CN101426128B - Detection system and method for stolen and lost packet - Google Patents
Detection system and method for stolen and lost packet Download PDFInfo
- Publication number
- CN101426128B CN101426128B CN200710165489A CN200710165489A CN101426128B CN 101426128 B CN101426128 B CN 101426128B CN 200710165489 A CN200710165489 A CN 200710165489A CN 200710165489 A CN200710165489 A CN 200710165489A CN 101426128 B CN101426128 B CN 101426128B
- Authority
- CN
- China
- Prior art keywords
- pixel
- model
- region
- interest
- color
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Expired - Fee Related
Links
Images
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N7/00—Television systems
- H04N7/18—Closed-circuit television [CCTV] systems, i.e. systems in which the video signal is not broadcast
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/40—Image enhancement or restoration by the use of histogram techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/20—Analysis of motion
- G06T7/269—Analysis of motion using gradient-based methods
Abstract
The present invention provides a system for monitoring the theft detection system and the loss of the parcel and its detection methods, including theft and the loss of the parcel described detection system including: image capture devices, surveillance of concern for the system to capture the scene and the new image; scene module records the scene using the scene model of history; scene comparison and model update module, in accordance with the pixel-level images and scenes of the new model compared to the results based on comparison to update the scene model, and use be recorded in pixel time model information to determine the status of each pixel, which was concerned about the image pixel tags to structure, in which the state of each pixel as the background, the outlook or concerns; concerned about the regional extraction module, the use of temporal and spatial information extraction instructions theft or loss occurred in the events package concerned about the region and the output of the concern about the region; display units, display instructions theft incidents have taken place in or loss of parcels of regional concern.
Description
Technical field
The present invention relates to a kind of surveillance, more particularly, relate to a kind of pilferage of in surveillance, using and lose parcel detection system and a kind of detection method based on improved change detecting method.
Background technology
In order to guarantee safety, surveillance camera is installed in the public place widely.Whether yet needing some extra labours to check has responsive incident to take place.At present, the heat subject in the field of surveillance systems be that research a kind ofly can detect, the intelligent camera of the action of tracking and analyst and other object.
In the technical scheme in early days, the user all can preserved and reported to the object that is changed or move further to check.But this scheme still needs the great amount of manpower resource, and its shortcoming in public places, and is particularly evident like shopping center or recreation center, because the people that walk about are almost all arranged in public places always.
Therefore, some researchers are studying the important relatively incident that detects how automatically, as stealing and losing and wrap up.Stealing and losing parcel is two kinds of the most responsive security incidents.The serious problem of stealing that yes, and when we during in the face of terrified bombing raid, lose to wrap up and also become extremely important.
But stealing and losing the detection of wrapping up is very difficult problem, especially in the up and-down crowd scene of much human.In addition, people and object can overlap each other, block in keeping watch on scene and repeat and occur.Following the trail of everyone almost is impossible with the motion of object.In real-time monitoring system, processing speed must just can work online very soon.
In the prior art, common way is to analyze the movement locus of each object or person in the scene.The 5th, 969, No. 755 United States Patent (USP)s, the 6th, 424, No. 370 United States Patent (USP)s and the 6th, 628, No. 835 U.S. Patent Publication the embodiment of said method.Disclosed method passes through to follow the trail of the motion of each object or person in the scene in above-mentioned patent, and indicates pilferage incident or loss package event to realize.The pilferage incident is by three order event indications, and these three incidents comprise: the people is near object; People and object move together; People and object leave scene together.Lose package event by two order event indications, comprising: people and object get into scene; The people leaves and object stays.In addition, the 6th, 628, No. 835 United States Patent (USP)s for example define also that the people is pacing up and down and such incidents such as two people clash.But generally speaking, above-mentioned patent belongs to same type.
The affair analytical method that is the basis with the movement locus detection depends on the cutting apart and tracking result of object excessively, and this method is not often very reliably in surveillance.Sometimes, a people can be considered to several isolated parts, can not judge the situation that two people move together sometimes.Under these complicated situations, it is impossible fully for a long time everyone being followed the trail of with object reliably.
In addition, the 6th, 819, No. 353 United States Patent (USP)s disclose the pilferage detection method of a kind of dependence based on the change-detection of traditional mixing Gaussian distribution.In change-detection, the historical color and the weights of each pixel are stored in the background model.Weights also mean the possibility that color occurs in this locations of pixels.For background color, its weights are higher; For foreground color, its weights are lower.
Obviously, this pilferage detection method that depends on change-detection purely is inappropriate for crowded places, keeps watch on like shopping center or recreation center, because in scene, there is too many mobile object.
Summary of the invention
Therefore, content of the present invention has been proposed to solve the above-mentioned and/or other problem that occurs in the prior art.One side of the present invention provides a kind of and has improved pilferage and the loss parcel detection system and the detection method thereof of stealing and losing the accuracy of parcel detection through removing some false alarms.
One side of the present invention is to provide a kind of and can wraps up detection system and a kind of pilferage thereof and lose the parcel detection method pilferage and loss that pilferage incident and loss package event are distinguished.
Above-mentioned and/or others of the present invention and effectiveness can be through providing a kind of pilferage and losing the parcel detection system and realize that this system comprises: image capture apparatus is used to keep watch on the scene of concern and is the system acquisition new images; Scene module uses the model of place that is made up of pixel model to write down the history of scene, and wherein, each pixel model is background model, concern model or foreground model; Scene comparison and model modification module; According to pixel scale new images and model of place are compared; Upgrade model of place according to comparing result, and utilize the temporal information that is documented in the pixel model to judge the state of each pixel, thereby obtain concerned pixel with the structure marking image; Wherein, the state of each pixel is background, prospect or concern; The region-of-interest extraction module utilizes space time information to extract indication and has taken place to steal or lose the region-of-interest of package event and export this region-of-interest; Display unit shows that the region-of-interest of pilferage incident or loss package event has taken place in indication.
Above-mentioned and/or others of the present invention and effectiveness can be through providing a kind of pilferage and losing the parcel detection method and realize that this method may further comprise the steps: catch new images; According to pixel scale new images and model of place are compared, and upgrade each pixel model of model of place according to comparing result, wherein, each pixel model is background model, pay close attention to model or background model; The temporal information that utilization is documented in the pixel model is judged the state of each pixel, thereby obtains concerned pixel with the structure marking image, and wherein, the state of each pixel is background, prospect or concern; Extract the region-of-interest that indication the pilferage incident has taken place or lost package event according to marking image, and export this region-of-interest; Show that the region-of-interest of pilferage time or loss package event has taken place in indication.
Description of drawings
From below in conjunction with the description of accompanying drawing to exemplary embodiment, above-mentioned and others of the present invention and advantage will become clear and be more readily understood, wherein:
Fig. 1 is the block diagram that illustrates according to the structure of pilferage of the present invention and loss parcel detection system;
Fig. 2 is the view of structure that the pixel model of a pixel is shown;
Fig. 3 is the flow chart that scene comparison and model modification module execution scene comparison and model modification is shown and exports the process of region-of-interest;
Fig. 4 is the view that the method for calculating difference image is shown;
Fig. 5 is the sketch map that the conversion between the color state is shown;
Fig. 6 illustrates the flow chart that the region-of-interest extraction module extracts the method for region-of-interest;
Fig. 7 A and Fig. 8 B illustrate the flow chart of stealing incident and the method for losing package event according to the identification of two embodiment of the present invention;
Fig. 8 be illustrate by pilferage of the present invention and lose the parcel detection system and method to losing package event and wrapping the picture of the result of experiment that two incidents carry out steathily;
Fig. 9 be illustrate by pilferage of the present invention and lose the parcel detection system and method to the picture of the result of experiment of stealing book and carrying out;
Figure 10 be illustrate by pilferage of the present invention and lose the parcel detection system and method to the picture of the result of experiment of stealing bag and carrying out.
Embodiment
To specify embodiments of the invention now, its example is shown in the drawings, and wherein, identical label is indicated components identical all the time.Below illustrate and describe embodiment to explain the present invention.
Fig. 1 is the block diagram that illustrates according to the structure of pilferage of the present invention and loss parcel detection system.
As shown in Figure 1, according to pilferage of the present invention with lose the parcel detection system and comprise: image capture device 10, this device is fixed, the new images 100 that is used to keep watch on the scene of concern and is used to detect for the detection system seizure; Scene logging modle 20 uses model of place 200 to write down the history of scenes; Scene comparison and model modification module 30, this module will be compared by image capture device 10 new images 100 of catching and the historical information that is recorded in the model of place 200, and upgrade model of place 200 according to result relatively; Region-of-interest extraction module 40, the zone that utilizes space time information to extract and examine generation pilferage incident or lose package event, and output region-of-interest; Event classification module 50 is to stealing and losing these two kinds of incidents of parcel and discern; The display unit (not shown) is used to show the result of detection.
Fig. 2 is the view of structure that the pixel model of a pixel in the model of place 200 is shown.
To explain the structure of model of place 200 below with reference to Fig. 2.Model of place 200 is made up of the plurality of pixels model 210 of each pixel.In model of place 200, use mixed Gaussian distribution (MoG) to represent that the recent color of each pixel is historical.
∑w
i·N(u
i，σ
i)
Wherein, w
iThe 213rd, each single Gauss's weights, u
i211 is each single Gauss's center, expression average color, σ
iThe 212nd, each single Gauss's variance.
If certain color often occurs, this color has bigger weights so.For the object that moves that the very short time on pixel, only occurs, the color of this object will have very little weights so.
Each color of pixel can be a several values.For example, under static background, pixel can have the color of background color or some foreground objects.And under dynamic background, pixel can have several background colors or object color.Each color is stored in the different possible pixel model 210 in the model of place.Therefore, each pixel can have one or more pixel model 210.
For a pixel, before stealing generation, this color of pixel is a background color, and after stealing generation, therefore this color of pixel flip-flop thinks that change has also taken place this pixel.After the short time, the weights of new color will increase, and greatly to surpassing threshold value, this color becomes background color then up to it.Therefore, in the present invention, the duration, (that is, d) pixel model of 214 records became after the background model duration at the background model state.If a pixel model is common foreground model, then the duration of this model is set as 0, and the duration of this model is set as 0.If after each the renewal, a model remains background model, then the increase by 1 of the duration of this model.In the present invention, the weights according to value and each pixel model of duration are divided into background model, foreground model and concern model with pixel model.Type how to confirm pixel model will hereinafter be described.
Fig. 3 is the flow chart that figure scene comparison and model modification module execution scene comparison and model modification is shown and exports the process of region-of-interest.
To carry out scene comparison and model modification and export the process of region-of-interest with reference to Fig. 3 illustrated in detail scene comparison and update module below.
At first, in step 310, in given new images, find Matching Model for pixel.
With the pixel of input and each the Gauss center μ in the model of place
iImmediate Gauss is found in 211 contrasts.If distance is less than σ
i* Thr
Var, then this pixel and this Model Matching (Thr in the present invention,
Var=3)
For example, as shown in Figure 5, the 210th, the Gauss of background model or concern model.The 220th, the Gauss of prospect.The 101st, the input pixel, it is not in any background Gauss's scope.The 102nd, the input pixel, itself and model 210 mate.
Then, in step 320, upgrade based on following regular execution model.
If new pixel belongs to a Gauss, then upgrade its center 211 and variance 212.Its weights 213 are increased a bit, another Gauss's weights 213 are reduced a bit.If new pixel does not belong to any Gauss, then deletion has minimum weights W
iOld Gauss, and to increase with this new color of pixel be the new Gauss at center.For background model with pay close attention to model, its duration is increased by 1, for other model, their duration 214 is reset to 0.
Then, in step 330, the state of pixel is confirmed as prospect, background or concern.
At first, be foreground model and background model the possible threshold value of devising a stratagem really: suppose that each pixel should be time of background to account for the percentage of whole time to be at least Thr
Bg(for example, in the present invention, Thr
Bg=85%).Then, the weights of plurality of pixels model that utilize each pixel are with these categories of model.To have weights addition, surpass Thr up to it than several models before the high weight
Bg, then these models are called as background model (comprise and pay close attention to model), and other model is a foreground model.If the duration d of model satisfies following conditions, then this model is for paying close attention to model, and this pixel is a concerned pixel.
0＜d＜Thr
Dur(Thr in the present invention,
Dur=100)
The meaning of above-mentioned condition is: after stealing or losing the package event generation, the pixel that the zone of above-mentioned incident takes place will become background from prospect gradually.Therefore, in the incipient stage of background, the Matching Model of pixel should be considered to steal event context or lose the parcel prospect.Briefly, in the following description, " pilferage background " or " losing the parcel prospect " model will be called as " concern model ".
In step 340, with the combinations of states of all pixels to form marking image.In this image, the value of each pixel can be prospect, background or concern.
In model of place, color of pixel can be corresponding to above-mentioned three kinds of states, prospect, background or concern.Shown the conversion of 3 kinds of models among Fig. 5.For a new model, because its weights are very little, so this model is a foreground model.When the weights of this new model increase, this model will be in the concern state in a period of time, become background model then; If background model does not occur in image for a long time, its weights will reduce, and this model will become foreground model.
Fig. 6 illustrates the flow chart that the region-of-interest extraction module extracts the method for region-of-interest;
Region-of-interest extraction module 40 comprises concerned pixel verification module, region extraction module and region-of-interest verification module.In region-of-interest is verified module 40, extract region-of-interest and verify according to pixel scale and regional rank.
As shown in Figure 6, the method that region-of-interest extracts comprises concerned pixel verification step 410, extracted region step 420 and region-of-interest verification step 430.Verify in the step 410 at concerned pixel, for real pilferage and loss parcel zone, after incident took place, the color of pixel in this zone should be constant.Even in this process, this zone is blocked by other foreground object, and its color also should be constant in the visible time period.Therefore, carry out concerned pixel identification,, then it is removed from concerned pixel if the color of a few frame interior pixels formerly is non-constant based on constant color.
In extracted region step 420, through using the coupling part marking image is analyzed, can extract one group of connection and steal pixel.Independently the zonule abandons as false alarm.
Verify in the step 430 at region-of-interest,, then it is reported as and steals the zone if should in a period of time, not change in the zone.Otherwise this is a false alarm.This verification is useful in scenario: if an object only has very little motion, then the inside of this object will be considered to take place steal because real pilferage keeps along with the motion of object changing.Through this check method, can successfully remove this false alarm.
Method according to 50 pairs of event classification modules of the present invention are stolen and the loss parcel is classified has two kinds of execution modes.
For stealing and losing package event, characteristics of image is different.After stealing, reformed zone should be a background, so it is similar to the near zone that also is assumed that background.After losing parcel, reformed zone is a parcel, so it is compared with background on every side and has the pictures different characteristic.According to above-mentioned principle, introduce stealing the zone and losing two kinds of execution modes that the parcel zone is distinguished at this.
Fig. 7 A show a kind of based on the characteristics of image matching method to stealing and losing the execution mode that the parcel zone is discerned.Shown in Fig. 7 A, this method comprises: in step 510, to the characteristic in the zone paid close attention to, like average color, color histogram, assess; In step 520, the same characteristic features of background pixel is on every side assessed; In step 530, these characteristics of image are compared; According to result calculated region-of-interest is confirmed as pilferage (step 540) or lost parcel (step 550).
In the pilferage incident, the color in the zone of being blocked by object is regional similar with on every side probably, therefore on the profile of object, do not have tangible gray scale.On the contrary, for losing package event, tangible border is arranged around the profile of object.Therefore, through the length and the quantity of computation bound pixel, can discern the pilferage incident and lose package event.
Fig. 7 B show a kind of based on the gradation of image method to stealing and losing the execution mode that the parcel zone is discerned.Shown in Fig. 7 B, this method comprises: in step 560, and the borderline average gradient of zoning; In step 570, average gradient and the threshold value calculated are compared; If average gradient surpasses threshold value, then region-of-interest is identified as and loses parcel zone (step 550), otherwise region-of-interest is identified as pilferage zone (step 540).
Fig. 8 be illustrate by pilferage according to the present invention and lose the parcel detection system and method to losing parcel and wrapping the result of experiment picture that two incidents are carried out steathily.As shown in Figure 8, can confirm according to image sequence, " losing parcel " incident has taken place earlier on an object, follow, " pilferage " incident has taken place again, and regional can the indication by the wire frame of different colours wrapped up with losing in the pilferage zone on another object.
Fig. 9 illustrates by pilferage according to the present invention and loses the picture of parcel detection system to the result of experiment of stealing book and carrying out.As shown in Figure 9, according to image sequence, can find out that a people has stolen a book.
Figure 10 be illustrate by pilferage according to the present invention and lose the parcel detection system and method to the picture of the result of experiment of stealing bag and carrying out.According to image sequence, can find out that a people has stolen a bag.In this experiment,, and stopped by several foreground objects that in this process this incident is still successfully detected though the object of being stolen is very little.
Therefore; According to pilferage of the present invention with lose the parcel detection system and pilferage and loss parcel detection method can improve and steal and lose reliability and the accuracy that parcel detects through removing some false alarms, and can to the pilferage incident with lose package event and distinguish.
Although shown and described exemplary embodiment of the present invention,, it should be appreciated by those skilled in the art, under the situation that does not break away from the principle of the present invention that limits claim and spirit, can carry out various modifications to these exemplary embodiments.
Claims (25)
1. steal and lose the parcel detection system for one kind, comprising:
Image capture apparatus is used to keep watch on the scene of concern and is the system acquisition new images;
Scene module uses the model of place that is made up of pixel model to write down the history of scene, and wherein, each pixel model is background model, concern model or foreground model;
Scene comparison and model modification module; According to pixel scale new images and model of place are compared; Upgrade model of place according to comparing result, and utilize the temporal information that is documented in the pixel model to judge the state of each pixel, thereby obtain concerned pixel with the structure marking image; Wherein, the state of each pixel is background, prospect or concern;
The region-of-interest extraction module utilizes space time information to extract indication and has taken place to steal or lose the region-of-interest of package event and export this region-of-interest;
Display unit shows that the region-of-interest of pilferage incident or loss package event has taken place in indication.
2. the system of claim 1 also comprises the event classification module, is used to distinguish the pilferage incident and loses package event.
3. system as claimed in claim 2; Wherein, The event classification module is assessed the characteristics of image of region-of-interest; To around the identical characteristics of image of background pixel assess, calculate the difference between this two parts characteristics of image, and according to the difference of calculating region-of-interest confirmed as and to be stolen the zone or to lose the parcel zone.
4. system as claimed in claim 3, wherein, characteristics of image is average color or color histogram.
5. system as claimed in claim 2; Wherein, the event classification module is calculated the average gradient on the border of region-of-interest, and average gradient and the threshold value calculated are compared; According to result relatively region-of-interest is confirmed as the pilferage zone or lost the parcel zone; If average gradient greater than threshold value, is then confirmed as region-of-interest and lost the parcel zone, otherwise for stealing the zone.
6. the system of claim 1; Wherein, the region-of-interest extraction module also comprises pixel verification module, and this pixel is verified module concerned pixel is verified; If non-constant in the color of concerned pixel several frames formerly, then this pixel is removed from concerned pixel.
7. the system of claim 1, wherein, the region-of-interest extraction module comprises that also region-of-interest verifies module; This region-of-interest is verified module the region-of-interest that extracts is verified, if region-of-interest is zonule independently, then this region-of-interest is removed as false alarm; If a region-of-interest did not change in a period of time; Then this region-of-interest is reported as and steals the zone, otherwise, this region-of-interest is removed as false alarm.
8. the system of claim 1, wherein, model of place is the set of pixel model, and distributes with mixed Gaussian and to describe each color of pixel and distribute:
∑w
i·N(u
i，σ
i)
Wherein, w
iThe weights of representing the color that each is possible, N (u
i, σ
i) function of the color that expresses possibility, u
iBe each single Gauss's center, represent the average color of the color that each is possible, σ
iBe variance.
9. system as claimed in claim 8; Wherein, scene comparison and model modification module confirm according to new images whether each color of pixel in the new images belongs to a Gauss in the mixed Gaussian distribution, if said color of pixel belongs to a Gauss in the mixed Gaussian distribution; Then upgrade this Gauss's center and variance; Increase this Gauss's weights and reduce other Gauss's weights, if said color of pixel does not belong to any Gauss, then deletion has the Gauss of minimum weights; And to increase with said color of pixel be the new Gauss at center, thereby realize the renewal to model of place.
10. system as claimed in claim 9, wherein, temporal information is that pixel model becomes the background model duration afterwards.
11. system as claimed in claim 10, wherein, scene comparison and model modification module are divided into foreground model or background model according to the weights and the predetermined first threshold of pixel model with pixel model; If pixel model is a foreground model behind the image update; Then the duration with this pixel model is made as 0, if pixel model is a background model behind the image update, then the duration with this pixel model increases a chronomere; If the duration of pixel model is greater than 0 and less than predetermined second threshold value; Then pixel is confirmed as concerned pixel, simultaneously, the pixel model corresponding with color of pixel becomes the concern model from background model.
12. system as claimed in claim 11; Wherein, Scene comparison and model modification module add the higher several weights in the pixel model together and surpass first threshold up to its result; To have than the pixel model of high weight and confirm as background model, other pixel model will be confirmed as foreground model.
13. steal and lose the parcel detection method, said method comprising the steps of for one kind:
Catch new images;
According to pixel scale new images and model of place are compared, and upgrade each pixel model of model of place according to comparing result, wherein, each pixel model is background model, pay close attention to model or background model;
The temporal information that utilization is documented in the pixel model is judged the state of each pixel, thereby obtains concerned pixel with the structure marking image, and wherein, the state of each pixel is background, prospect or concern;
Extract the region-of-interest that indication the pilferage incident has taken place or lost package event according to marking image, and export this region-of-interest;
Show that the region-of-interest of pilferage time or loss package event has taken place in indication.
14. method as claimed in claim 13 also comprises the step that pilferage incident and loss package event are discerned.
15. method as claimed in claim 14, wherein, identification pilferage incident comprises with the step of losing package event:
Characteristics of image to region-of-interest is assessed;
Same characteristic features to background pixel is on every side assessed;
Calculate the difference between this two parts characteristics of image;
According to the difference of calculating region-of-interest is confirmed as the pilferage zone or lost the parcel zone.
16. method as claimed in claim 15, wherein, characteristics of image is average color or color histogram.
17. method as claimed in claim 14, wherein, the step that pilferage incident and loss package event are discerned comprises:
Calculate the average gradient on the border of region-of-interest;
Average gradient and the threshold value calculated are compared;
According to relatively result region-of-interest is confirmed as and to be stolen the zone or to lose the parcel zone, if difference greater than threshold value, then region-of-interest is confirmed as and loses the parcel zone, on the contrary for stealing the zone.
18. method as claimed in claim 13, wherein, extraction step also comprises:
According to pixel scale concerned pixel is verified, if non-constant in the color of concerned pixel several frames formerly, then it is removed from concerned pixel with this pixel.
19. method as claimed in claim 13, wherein, extraction step also comprises:
Region-of-interest to extracting is verified, if region-of-interest is zonule independently, then with this region-of-interest removal.
20. method as claimed in claim 13, wherein, the step that the region-of-interest that extracts is verified also comprises:
Judge whether the region-of-interest that is extracted variation has taken place, if this region-of-interest does not change, then this region-of-interest is reported as the zone of pilferage in a period of time, otherwise, this region-of-interest is removed as false alarm.
21. method as claimed in claim 13, wherein, model of place is the set of pixel model, and distributes with mixed Gaussian and to describe each color of pixel and distribute:
∑w
i·N(u
i，σ
i)，
Wherein, w
iThe weights of representing the color that each is possible, N (u
i, σ
i) function of the color that expresses possibility, u
iBe each single Gauss's center, represent the average color σ of the color that each is possible
iBe variance.
22. method as claimed in claim 21, wherein, the step of upgrading model of place according to new images comprises:
Confirm whether each color of pixel in the new images belongs to a Gauss in the mixed Gaussian distribution;
If said color of pixel belongs to the Gauss of mixed Gaussian in distributing, then upgrade this Gauss's center and variance, increase this Gauss's weights and reduce other Gauss's weights;
If said color of pixel does not belong to any Gauss, then deletion has a Gauss of minimum weights, and to increase with said color of pixel be the new Gauss at center, thereby realizes the renewal to model of place.
23. method as claimed in claim 22, wherein, temporal information is that pixel model becomes the background model duration afterwards.
24. method as claimed in claim 23 wherein, judges that the method for the state of each pixel comprises:
Weights and predetermined first threshold according to pixel model are divided into foreground model or background model with pixel model;
If pixel model is a foreground model behind the image update, then the duration with this pixel model is made as 0;
If pixel model still is a background model behind the image update, then this pixel model duration increases a chronomere;
If the duration of pixel model is then confirmed as concerned pixel with pixel greater than 0 and less than predetermined second threshold value, simultaneously, the pixel model corresponding with color of pixel becomes the concern model from background model.
25. method as claimed in claim 24, wherein, the method for distinguishing foreground model and background model comprises:
Higher several weights in the pixel model are added together, surpass first threshold up to its result, the pixel model that then has than high weight is a background model, and other pixel model is a foreground model.
Priority Applications (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN200710165489A CN101426128B (en) | 2007-10-30 | 2007-10-30 | Detection system and method for stolen and lost packet |
| KR1020080011392A KR20090044957A (en) | 2007-10-30 | 2008-02-04 | Theft and left baggage survellance system and meothod thereof |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN200710165489A CN101426128B (en) | 2007-10-30 | 2007-10-30 | Detection system and method for stolen and lost packet |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN101426128A CN101426128A (en) | 2009-05-06 |
| CN101426128B true CN101426128B (en) | 2012-10-03 |
Family
ID=40616421
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN200710165489A Expired - Fee Related CN101426128B (en) | 2007-10-30 | 2007-10-30 | Detection system and method for stolen and lost packet |
Country Status (2)
| Country | Link |
|---|---|
| KR (1) | KR20090044957A (en) |
| CN (1) | CN101426128B (en) |
Families Citing this family (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102256104B (en) * | 2010-05-20 | 2013-12-11 | 鸿富锦精密工业（深圳）有限公司 | Hand-held device and method for dynamically monitoring specific area by using same |
| KR101237970B1 (en) | 2011-01-17 | 2013-02-28 | 포항공과대학교 산학협력단 | Image survailance system and method for detecting left-behind/taken-away of the system |
| CN103514694B (en) * | 2013-09-09 | 2016-05-11 | 重庆邮电大学 | A kind of intrusion detection monitoring system |
| KR102366521B1 (en) | 2015-01-19 | 2022-02-23 | 한화테크윈 주식회사 | System and Method for Detecting Moving Objects |
| CN109040614A (en) * | 2018-08-01 | 2018-12-18 | 安徽宇烁光电科技有限公司 | A kind of more picture video processor scene processing systems |
| CN109409315B (en) * | 2018-11-07 | 2022-01-11 | 浩云科技股份有限公司 | Method and system for detecting remnants in panel area of ATM (automatic Teller machine) |
| CN111914670A (en) * | 2020-07-08 | 2020-11-10 | 浙江大华技术股份有限公司 | Method, device and system for detecting left-over article and storage medium |
| CN114743326A (en) * | 2022-04-07 | 2022-07-12 | 武汉东湖学院 | Intelligent manufacturing workshop anti-theft early warning system with intelligent identification function |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP0805405A2 (en) * | 1996-02-05 | 1997-11-05 | Texas Instruments Incorporated | Motion event detection for video indexing |
| US6424370B1 (en) * | 1999-10-08 | 2002-07-23 | Texas Instruments Incorporated | Motion based event detection system and method |
| US6628835B1 (en) * | 1998-08-31 | 2003-09-30 | Texas Instruments Incorporated | Method and system for defining and recognizing complex events in a video sequence |
| CN1798331A (en) * | 2004-12-29 | 2006-07-05 | 上海乐金广电电子有限公司 | Method of recording images in digital image monitoring device |
-
2007
- 2007-10-30 CN CN200710165489A patent/CN101426128B/en not_active Expired - Fee Related
-
2008
- 2008-02-04 KR KR1020080011392A patent/KR20090044957A/en not_active Application Discontinuation
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP0805405A2 (en) * | 1996-02-05 | 1997-11-05 | Texas Instruments Incorporated | Motion event detection for video indexing |
| US6628835B1 (en) * | 1998-08-31 | 2003-09-30 | Texas Instruments Incorporated | Method and system for defining and recognizing complex events in a video sequence |
| US6424370B1 (en) * | 1999-10-08 | 2002-07-23 | Texas Instruments Incorporated | Motion based event detection system and method |
| CN1798331A (en) * | 2004-12-29 | 2006-07-05 | 上海乐金广电电子有限公司 | Method of recording images in digital image monitoring device |
Also Published As
| Publication number | Publication date |
|---|---|
| KR20090044957A (en) | 2009-05-07 |
| CN101426128A (en) | 2009-05-06 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN101426128B (en) | Detection system and method for stolen and lost packet | |
| US9355308B2 (en) | Auditing video analytics through essence generation | |
| US20080074496A1 (en) | Video analytics for banking business process monitoring | |
| US8208733B2 (en) | Method, medium, and apparatus with estimation of background change | |
| Tian et al. | Real-time detection of abandoned and removed objects in complex environments | |
| US20080106599A1 (en) | Object density estimation in video | |
| KR102217253B1 (en) | Apparatus and method for analyzing behavior pattern | |
| Lee et al. | ArchCam: Real time expert system for suspicious behaviour detection in ATM site | |
| CN103810711A (en) | Keyframe extracting method and system for monitoring system videos | |
| CN101208710A (en) | Target detection and tracking from overhead video streams | |
| CN109711318B (en) | Multi-face detection and tracking method based on video stream | |
| CN107222660A (en) | A kind of distributed network visual monitor system | |
| CN102004904A (en) | Automatic teller machine-based safe monitoring device and method and automatic teller machine | |
| CN103605971A (en) | Method and device for capturing face images | |
| CN102176266A (en) | Visual behavior early warning prompting method and system for automatic teller machine (ATM) bank card | |
| Ullah et al. | Gaussian mixtures for anomaly detection in crowded scenes | |
| Kaltsa et al. | Swarm-based motion features for anomaly detection in crowds | |
| Kawakatsu et al. | Traffic surveillance system for bridge vibration analysis | |
| KR102142315B1 (en) | ATM security system based on image analyses and the method thereof | |
| Mathew et al. | Detecting new stable objects in surveillance video | |
| Cheng et al. | A multiscale parametric background model for stationary foreground object detection | |
| US20200394802A1 (en) | Real-time object detection method for multiple camera images using frame segmentation and intelligent detection pool | |
| CN110913209B (en) | Camera shielding detection method and device, electronic equipment and monitoring system | |
| CN111291728A (en) | Detection system, detection equipment and detection method for illegal crossing of transmission belt behavior | |
| CN104270619B (en) | A kind of security alarm method and device |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20121003 Termination date: 20161030 |
|
| CF01 | Termination of patent right due to non-payment of annual fee |