinates of targets point to reference point coordinate and the angle α ' of consult straight line in previous frame image, by θ ₁', θ ₂' ... θ _m' sort according to the size of numerical value with α ', select the θ ' minimum value being greater than α ' to be denoted as T ₁', select the θ ' maximum being less than α ' to be denoted as T ₂', record T ₁', T ₂three-dimensional coordinate (x after the line segment end points conversion of ' correspondence ₁', y ₁') and (x ₂', y ₂'), record the three-dimensional coordinate after the conversion of now moving target (x ', y ').

D, calculate T respectively ₁, T ₂three-dimensional coordinate (x after corresponding line segment end points conversion ₁, y ₁) and (x ₂, y ₂) change with reference point after the distance d of three-dimensional coordinate (X, Y) ₁, d ₂, three-dimensional coordinate (x, y) and the reference point calculated after moving target conversion change after the distance d of three-dimensional coordinate (X, Y).

d＝((X-x) ²+(Y-y) ²) ^1/2(10)

Judge d and d ₁and d ₂size, likely occur three kinds of result: d compare d ₁and d ₂all large, d compares d ₁and d ₂all little, d is between d ₁and d ₂between, be denoted as result 1.1 respectively, 1.2,1.3.

E, calculate T respectively ₁', T ₂three-dimensional coordinate (x after the line segment end points conversion of ' correspondence ₁', y ₁') and (x ₂', y ₂') change with reference point after the distance d of three-dimensional coordinate (X, Y) ₁', d ₂', three-dimensional coordinate (x, y) and the reference point calculated after moving target conversion change after the distance d ' of three-dimensional coordinate (X, Y).

Judge d ' and d ₁' and d ₂' size, likely occur three kinds of result: d ' compare d ₁' and d ₂' all large, d ' compares d ₁' and d ₂' all little, d ' is between d ₁' and d ₂' between, be denoted as result 2.1 respectively, 2.2,2.3.

F, carry out the judgement of the direction of motion according to result.

1.1,2.1 combinations as a result: account for motion target is greater than the line segment end points of virtual door setting from the distance of reference point from the distance of reference point always, does not occur passing through virtual door situation.

As a result 1.1,2.2 combination: account for motion target apart from reference point distance occur from be less than be greater than virtual door arrange line segment end points from the distance of reference point, pass through virtual door situation, direction is from inside to outside.

As a result 1.1,2.3 combination: account for motion target apart from reference point distance occur from be less than be greater than virtual door arrange line segment end points from the distance of reference point, pass through virtual door situation, direction is from inside to outside.

As a result 1.2,2.1 combination: account for motion target apart from reference point distance occur from be greater than be less than virtual door arrange line segment end points from the distance of reference point, pass through virtual door situation, direction is from outside to inside.

1.2,2.2 combinations as a result: account for motion target is less than the line segment end points of virtual door setting from the distance of reference point from the distance of reference point always, does not occur passing through virtual door situation.

As a result 1.2,2.3 combination: account for motion target apart from reference point distance occur from be greater than be less than virtual door arrange line segment end points from the distance of reference point, pass through virtual door situation, direction is from outside to inside.

As a result 1.3,2.1 combination: account for motion target apart from reference point distance occur from be greater than be less than virtual door arrange line segment end points from the distance of reference point, pass through virtual door situation, direction is from outside to inside.

As a result 1.3,2.2 combination: account for motion target apart from reference point distance occur from be less than be greater than virtual door arrange line segment end points from the distance of reference point, pass through virtual door situation, direction is from inside to outside.

1.3,2.3 combinations as a result: the line segment end points that account for motion target is arranged between virtual door from the distance of reference point always from reference point distance between, do not occur passing through virtual door situation, do not report to the police.

By such method, the generation of event can be extracted according to the motion state of target, judge the direction of motion of target and whether pass through virtual door, reaching comparatively accurate and detailed Event Distillation effect.

In one embodiment, the three-dimensional coordinate information of target obtained according to video image is the three-dimensional coordinate information under camera coordinates system, and the three-dimensional coordinate information of virtual door is the three-dimensional coordinate information under earth axes.Need the three-dimensional coordinate information unification of the three-dimensional coordinate information of target and virtual door under the same coordinate system.In one embodiment, according to the relation of camera coordinates system and earth axes, the three-dimensional coordinate information of target under camera coordinates system is transformed into earth axes.Virtual door can be perpendicular to the door region on ground, by unified to earth axes for the coordinate system of virtual door and target, relative position relation both can only judging according to the horizontal coordinate information of virtual door and target, the relative position relation decision event according to the two occurs.

By such method, can by the three-dimensional coordinate information unification of virtual door and target in earth axes, the position relationship both judging in the same coordinate system, improves the accuracy of Event Distillation.When virtual door is perpendicular to ground, the relative position relation both only judging according to horizontal coordinate information, reduces the complexity of Event Distillation.

In one embodiment, the three-dimensional coordinate information of virtual door is the three-dimensional coordinate information under camera coordinates system, or can according to the relation of camera coordinates system and earth axes, the three-dimensional coordinate information of virtual door under earth axes is transformed into camera coordinates system, thus obtaining target and the relative position relation of virtual door under camera coordinates system, the relative position relation decision event according to the two occurs.

By such method, can by the three-dimensional coordinate information unification of virtual door and target in camera coordinates system, the position relationship both judging in the same coordinate system, improves the accuracy of Event Distillation.Owing to not needing to change the three-dimensional coordinate information of target, simplify the step of data processing.

In one embodiment, the three-dimensional coordinate information of target obtained according to video image is the three-dimensional coordinate information under earth axes, and the three-dimensional coordinate information of virtual door is the three-dimensional coordinate information under earth axes.Relative position relation both can only judging according to the horizontal coordinate information of virtual door and target, the relative position relation decision event according to the two occurs.

By such method, only occur according to the horizontal coordinate information decision event of the two, the complexity of Event Distillation can be reduced.

In one embodiment, the position relationship of target and virtual door comprise be positioned at virtual door, be positioned at virtual outdoors, be positioned at virtual door region, from outside to inside through virtual door, from inside to outside through virtual door, from outside to inside move and through virtual door or from inside to outside move and do not pass virtual door.In these events, can according to real needs judge which or which as needing the event of reporting to the police, as being positioned at virtual door, from outside to inside through virtual door etc.Such method, can facilitate user to select the event needing to report to the police according to concrete use scenes, thus increase the available scene of this method.

In one embodiment, also warning function is comprised.The predetermined event needing to report to the police, as being positioned at virtual door, being in virtual door region, passing through virtual door from outside to inside and/or from inside to outside passing through virtual door etc.When these events occur, trigger alarm.By such method, can realize clearly providing warning message to user, help user to process event in time.

In one embodiment, target type analysis is also comprised.Target type comprises people, animal or car.By images match, obtain the type of target, thus enrich Event Distillation information.By such method, the target type needing to report to the police can be selected, reduce the workload of user.

In one embodiment, the frame number that statistical phenomeon is extracted, when the extraction of event reaches predetermined frame number, just determines that event occurs.By such method, erroneous judgement can be prevented, filter the false-alarm of a part.

The schematic diagram of an embodiment of video monitoring apparatus of the present invention as shown in Figure 6.Wherein, 601 is video acquisition module, and for obtaining the video image of guarded region, video acquisition module can be single or multiple 2D video camera, 3D video camera or distance sensitive device; The video image collected is sent to three-dimensional coordinate determination module 602 by video acquisition module 601, three-dimensional coordinate determination module 602 obtains the three-dimensional coordinate information of target according to video image, this target can be obtained by the contrast of current image and background image, or can be obtained by the contrast of a few two field picture in front and back, this target can be motion or static object, also can be the pixel changed in image or point group.The three-dimensional coordinate information of target is sent to Event Distillation module 603 by three-dimensional coordinate determination module 602; Event Distillation module 603, according to the three-dimensional coordinate information of target and the three-dimensional coordinate information determination target of virtual door and the position relationship of virtual door, extracts the generation of event.

By such device, the three-dimensional coordinate information of target is obtained according to video image, position relationship both judging based on the three-dimensional coordinate information of virtual door and target, thus the event of extraction occurs, effectively prevent the event caused due to transparent effect in two dimensional image to judge by accident, improve the accuracy that event judges.

Virtual door is door region perpendicular to the ground, and the intersection on virtual door and ground can be straight line, line segment or broken line.Such device, can delimit the border in the region of wanting monitoring and protection as much as possible, and all monitor from ground to space, improves the comprehensive and accurate property of Event Distillation.

Virtual door upwards extends on the basis of this straight line, line segment or broken line, highly can be unlimited, also can predetermined altitude.Virtual door can be arranged by the mode of the boundary line arranging virtual door and ground; Also can directly be set by the mode of delimiting convex polygon, this polygon is perpendicular to ground, and polygonal lower boundary is the intersection on virtual door and ground; Can also be set by the mode of the distance arranging virtual door and camera; Or first set the boundary line on virtual door extended surface and ground, then arrange virtual door region, virtual door up-and-down boundary can be specified by user images, or setting height.Arranging of virtual door can by the Interface realization of external program.Such device, can need the virtual door of free setting according to monitoring, have more flexibility, make video monitoring regional more targeted.

Fig. 7 is the schematic diagram of an embodiment of three-dimensional coordinate determination module in video monitoring devices of the present invention.Wherein, 701 is frame contrast unit, and the successive frame for video video acquisition module obtained contrasts, or the video image of present frame and background image is contrasted, and obtains pixel or point group that in video image, pixel value changes.According to the difference of video image kind, pixel value can be color information, three-dimensional coordinate information or depth information.702 is target determination unit, for for the frame contrast change point that extracts from video image of unit 701 or point group, by screening or arranging the operations such as thresholding, extracts target.Three-dimensional coordinate extraction unit 703 is according to the three-dimensional coordinate information of video image determination target.

Such device, can capture the point of change or point group as target, thus improve the sensitivity of monitoring, reduce the probability omitted.

In one embodiment, the video image that video acquisition module obtains is plane video image.As shown in Figure 8,801 is plane coordinates acquiring unit to three-dimensional coordinate determination module, for obtaining the plane coordinates information of target from the plane video image obtained.802 is three-dimensional coordinate extraction unit, and the plane coordinates information of the target utilizing plane coordinates acquiring unit 801 to obtain is changed coordinate, obtains the three-dimensional coordinate information of target.Can be 3D coordinate by above formula (6) realizes 2D Coordinate Conversion.

Such device, can be 2D video camera in video acquisition module, when the video image of acquisition is single 2D video image, realize the three-dimensional coordinate information obtaining target, carried out the judgement of event by three-dimensional coordinate information, thus avoid the event erroneous judgement because transparent effect causes.

In one embodiment, as shown in Figure 9,901 is frame contrast unit to three-dimensional coordinate determination module, for contrasting the plane video image of successive frame, or plane video image and background image being contrasted, obtaining pixel or the point group of change.902 is target determination unit, for extracting target from the change point or point group of the acquisition of frame contrast unit 901.903 is plane coordinates acquiring unit, for obtaining the plane coordinates information of target from plane video image, 904 is three-dimensional coordinate extraction unit, for obtaining the three-dimensional coordinate information of target according to the plane coordinates information of target and the operation relation of plane coordinates and three-dimensional coordinate.

Such device, can according to monoplane video image, the point changed by the contrast acquisition of frame or point group are as target, improve the sensitivity of monitoring, reduce the probability omitted, achieve the three-dimensional coordinate information obtaining target, carried out the judgement of event by three-dimensional coordinate information, thus avoid the event erroneous judgement because transparent effect causes.

In one embodiment, the video image that video acquisition module obtains is the plane video image of multiple the same area taken from different camera site.As shown in Figure 8,801 is 3D reconstruction unit to three-dimensional coordinate determination module, for the plane video image of multiple the same area taken from different camera site is carried out 3D reconstruction, obtains 3D video image; 802 is three-dimensional coordinate extraction unit, and the 3D video image for obtaining according to 3D reconstruction unit 801 obtains the three-dimensional coordinate information of target.

Such device, the multiple plane video image from diverse location shooting of same guarded region can be obtained, multiple plane video image is carried out 3D and rebuild acquisition 3D video image, the determination of target three-dimensional coordinate is carried out according to 3D video image, thus three-dimensional coordinate information can be used to carry out the extraction of event, when avoiding using two-dimensional coordinate to carry out Event Distillation, due to the event erroneous judgement that transparent effect causes.

In one embodiment, as shown in Figure 9,901 is 3D reconstruction unit to three-dimensional coordinate determination module, multiple plane video image from diverse location shooting is carried out 3D reconstruction, obtains 3D video image.902 is frame contrast unit, for being contrasted by the front and back frame of the 3D video image after reconstruction, or 3D video image and background image is contrasted, the point that acquisition three-dimensional coordinate information or color information change or point group.903 is Objective extraction unit, in the frame contrast change point determined of unit 902 or point group, extract target.904 is three-dimensional coordinate extraction unit, rebuilds the 3D video image obtained according to 3D reconstruction unit 901, obtains the three-dimensional coordinate information of target.

Such device, when can be the plane video image of multiple different camera site at the image obtained, be rebuild by 3D and obtain 3D video image, in 3D video image, obtain change point or point group as target, improve the sensitivity of monitoring, reduce the probability omitted.Three-dimensional coordinate information can be used to carry out the extraction of event, when avoiding using two-dimensional coordinate to carry out Event Distillation, due to the event erroneous judgement that transparent effect causes.

In one embodiment, video acquisition module is distance sensitive device, and the video image of acquisition is depth image, then three-dimensional coordinate determination module obtains the three-dimensional coordinate information of target according to depth image.Such device, the three-dimensional coordinate information of target is obtained by depth image, position relationship both judging based on the three-dimensional coordinate information of virtual door and target, thus the event of extraction occurs, effectively prevent the event caused due to transparent effect in two dimensional image to judge by accident, improve the accuracy that event judges.

In one embodiment, as shown in Figure 7, frame contrast unit 701 contrasts the depth image of successive frame, or depth image and background depth image is contrasted, the pixel that acquisition depth information changes or point group; Objective extraction unit 702 extracts target from change point or point group; Three-dimensional coordinate extraction unit 703, according to depth image, obtains the three-dimensional coordinate information of target.

Such device, according to the change detection change point of depth information in depth image or point group as target, can improve the sensitivity of monitoring, reduces the probability omitted.Three-dimensional coordinate information can be used to carry out the extraction of event, when avoiding using two-dimensional coordinate to carry out Event Distillation, due to the event erroneous judgement that transparent effect causes.

In one embodiment, video acquisition module is the equipment comprising distance sensitive device and 2D video camera, and as Kinect, PMD CARMERA or MESA SR etc., the image of acquisition is depth image and plane video image.The successive frame of plane video image can be contrasted, or contrast with background image, determine the point that color information changes or point group, from change point or point group, extract target.The three-dimensional coordinate information of target is obtained according to depth image.

Such device, plane video image that is higher according to definition, that have color information can obtain target, prevent because depth information change not obvious causing is mistaken for noise, thus reduce the probability leaking captured target, make monitoring more reliable.

In one embodiment, three-dimensional coordinate determination module also comprises track determining unit, track determining unit by the movement locus of the positional information determination target of record object in multi-frame video image, then by the three-dimensional coordinate information of the movement locus of three-dimensional coordinate extraction unit determination target.

Extract the generation of event according to the three-dimensional coordinate information of the movement locus of target and virtual door, the event extracted can comprise: from outside to inside through virtual door, from inside to outside through virtual door, from outside to inside motion and through virtual door, from inside to outside motion and not through virtual door.By such device, can continuing to monitor target be realized, improve the accuracy of Event Distillation.

In one embodiment, the schematic diagram of video monitoring apparatus of the present invention as shown in Figure 10.Wherein, 1001,1002 and 1003 are respectively video acquisition module, three-dimensional coordinate determination module and Event Distillation module, and its course of work is similar to Fig. 6.Video monitoring apparatus can also comprise alarm module 1005.When Event Distillation module extracts the event of predetermined needs warning, send alarm signal, event information to alarm module 1005 trigger alarm.Such device, can realize clearly providing warning message to user, helps user to process event in time.Make a reservation for need the event of reporting to the police to comprise and be positioned at virtual door, be in virtual door region, pass through virtual door from outside to inside and/or from inside to outside pass through virtual door etc.

In one embodiment, video monitoring apparatus also comprises target type analysis module 1004.The video image coupling that target type analysis module 1004 is obtained by video acquisition module obtains target type, and target type can comprise people, animal and/or car.By images match, obtain the type of target, thus enrich Event Distillation information.In one embodiment, alarm module 1005 can obtain target type, provides Target type information to user.By such device, the target type needing to report to the police can be selected, reduce the workload of user.

In one embodiment, Event Distillation module adds up the frame number of the event generation of extracting, and when the event extracted reaches predetermined frame number, just determines that event occurs.Such device can prevent erroneous judgement, filters the false-alarm of a part.

Finally should be noted that: above embodiment is only in order to illustrate that technical scheme of the present invention is not intended to limit; Although with reference to preferred embodiment to invention has been detailed description, those of ordinary skill in the field are to be understood that: still can modify to the specific embodiment of the present invention or carry out equivalent replacement to portion of techniques feature; And not departing from the spirit of technical solution of the present invention, it all should be encompassed in the middle of the technical scheme scope of request of the present invention protection.

Claims (30)

1. a video monitoring method, is characterized in that:

Obtain video image;

The three-dimensional coordinate information of target is obtained according to described video image;

Position relationship based on described target and virtual door extracts event generation, and wherein, described virtual door comprises three-dimensional coordinate information.

2. method according to claim 1, is characterized in that, described virtual door is door region perpendicular to the ground, and the intersection on described virtual door and ground is straight line, line segment or broken line.

3. method according to claim 1, is characterized in that, the described three-dimensional coordinate information according to described video image acquisition target comprises:

Video image described in contrast successive frame, or described video image and background image are contrasted, obtain the change point in described video image or point group;

From described change point or point group, extraction point or point group are as target;

Determine the three-dimensional coordinate information of described target.

4. method according to claim 1, is characterized in that, described video image is plane video image;

The described three-dimensional coordinate information according to described video image acquisition target comprises:

The plane coordinates information of described target is obtained by described plane video image;

Carry out 3D reconstruction according to described plane coordinates information by 3D algorithm for reconstructing, obtain the described three-dimensional coordinate information of described target.

5. method according to claim 4, is characterized in that, the equipment obtaining described video image comprises a 2D video camera.

6. method according to claim 1, is characterized in that, described video image comprises the plane video image of multiple different camera site;

The described three-dimensional coordinate information according to described video image acquisition target comprises:

Multiple described plane video image is carried out 3D and rebuild acquisition 3D video image;

The three-dimensional coordinate information of described target is obtained according to described 3D video image.

7. method according to claim 6, is characterized in that, the equipment obtaining described video image comprises two or more 2D video camera or the 3D video camera based on binocular vision.

8. method according to claim 1, is characterized in that, described video image is depth image;

The described three-dimensional coordinate information according to described video image acquisition target is obtain the three-dimensional coordinate information of target according to described depth image.

9. method according to claim 8, is characterized in that, the equipment obtaining described depth image comprises distance sensitive device or 3D video camera.

10. method according to claim 1, is characterized in that, the described position relationship extraction event based on described target and virtual door occurs as:

Extract event according to the position relationship of the horizontal coordinate information in the three-dimensional coordinate information of described target and virtual door to occur, wherein, described virtual door comprises the horizontal coordinate information in three-dimensional coordinate.

11. methods according to claim 1, is characterized in that, also comprise:

According to multiframe, video image determines the movement locus of described target;

Determine the three-dimensional coordinate information of the described movement locus of described target;

Extract event based on the described movement locus of described target and the position relationship of described virtual door to occur.

12. methods according to claim 1, it is characterized in that, described event comprise be positioned at virtual door, be positioned at virtual outdoors, be positioned at virtual door region, from outside to inside through virtual door, from inside to outside through virtual door, from outside to inside motion and through virtual door and/or from inside to outside motion and not through virtual door.

13. methods according to claim 1, is characterized in that, the type of described target comprises people, animal and/or car.

14. methods according to claim 1, is characterized in that, also comprise, if extract scheduled event, then send warning message, and described warning message includes invades positional information and/or intrusion directional information.

15. methods according to claim 1, is characterized in that, the position relationship extraction event based on described target and virtual door comprises, the continuous frame number of statistical phenomeon, and when described frame number is greater than predetermined warning frame number, decision event occurs.

16. 1 kinds of video monitoring apparatus, is characterized in that, comprising:

Video acquisition module, for obtaining video image;

Three-dimensional coordinate determination module, for obtaining the three-dimensional coordinate information of described target according to described video image;

Event Distillation module, occur for extracting event based on the position relationship of described target and virtual door, wherein, described virtual door comprises three-dimensional coordinate information.

17. devices according to claim 16, is characterized in that, described virtual door is door region perpendicular to the ground, and the intersection on described virtual door and ground is straight line, line segment or broken line.

18. devices according to claim 17, is characterized in that, described Target Acquisition module comprises:

Frame contrast unit, for contrasting video image described in successive frame, or contrasting described video image and background image, obtaining the change point in described video image or point group;

Target determination unit, for extraction point or point group from described change point or point group as target;

Three-dimensional coordinate extraction unit, for determining the three-dimensional coordinate information of described target.

19. devices according to claim 16, is characterized in that, described video image is plane video image;

Described three-dimensional coordinate determination module comprises:

Plane coordinates acquiring unit, for obtaining the plane coordinates information of described target by described plane video image;

Three-dimensional coordinate extraction unit, for carrying out 3D reconstruction according to described plane coordinates information by 3D algorithm for reconstructing, obtains the described three-dimensional coordinate information of described target.

20. devices according to claim 19, is characterized in that, described video acquisition module comprises a 2D video camera.

21. devices according to claim 16, is characterized in that, described video image comprises the plane video image of multiple different camera site;

Described three-dimensional coordinate determination module comprises:

3D reconstruction unit, rebuilds acquisition 3D video image for multiple described plane video image being carried out 3D;

Three-dimensional coordinate extraction unit, for according to described 3D video image, obtains the three-dimensional coordinate information of target.

22. devices according to claim 21, is characterized in that, described video acquisition module comprises two or more 2D video camera or the 3D video camera based on binocular vision.

23. devices according to claim 16, is characterized in that, described video image is depth image;

Described three-dimensional coordinate determination module, for obtaining the three-dimensional coordinate information of target according to described depth image.

24. devices according to claim 23, is characterized in that, described video acquisition module comprises distance sensitive device or 3D video camera.

25. devices according to claim 16, it is characterized in that, described Event Distillation module occurs for extracting event according to the position relationship of the horizontal coordinate information in the three-dimensional coordinate information of described target and virtual door, and wherein, described virtual door comprises the horizontal coordinate information in three-dimensional coordinate.

26. devices according to claim 16, is characterized in that,

Described three-dimensional coordinate determination module also comprises:

Track determining unit, for determining the movement locus of target described in described video image;

Described three-dimensional coordinate extraction unit, also for determining the three-dimensional coordinate information of the described movement locus of described target;

Described Event Distillation module, also occurs for extracting event based on the described movement locus of described target and the position relationship of described virtual door.

27. devices according to claim 16, it is characterized in that, described event comprise be positioned at virtual door, be positioned at virtual outdoors, be positioned at virtual door region, from outside to inside through virtual door, from inside to outside through virtual door, from outside to inside motion and through virtual door and/or from inside to outside motion and not through virtual door.

28. devices according to claim 16, is characterized in that, also comprise target type analysis module, and for evaluating objects type, the type of described target comprises people, animal and/or car.

29. devices according to claim 16, is characterized in that, also comprise alarm module, and for sending warning message according to the scheduled event extracted, described warning message includes invades positional information and/or intrusion directional information.

30. devices according to claim 16, is characterized in that, described Event Distillation module is also for the continuous frame number of statistical phenomeon, and when described frame number is greater than predetermined warning frame number, decision event occurs.