CN104463220A - License plate detection method and system - Google Patents
License plate detection method and system Download PDFInfo
- Publication number
- CN104463220A CN104463220A CN201410800033.4A CN201410800033A CN104463220A CN 104463220 A CN104463220 A CN 104463220A CN 201410800033 A CN201410800033 A CN 201410800033A CN 104463220 A CN104463220 A CN 104463220A
- Authority
- CN
- China
- Prior art keywords
- license plate
- area
- gray
- rectangular scanning
- edge
- 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.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/50—Context or environment of the image
- G06V20/56—Context or environment of the image exterior to a vehicle by using sensors mounted on the vehicle
- G06V20/58—Recognition of moving objects or obstacles, e.g. vehicles or pedestrians; Recognition of traffic objects, e.g. traffic signs, traffic lights or roads
- G06V20/584—Recognition of moving objects or obstacles, e.g. vehicles or pedestrians; Recognition of traffic objects, e.g. traffic signs, traffic lights or roads of vehicle lights or traffic lights
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Theoretical Computer Science (AREA)
- Image Analysis (AREA)
Abstract
The embodiment of the invention provides a license plate detection method and system. The method includes the steps of obtaining a license plate image, carrying out gray-scale treatment on the license plate image to obtain a gray-scale image, carrying out discrete cosine transformation, sign conversion, square operation, inverse discrete cosine transform and binarization processing sequentially on the gray-scale image to obtain an edge obvious feature graph, searching for the area with the edge density larger than a preset threshold in the edge obvious feature graph to obtain an uncertain license plate area, and detecting the uncertain license plate area by means of a license plate classifier to obtain a license plate area. The method and system are more suitable for hardware implementation, and an algorithm can be transplanted into an embedded system more easily.
Description
Technical field
The present invention relates to technical field of image processing, particularly relate to a kind of detection method of license plate and system.
Background technology
At present, Car license recognition equipment has been widely used in automatic candid photograph and the identification that number plate of vehicle is carried out in the regions such as parking lot, urban road, highway, license plate recognition technology can be divided into that car plate detects, car plate detects, three steps voted by License Plate Segmentation and identification and car plate, wherein, car plate detects, refer to and wherein license plate area detected in the license plate image comprising license plate area, because, license plate area in license plate image is only often the sub-fraction in license plate image, therefore, car plate detects a step the most consuming time in license plate recognition technology often.Universal along with high-definition camera, Car license recognition equipment use high-definition camera becomes trend, how can in high-definition picture fast search full figure find license plate area to become a new challenge.In prior art, after acquisition license plate image, usually directly car plate sorter is adopted, as the AdaBoost car plate sorter based on Ha Er Haar feature, the license plate image obtained is detected, namely Haar detection is carried out to the license plate image full figure obtained, use car plate sorter accurately can detect the license plate area in the license plate image of acquisition, but, because the license plate image area obtained is large, pixel is many, required license plate area is obtained in order to detect in the license plate image obtained, need through numerous and diverse calculating, operand is huge, if be applied in embedded device, its speed cannot meet the demands, be not suitable for hardware device to realize.
Summary of the invention
In view of this, the embodiment of the present invention provides a kind of detection method of license plate and system, directly use car plate sorter to detect license plate area to solve in prior art to the license plate image obtained, in the process obtaining license plate area, operand is huge, if be applied in embedded device, its speed cannot meet the demands, and not be suitable for the problem that hardware device realizes.
For achieving the above object, the embodiment of the present invention provides following technical scheme:
A kind of detection method of license plate, comprising:
Obtain license plate image, and gray processing process is carried out to described license plate image, obtain gray-scale map;
Discrete cosine transform, sign conversion, square operation, inverse discrete cosine transformation and binary conversion treatment are carried out successively to described gray-scale map, obtains edge notable feature figure;
Search for marginal density in described edge notable feature figure and be greater than the region of predetermined threshold, obtain doubtful license plate area;
Use car plate sorter to detect to described doubtful license plate area, obtain license plate area.
Wherein, also comprise after obtaining edge notable feature figure described in:
Generate the integrogram of described edge notable feature figure, search for marginal density in described integrogram and be greater than the region of predetermined threshold, obtain doubtful license plate area.
Wherein, the region that in the described edge notable feature figure of described search, marginal density is greater than predetermined threshold comprises:
Get predetermined number preliminary dimension rectangular scanning frame and staggered scanning is carried out to described edge notable feature figure;
Each rectangular scanning frame is a mobile pixel often, calculates the gray-scale value of all pixels in it;
Retain the rectangular scanning frame that average edge gray-scale value is greater than predetermined threshold.
Wherein, also comprise after described reservation average edge gray-scale value is greater than the rectangular scanning frame of predetermined threshold: utilize the area registration of rectangular scanning frame with a grain of salt carry out cluster operation;
Described utilization the area registration of rectangular scanning frame with a grain of salt carry out cluster operation and comprise:
Calculate in rectangular scanning frame with a grain of salt between two between area registration;
If the area registration between two rectangular scanning frames is greater than predetermined threshold, then think that these two rectangular scanning frames are same class, according to described between two between area registration rectangular scanning frame rectangle with a grain of salt is classified, retain the rectangular scanning frame that in every class rectangular scanning frame, average edge gray-scale value is maximum.
Wherein, also comprise after obtaining doubtful license plate area described in:
Horizontal projection and vertical projection are carried out to described doubtful license plate area, obtains the rear doubtful license plate area of projection, use car plate sorter to detect to license plate area doubtful after described projection, obtain license plate area.
Wherein, described car plate sorter comprises:
Obtain car plate training sample;
Getting the part comprising car plate in described car plate training sample is positive sample, and the part not comprising car plate of getting in described license plate image sample is negative sample;
Use Ha Er Haar feature to characterize described positive sample and negative sample, form Haar proper vector;
Utilize iterative algorithm to train described Haar proper vector, obtain car plate sorter.
Wherein, also comprise after obtaining license plate area described in:
Segmentation of the characters and their identification is carried out to described license plate area, obtains the degree of confidence of described license plate area;
If described degree of confidence is greater than predetermined threshold, then illustrate there is no flase drop.
A kind of car plate detection system, comprising: acquisition module, computing module, search module and detection module; Wherein,
Described acquisition module, for obtaining license plate image, and carrying out gray processing process to described license plate image, obtaining gray-scale map;
Described computing module, for carrying out discrete cosine transform, sign conversion, square operation, inverse discrete cosine transformation and binary conversion treatment successively to described gray-scale map, obtains edge notable feature figure;
Described search module, being greater than the region of predetermined threshold for searching for marginal density in described edge notable feature figure, obtaining doubtful license plate area;
Described detection module, for using car plate sorter to detect to described doubtful license plate area, obtains license plate area.
Wherein, described search module comprises: scanning element, the first computing unit and the first stick unit; Wherein,
Described scanning element, carries out staggered scanning for getting predetermined number preliminary dimension rectangular scanning frame to described edge notable feature figure;
Described computing unit, for each rectangular scanning frame often mobile pixel, calculates the gray-scale value of all pixels in it;
Described stick unit, is greater than the rectangular scanning frame of predetermined threshold for retaining average edge gray-scale value.
Wherein, described search module also comprises: the second computing unit and the second stick unit; Wherein,
Described second computing unit, for calculate in rectangular scanning frame with a grain of salt between two between area registration;
Described second stick unit, if be greater than predetermined threshold for the area registration between two rectangular scanning frames, then think that these two rectangular scanning frames are same class, according to described between two between area registration rectangular scanning frame rectangle with a grain of salt is classified, retain the rectangular scanning frame that in every class rectangular scanning frame, average edge gray-scale value is maximum.
Based on technique scheme, the detection method of license plate that the embodiment of the present invention provides and system, after acquisition license plate image, by carrying out gray processing process to the license plate image obtained, obtain gray-scale map, then successively discrete cosine transform is carried out to the gray-scale map obtained, sign changes, square operation, inverse discrete cosine transformation and binary conversion treatment, obtain edge notable feature figure, search for marginal density in this edge notable feature figure again and be greater than the region of predetermined threshold, obtain doubtful license plate area, last only needs detect by using car plate sorter to the doubtful license plate area obtained, just required license plate area can be obtained.After acquisition license plate image, by utilizing the notable feature that in license plate image, the marginal density of license plate area is larger, only car plate sorter is used to detect to the doubtful license plate area obtained after obtaining doubtful license plate area, just license plate area accurately can be obtained, do not re-use car plate sorter and full figure detection is carried out to the license plate image obtained, decrease the calculated amount detecting and obtain license plate area in license plate image, and, edge notable feature figure is obtained by the method for discrete cosine transform, algorithm is more prone to be transplanted in embedded system, is more suitable for hardware implementing.
Accompanying drawing explanation
In order to be illustrated more clearly in the embodiment of the present invention or technical scheme of the prior art, be briefly described to the accompanying drawing used required in embodiment or description of the prior art below, apparently, accompanying drawing in the following describes is only embodiments of the invention, for those of ordinary skill in the art, under the prerequisite not paying creative work, other accompanying drawing can also be obtained according to the accompanying drawing provided.
The process flow diagram of the detection method of license plate that Fig. 1 provides for the embodiment of the present invention;
Another process flow diagram of the detection method of license plate that Fig. 2 provides for the embodiment of the present invention;
Search for marginal density in edge notable feature figure in the detection method of license plate that Fig. 3 provides for the embodiment of the present invention and be greater than the method flow diagram in the region of predetermined threshold;
The area registration of rectangular scanning frame with a grain of salt is utilized to carry out the method flow diagram of cluster operation in the detection method of license plate that Fig. 4 provides by the embodiment of the present invention;
A method flow diagram again of the detection method of license plate that Fig. 5 provides for the embodiment of the present invention;
The method flow diagram of car plate sorter is trained in the detection method of license plate that Fig. 6 provides for the embodiment of the present invention;
The method flow diagram of flase drop is judged whether in the detection method of license plate that Fig. 7 provides for the embodiment of the present invention;
The system chart of the car plate detection system that Fig. 8 provides for the embodiment of the present invention;
The structured flowchart of search module in the system chart of the car plate detection system that Fig. 9 provides for the embodiment of the present invention;
Another structured flowchart of search module in the system chart of the car plate detection system that Figure 10 provides for the embodiment of the present invention.
Embodiment
Below in conjunction with the accompanying drawing in the embodiment of the present invention, be clearly and completely described the technical scheme in the embodiment of the present invention, obviously, described embodiment is only the present invention's part embodiment, instead of whole embodiments.Based on the embodiment in the present invention, those of ordinary skill in the art, not making the every other embodiment obtained under creative work prerequisite, belong to the scope of protection of the invention.
The process flow diagram of the detection method of license plate that Fig. 1 provides for the embodiment of the present invention, this detection method of license plate only needs to use the car plate sorter based on Ha Er Haar features training to detect to the doubtful license plate area obtained, just license plate area accurately can be obtained, decrease the calculated amount detecting and obtain license plate area in license plate image, and obtain edge notable feature figure by the method for discrete cosine transform, be more suitable for hardware implementing, algorithm is more prone to be transplanted in embedded system; With reference to Fig. 1, this detection method of license plate can comprise:
Step S100: obtain license plate image, and gray processing process is carried out to described license plate image, obtain gray-scale map;
License plate image refers to the image comprising automotive license plate collected by image capture devices such as video cameras at places such as gateways, parking lot.
Optionally, the image capture device obtaining license plate image can be video camera.
Step S110: discrete cosine transform, sign conversion, square operation, inverse discrete cosine transformation and binary conversion treatment are carried out successively to described gray-scale map, obtains edge notable feature figure;
Optionally, if set the wide as M of the two-dimentional license plate image of input, height is N, and so, any point on license plate image can be expressed as: f (x, y), wherein, x=0,1,2, ..., M-1, y=0,1,2, ..., N-1, if the cosine transform coefficient representing two-dimentional license plate image with F (u, v), so has u=0,1,2 ..., M-1, v=0,1,2 ..., N-1, the discrete cosine transform of two-dimentional license plate image can be expressed as:
Wherein,
Optionally, carry out sign conversion to the image after discrete cosine transform, each numerical value of Sign function to input is changed, Ke Yiwei:
Carrying out square operation to the image after sign conversion, after have passed through square operation, was that the point of-1 can become 1 originally.
Optionally, when carrying out inverse discrete cosine transformation to the figure after square transport, if set the size of image as MxN, the cosine transform coefficient representing two dimensional image with F (u, v), so has u=0, and 1,2 ..., M-1, v=0,1,2 ..., N-1.For any point f (x, y) on time-domain diagram picture, x=0,1,2 ..., M-1, y=0,1,2 ..., N-1, inverse discrete cosine transformation can be expressed as:
Wherein
Carry out binaryzation operation to the image obtained after inverse discrete cosine transformation, be not that the some expression of 0 has certain marginal information after operation, this figure is edge notable feature figure.
Optionally, after obtaining edge notable feature figure, the integrogram of this edge notable feature figure can be generated, search for marginal density in this integrogram and be greater than the region of predetermined threshold, obtain doubtful license plate area.To notable feature figure formation product component, use integrogram can accelerate calculate, avoid a square frame institute a little notable feature value addition double counting.
Step S120: search for marginal density in described edge notable feature figure and be greater than the region of predetermined threshold, obtain doubtful license plate area;
Optionally, staggered scanning can be carried out by getting predetermined number preliminary dimension rectangular scanning frame to described edge notable feature figure; Each rectangular scanning frame is a mobile pixel often, calculates the gray-scale value of all pixels in it; Retain the rectangular scanning frame that average edge gray-scale value is greater than predetermined threshold, search for marginal density in described edge notable feature figure and be greater than the region of predetermined threshold, obtain doubtful license plate area.
Optionally, three dimension scales can be used for 1:1.5:2.25 is to scan edge notable feature figure, the square frame of each yardstick is 1:1.5 from the top of figure to nethermost ratio, and the size of frame becomes large in proportion gradually.The size of the such as frame of the first yardstick is 90x30, and when this frame slides into image base time, size variation is 135x45.Namely three yardsticks are on diverse location, and the size of square frame presets.The square frame of 3 different scales is used to carry out staggered scanning to integrogram.
Optionally, in the described edge notable feature figure of search, marginal density is greater than in the operation in the region of predetermined threshold, retain can also to utilize after average edge gray-scale value is greater than the rectangular scanning frame of predetermined threshold the area registration of rectangular scanning frame with a grain of salt carry out cluster operation, remove the frame scanning the repetition obtained, reduce the quantity of the doubtful license plate area using car plate sorter to check, accelerate the speed detecting license plate area further.
Optionally, utilize the area registration of rectangular scanning frame with a grain of salt carry out cluster operation and can be, by calculate in rectangular scanning frame with a grain of salt between two between area registration, if the area registration between two rectangular scanning frames is greater than predetermined threshold, then think that these two rectangular scanning frames are same class, according to described between two between area registration rectangular scanning frame rectangle with a grain of salt is classified, retain the rectangular scanning frame that in every class rectangular scanning frame, average edge gray-scale value is maximum.
Optionally, for general car plate, the marginal density of the character in car plate is the highest, the marginal density that character adds part frame will reduce, search for the frame obtained and only represent the region that in image, marginal density is larger, complete car plate may be comprised, also part car plate may only be comprised, not necessarily accurate to the location of car plate, therefore, can obtain predetermined scope to searching for the doubtful license plate area obtained, marginal information is extracted in the region after expansion first, and then edge figure carries out horizontal projection and vertical projection.
Optionally, the doubtful license plate area searched respectively can be expanded 20% of little figure width by left and right, song diffuses into 30% of height up and down.
Optionally, can be added by pixel value of each row square frame, then calculate the pixel value of each row and mean value, to turn right search from Far Left, find the first row being greater than mean value, the coordinate marking these row is left hand edge.To turn left search from rightmost, find the first row being greater than mean value, the coordinate marking these row is that right hand edge is to carry out horizontal projection.
Optionally, can by the pixel value of every a line of square frame be added, then calculate the pixel value of every a line and mean value, from topmost down searching for, find the first row being greater than mean value, the coordinate marking these row is coboundary, from up searching for bottom, find the first row being greater than mean value, the coordinate marking these row is that lower limb is to carry out vertical projection.
Because the marginal density of general license plate area is larger, the marginal density of car plate neighboring area is smaller, more accurately can find the position of car plate by vertical and level with projection; In addition for some background areas, if do not have stronger edge difference, such as thick grass, square frame cannot shrink after expanding.Some less object there will be after vertical and horizontal projection long-pending too small in addition, and minimum value and maximal value therefore by arranging square frame can filter the empty inspection of a part.
Step S130: use car plate sorter to detect to described doubtful license plate area, obtain license plate area.
License plate area refers in license plate image the region comprising car plate, in general, only has sub-fraction to be license plate area in the whole figure of a width license plate image.
Optionally, the car plate sorter of use can be the AdaBoost car plate sorter based on Ha Er Haar features training.
Optionally, can by obtaining car plate training sample; Getting the part comprising car plate in the car plate training sample of acquisition is positive sample, and the part not comprising car plate of getting in the license plate image of acquisition is negative sample; Use Haar feature to characterize described positive sample and negative sample, form Haar proper vector; Utilize iterative algorithm to train described Haar proper vector, obtain car plate sorter.
Wherein, Haar feature is a kind of rectangular characteristic, and rectangular characteristic is to some simple graphic structure such as edges, and line segment is more responsive, but can only describe particular orientation, more roughly.But for a detecting device, the inside comprises a hundreds of thousands different rectangular characteristic, then by using Adaboost algorithm to train, just can obtain a strong classifier.
The feature templates of each Haar combines by the rectangle of two or more congruence is adjacent, adularescent and black two kinds of rectangles in feature templates, and by this template definition be white rectangle pixel and deduct black rectangle pixel and.Feature templates can be placed arbitrarily with arbitrary dimension in subwindow, and each form is called a feature.
Optionally, the iterative algorithm of use can be Adaboost algorithm, and Adaboost algorithm is a kind of adaptive boosting algorithm, and its basic thought is when sorter is correctly classified to some sample, then reduce the weights of these samples.When mis-classification, then increase the weights of these samples, allow learning algorithm concentrate in follow-up study and more difficult training sample is learnt, finally obtain the sorter that a recognition accuracy is desirable.The training of every one deck adopts minimum permission verification and measurement ratio and maximum permission false drop rate as strong classifier iteration stopping foundation, when every one deck strong classifier and when all reaching the setting value before training, namely this grade of training complete.The training negative sample of lower one deck strong classifier is produced from this layer by the negative sample of mis-classification.
Optionally, can by carrying out segmentation of the characters and their identification to have judged whether flase drop to the license plate area obtained after obtaining license plate area, after segmentation of the characters and their identification is carried out to license plate area, the degree of confidence of license plate area will be obtained, judge whether the license plate area degree of confidence obtained is greater than predetermined threshold, if be greater than, illustrates do not have flase drop.
Based on technique scheme, the detection method of license plate that the embodiment of the present invention provides and system, after acquisition license plate image, by carrying out gray processing process to the license plate image obtained, obtain gray-scale map, then successively discrete cosine transform is carried out to the gray-scale map obtained, sign changes, square operation, inverse discrete cosine transformation and binary conversion treatment, obtain edge notable feature figure, search for marginal density in this edge notable feature figure again and be greater than the region of predetermined threshold, obtain doubtful license plate area, last only needs detect by using car plate sorter to the doubtful license plate area obtained, just required license plate area can be obtained.After acquisition license plate image, by utilizing the notable feature that in license plate image, the marginal density of license plate area is larger, only car plate sorter is used to detect to the doubtful license plate area obtained after obtaining doubtful license plate area, just license plate area accurately can be obtained, do not re-use car plate sorter and full figure detection is carried out to the license plate image obtained, decrease the calculated amount detecting and obtain license plate area in license plate image, and, edge notable feature figure is obtained by the method for discrete cosine transform, algorithm is more prone to be transplanted in embedded system, is more suitable for hardware implementing.
Optionally, Fig. 2 shows another process flow diagram of the detection method of license plate that the embodiment of the present invention provides, and with reference to Fig. 2, this car plate detects other method and can comprise:
Step S200: obtain license plate image, and gray processing process is carried out to described license plate image, obtain gray-scale map;
Step S210: discrete cosine transform, sign conversion, square operation, inverse discrete cosine transformation and binary conversion treatment are carried out successively to described gray-scale map, obtains edge notable feature figure;
Step S220: the integrogram generating described edge notable feature figure;
Each point (x, y) on integrogram contains the notable feature value from point (0,0) to all pixel of point (x, y).
Optionally, if represent notable feature figure with I (x, y), II (x, y) represents the integrogram calculated, so:
An arbitrary rectangle can use following mode to calculate:
If the upper left corner of rectangle is (x
lt, y
lt), the coordinate in the lower right corner is (x
rb, y
rb), so the integrogram of this rectangle can calculate with following formula:
SUM
D＝II(x
rb,y
rb)-II(x
lt,y
rb)-II(x
rb,y
lt)+II(x
lt,y
lt)。
Step S230: search for marginal density in described integrogram and be greater than the region of predetermined threshold, obtain doubtful license plate area;
To notable feature figure formation product component, use integrogram can accelerate calculate, avoid a square frame institute a little notable feature value addition double counting.
Step S240: use the car plate sorter based on Ha Er Haar features training to detect to described doubtful license plate area, obtain license plate area.
Optionally, Fig. 3 shows and searches for marginal density in edge notable feature figure in the detection method of license plate that the embodiment of the present invention provides and be greater than the method flow diagram in the region of predetermined threshold, with reference to Fig. 3, the method that in this search edge notable feature figure, marginal density is greater than the region of predetermined threshold can comprise:
Step S300: get predetermined number preliminary dimension rectangular scanning frame and staggered scanning is carried out to described edge notable feature figure;
Optionally, three dimension scales can be used for 1:1.5:2.25 is to scan edge notable feature figure, the square frame of each yardstick is 1:1.5 from the top of figure to nethermost ratio, and the size of frame becomes large in proportion gradually.The size of the such as frame of the first yardstick is 90x30, and when this frame slides into image base time, size variation is 135x45.Namely three yardsticks are on diverse location, and the size of square frame presets.The square frame of 3 different scales is used to carry out staggered scanning to integrogram.
Step S310: each rectangular scanning frame is a mobile pixel often, calculates the gray-scale value of all pixels in it;
Step S320: retain the rectangular scanning frame that average edge gray-scale value is greater than predetermined threshold.
Edge gray-scale value refers to, the gray-scale value of rectangular scanning edge edge pixel, and average edge gray-scale value refers to the mean value of the gray-scale value of rectangular scanning edge edge pixel, and generally speaking, gray-scale value higher then marginal density in edge is larger.
Optionally, Fig. 4 show in the detection method of license plate that the embodiment of the present invention provides utilize the area registration of rectangular scanning frame with a grain of salt carry out the method flow diagram of cluster operation, with reference to Fig. 4, this utilization the area registration of the rectangular scanning frame with a grain of salt method of carrying out cluster operation can comprise:
Step S400: calculate in rectangular scanning frame with a grain of salt between two between area registration;
Step S410: if the area registration between two rectangular scanning frames is greater than predetermined threshold, then think that these two rectangular scanning frames are same class, according to described between two between area registration rectangular scanning frame rectangle with a grain of salt is classified, retain the rectangular scanning frame that in every class rectangular scanning frame, average edge gray-scale value is maximum.
As long as rectangular scanning frame overlapping area between any two reaches threshold value, just think that the frame that these two rectangular scanning frames are swept to belongs to a class, for of a sort frame, only need to retain the frame that wherein edge gray-scale value is larger, reduce the quantity of the doubtful license plate area using car plate sorter to check, accelerate the speed detecting license plate area further.
Optionally, Fig. 5 shows a method flow diagram again of the detection method of license plate that the embodiment of the present invention provides, and with reference to Fig. 5, this car plate detects a method again and can comprise:
Step S500: obtain license plate image, and gray processing process is carried out to described license plate image, obtain gray-scale map;
Step S510: discrete cosine transform, sign conversion, square operation, inverse discrete cosine transformation and binary conversion treatment are carried out successively to described gray-scale map, obtains edge notable feature figure;
Step S520: search for marginal density in described edge notable feature figure and be greater than the region of predetermined threshold, obtain doubtful license plate area;
Step S530: carry out horizontal projection and vertical projection to described doubtful license plate area, obtains the rear doubtful license plate area of projection;
Optionally, can be added by pixel value of each row square frame, then calculate the pixel value of each row and mean value, to turn right search from Far Left, find the first row being greater than mean value, the coordinate marking these row is left hand edge.To turn left search from rightmost, find the first row being greater than mean value, the coordinate marking these row is that right hand edge is to carry out horizontal projection.
Optionally, can by the pixel value of every a line of square frame be added, then calculate the pixel value of every a line and mean value, from topmost down searching for, find the first row being greater than mean value, the coordinate marking these row is coboundary, from up searching for bottom, find the first row being greater than mean value, the coordinate marking these row is that lower limb is to carry out vertical projection.
Because the marginal density of general license plate area is larger, the marginal density of car plate neighboring area is smaller, more accurately can find the position of car plate by vertical and level with projection.
Step S540: use the car plate sorter based on Ha Er Haar features training to detect to license plate area doubtful after described projection, obtain license plate area.
Optionally, Fig. 6 shows the method flow diagram of training car plate sorter in the detection method of license plate that the embodiment of the present invention provides, and with reference to Fig. 6, the method for this training car plate sorter can comprise:
Step S600: obtain car plate training sample;
Car plate training sample refers to the license plate image for training car plate sorter.
Optionally, use video camera take under each period and weather condition and store the video of gateway, a large amount of parking lot, obtain car plate training sample.
Step S610: getting the part comprising car plate in described car plate training sample is positive sample, the part not comprising car plate of getting in described license plate image sample is negative sample;
Optionally, can be intercepted the part comprising car plate in car plate training sample by artificial mode be positive sample, and the part not comprising car plate intercepted in license plate image sample by artificial mode is negative sample.
Step S620: use Haar feature to characterize described positive sample and negative sample, form Haar proper vector;
Step S630: utilize iterative algorithm to train described Haar proper vector, obtain car plate sorter.
Optionally, the iterative algorithm of use can be Adaboost algorithm, finally the available Adaboost car plate sorter based on Haar feature.
Optionally, Fig. 7 shows the method flow diagram judging whether flase drop in the detection method of license plate that the embodiment of the present invention provides, and with reference to Fig. 7, this judges whether that the method for flase drop can comprise:
Step S700: carry out segmentation of the characters and their identification to described license plate area, obtains the degree of confidence of described license plate area;
After segmentation of the characters and their identification is carried out to license plate area, the degree of confidence of this license plate area will be obtained.
Step S710: judge that described degree of confidence is greater than predetermined threshold;
Step S720: if be greater than, then judge do not have flase drop.
The detection method of license plate that the embodiment of the present invention provides, only need to use car plate sorter to detect to the doubtful license plate area obtained, just license plate area accurately can be obtained, decrease the calculated amount detecting and obtain license plate area in license plate image, and obtain edge notable feature figure by the method for discrete cosine transform, algorithm is more prone to be transplanted in embedded system, is more suitable for hardware implementing.
Be introduced the car plate detection system that the embodiment of the present invention provides below, car plate detection system described below can mutual corresponding reference with above-described detection method of license plate.
The system chart of the car plate detection system that Fig. 8 provides for the embodiment of the present invention, with reference to Fig. 8, this car plate detection system can comprise: acquisition module 100, computing module 200, search module 300 and detection module 400; Wherein,
Acquisition module 100, for obtaining license plate image, and carrying out gray processing process to described license plate image, obtaining gray-scale map;
Computing module 200, for carrying out discrete cosine transform, sign conversion, square operation, inverse discrete cosine transformation and binary conversion treatment successively to described gray-scale map, obtains edge notable feature figure;
Search module 300, being greater than the region of predetermined threshold for searching for marginal density in described edge notable feature figure, obtaining doubtful license plate area;
Detection module 400, for using car plate sorter to detect to described doubtful license plate area, obtains license plate area.
Optionally, Fig. 9 shows the structured flowchart of search module 300 in the system chart of the car plate detection system that the embodiment of the present invention provides, and with reference to Fig. 9, this search module 300 can comprise: scanning element 310, first computing unit 320 and the first stick unit 330; Wherein,
Scanning element 310, carries out staggered scanning for getting predetermined number preliminary dimension rectangular scanning frame to described edge notable feature figure;
First computing unit 320, for each rectangular scanning frame often mobile pixel, calculates the gray-scale value of all pixels in it;
First stick unit 330, is greater than the rectangular scanning frame of predetermined threshold for retaining average edge gray-scale value.
Optionally, Figure 10 shows another structured flowchart of search module 300 in the system chart of the car plate detection system that the embodiment of the present invention provides, and with reference to Figure 10, this search module 300 can also comprise: the second computing unit 340 and the second stick unit 350; Wherein,
Second computing unit 340, for calculate in rectangular scanning frame with a grain of salt between two between area registration;
Second stick unit 350, if be greater than predetermined threshold for the area registration between two rectangular scanning frames, then think that these two rectangular scanning frames are same class, according to described between two between area registration rectangular scanning frame rectangle with a grain of salt is classified, retain the rectangular scanning frame that in every class rectangular scanning frame, average edge gray-scale value is maximum.
The car plate detection system that the embodiment of the present invention provides, only need to use car plate sorter to detect to the doubtful license plate area obtained, just license plate area accurately can be obtained, decrease the calculated amount detecting and obtain license plate area in license plate image, and obtain edge notable feature figure by the method for discrete cosine transform, algorithm is more prone to be transplanted in embedded system, is more suitable for hardware implementing.
In this instructions, each embodiment adopts the mode of going forward one by one to describe, and what each embodiment stressed is the difference with other embodiments, between each embodiment identical similar portion mutually see.For device disclosed in embodiment, because it corresponds to the method disclosed in Example, so description is fairly simple, relevant part illustrates see method part.
Professional can also recognize further, in conjunction with unit and the algorithm steps of each example of embodiment disclosed herein description, can realize with electronic hardware, computer software or the combination of the two, in order to the interchangeability of hardware and software is clearly described, generally describe composition and the step of each example in the above description according to function.These functions perform with hardware or software mode actually, depend on application-specific and the design constraint of technical scheme.Professional and technical personnel can use distinct methods to realize described function to each specifically should being used for, but this realization should not thought and exceeds scope of the present invention.
The software module that the method described in conjunction with embodiment disclosed herein or the step of algorithm can directly use hardware, processor to perform, or the combination of the two is implemented.Software module can be placed in the storage medium of other form any known in random access memory (RAM), internal memory, ROM (read-only memory) (ROM), electrically programmable ROM, electrically erasable ROM, register, hard disk, moveable magnetic disc, CD-ROM or technical field.
To the above-mentioned explanation of the disclosed embodiments, professional and technical personnel in the field are realized or uses the present invention.To be apparent for those skilled in the art to the multiple amendment of these embodiments, General Principle as defined herein can without departing from the spirit or scope of the present invention, realize in other embodiments.Therefore, the present invention can not be restricted to these embodiments shown in this article, but will meet the widest scope consistent with principle disclosed herein and features of novelty.
Claims (10)
1. a detection method of license plate, is characterized in that, comprising:
Obtain license plate image, and gray processing process is carried out to described license plate image, obtain gray-scale map;
Discrete cosine transform, sign conversion, square operation, inverse discrete cosine transformation and binary conversion treatment are carried out successively to described gray-scale map, obtains edge notable feature figure;
Search for marginal density in described edge notable feature figure and be greater than the region of predetermined threshold, obtain doubtful license plate area;
Use car plate sorter to detect to described doubtful license plate area, obtain license plate area.
2. detection method of license plate according to claim 1, is characterized in that, described in obtain edge notable feature figure after also comprise:
Generate the integrogram of described edge notable feature figure, search for marginal density in described integrogram and be greater than the region of predetermined threshold, obtain doubtful license plate area.
3. detection method of license plate according to claim 1, is characterized in that, the region that in the described edge notable feature figure of described search, marginal density is greater than predetermined threshold comprises:
Get predetermined number preliminary dimension rectangular scanning frame and staggered scanning is carried out to described edge notable feature figure;
Each rectangular scanning frame is a mobile pixel often, calculates the gray-scale value of all pixels in it;
Retain the rectangular scanning frame that average edge gray-scale value is greater than predetermined threshold.
4. detection method of license plate according to claim 3, is characterized in that, also comprises after described reservation average edge gray-scale value is greater than the rectangular scanning frame of predetermined threshold: utilize the area registration of rectangular scanning frame with a grain of salt carry out cluster operation;
Described utilization the area registration of rectangular scanning frame with a grain of salt carry out cluster operation and comprise:
Calculate in rectangular scanning frame with a grain of salt between two between area registration;
If the area registration between two rectangular scanning frames is greater than predetermined threshold, then think that these two rectangular scanning frames are same class, according to described between two between area registration rectangular scanning frame rectangle with a grain of salt is classified, retain the rectangular scanning frame that in every class rectangular scanning frame, average edge gray-scale value is maximum.
5. detection method of license plate according to claim 1, is characterized in that, described in obtain doubtful license plate area after also comprise:
Horizontal projection and vertical projection are carried out to described doubtful license plate area, obtains the rear doubtful license plate area of projection, use car plate sorter to detect to license plate area doubtful after described projection, obtain license plate area.
6. detection method of license plate according to claim 1, is characterized in that, described car plate sorter comprises:
Obtain car plate training sample;
Getting the part comprising car plate in described car plate training sample is positive sample, and the part not comprising car plate of getting in described license plate image sample is negative sample;
Use Ha Er Haar feature to characterize described positive sample and negative sample, form Haar proper vector;
Utilize iterative algorithm to train described Haar proper vector, obtain car plate sorter.
7. detection method of license plate according to claim 1, is characterized in that, described in obtain license plate area after also comprise:
Segmentation of the characters and their identification is carried out to described license plate area, obtains the degree of confidence of described license plate area;
If described degree of confidence is greater than predetermined threshold, then illustrate there is no flase drop.
8. a car plate detection system, is characterized in that, comprising: acquisition module, computing module, search module and detection module; Wherein,
Described acquisition module, for obtaining license plate image, and carrying out gray processing process to described license plate image, obtaining gray-scale map;
Described computing module, for carrying out discrete cosine transform, sign conversion, square operation, inverse discrete cosine transformation and binary conversion treatment successively to described gray-scale map, obtains edge notable feature figure;
Described search module, being greater than the region of predetermined threshold for searching for marginal density in described edge notable feature figure, obtaining doubtful license plate area;
Described detection module, for using car plate sorter to detect to described doubtful license plate area, obtains license plate area.
9. car plate detection system according to claim 8, is characterized in that, described search module comprises: scanning element, the first computing unit and the first stick unit; Wherein,
Described scanning element, carries out staggered scanning for getting predetermined number preliminary dimension rectangular scanning frame to described edge notable feature figure;
Described first computing unit, for each rectangular scanning frame often mobile pixel, calculates the gray-scale value of all pixels in it;
Described first stick unit, is greater than the rectangular scanning frame of predetermined threshold for retaining average edge gray-scale value.
10. car plate detection system according to claim 9, is characterized in that, described search module also comprises: the second computing unit and the second stick unit; Wherein,
Described second computing unit, for calculate in rectangular scanning frame with a grain of salt between two between area registration;
Described second stick unit, if be greater than predetermined threshold for the area registration between two rectangular scanning frames, then think that these two rectangular scanning frames are same class, according to described between two between area registration rectangular scanning frame rectangle with a grain of salt is classified, retain the rectangular scanning frame that in every class rectangular scanning frame, average edge gray-scale value is maximum.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410800033.4A CN104463220A (en) | 2014-12-19 | 2014-12-19 | License plate detection method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410800033.4A CN104463220A (en) | 2014-12-19 | 2014-12-19 | License plate detection method and system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN104463220A true CN104463220A (en) | 2015-03-25 |
Family
ID=52909234
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410800033.4A Pending CN104463220A (en) | 2014-12-19 | 2014-12-19 | License plate detection method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104463220A (en) |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104867080A (en) * | 2015-05-18 | 2015-08-26 | 健康宝互联网技术有限公司 | Identity recognition method for self-service health examination |
| CN105182320A (en) * | 2015-07-14 | 2015-12-23 | 安徽清新互联信息科技有限公司 | Depth measurement-based vehicle distance detection method |
| CN105488532A (en) * | 2015-12-03 | 2016-04-13 | 深圳市捷顺科技实业股份有限公司 | License plate screening method and device |
| CN106096607A (en) * | 2016-06-12 | 2016-11-09 | 湘潭大学 | A kind of licence plate recognition method |
| CN106683073A (en) * | 2015-11-11 | 2017-05-17 | 杭州海康威视数字技术股份有限公司 | License plate detection method, camera and server |
| CN106778731A (en) * | 2017-01-13 | 2017-05-31 | 深圳市华尊科技股份有限公司 | A kind of license plate locating method and terminal |
| CN106778760A (en) * | 2017-01-18 | 2017-05-31 | 博康智能信息技术有限公司 | A kind of license plate image line of demarcation approximating method and device |
| CN111063199A (en) * | 2019-12-19 | 2020-04-24 | 深圳市捷顺科技实业股份有限公司 | Method and device for associating vehicle with license plate and terminal equipment |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20080050036A1 (en) * | 2006-08-25 | 2008-02-28 | Portalplayer, Inc. | Method and system for performing two-dimensional transform on data value array with reduced power consumption |
| CN102663405A (en) * | 2012-05-14 | 2012-09-12 | 武汉大学 | Prominence and Gaussian mixture model-based method for extracting foreground of surveillance video |
| CN103559793A (en) * | 2013-11-18 | 2014-02-05 | 哈尔滨工业大学 | Detecting method and device for sun shield in car |
| CN103745197A (en) * | 2013-12-27 | 2014-04-23 | 深圳市捷顺科技实业股份有限公司 | Detection method of license plate and device thereof |
-
2014
- 2014-12-19 CN CN201410800033.4A patent/CN104463220A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20080050036A1 (en) * | 2006-08-25 | 2008-02-28 | Portalplayer, Inc. | Method and system for performing two-dimensional transform on data value array with reduced power consumption |
| CN102663405A (en) * | 2012-05-14 | 2012-09-12 | 武汉大学 | Prominence and Gaussian mixture model-based method for extracting foreground of surveillance video |
| CN103559793A (en) * | 2013-11-18 | 2014-02-05 | 哈尔滨工业大学 | Detecting method and device for sun shield in car |
| CN103745197A (en) * | 2013-12-27 | 2014-04-23 | 深圳市捷顺科技实业股份有限公司 | Detection method of license plate and device thereof |
Non-Patent Citations (1)
| Title |
|---|
| 卢佩: "基于类生物视觉特性的目标识别", 《中国优秀博士学位论文全文数据库 信息科技辑》 * |
Cited By (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104867080A (en) * | 2015-05-18 | 2015-08-26 | 健康宝互联网技术有限公司 | Identity recognition method for self-service health examination |
| CN104867080B (en) * | 2015-05-18 | 2018-09-25 | 健康宝互联网技术有限公司 | A kind of personal identification method when self-service physical examination |
| CN105182320A (en) * | 2015-07-14 | 2015-12-23 | 安徽清新互联信息科技有限公司 | Depth measurement-based vehicle distance detection method |
| CN106683073A (en) * | 2015-11-11 | 2017-05-17 | 杭州海康威视数字技术股份有限公司 | License plate detection method, camera and server |
| CN106683073B (en) * | 2015-11-11 | 2020-02-18 | 杭州海康威视数字技术股份有限公司 | License plate detection method, camera and server |
| CN105488532A (en) * | 2015-12-03 | 2016-04-13 | 深圳市捷顺科技实业股份有限公司 | License plate screening method and device |
| CN105488532B (en) * | 2015-12-03 | 2019-03-05 | 深圳市捷顺科技实业股份有限公司 | A kind of license plate screening technique and device |
| CN106096607A (en) * | 2016-06-12 | 2016-11-09 | 湘潭大学 | A kind of licence plate recognition method |
| CN106778731A (en) * | 2017-01-13 | 2017-05-31 | 深圳市华尊科技股份有限公司 | A kind of license plate locating method and terminal |
| CN106778731B (en) * | 2017-01-13 | 2019-09-06 | 深圳市华尊科技股份有限公司 | A kind of license plate locating method and terminal |
| CN106778760A (en) * | 2017-01-18 | 2017-05-31 | 博康智能信息技术有限公司 | A kind of license plate image line of demarcation approximating method and device |
| CN111063199A (en) * | 2019-12-19 | 2020-04-24 | 深圳市捷顺科技实业股份有限公司 | Method and device for associating vehicle with license plate and terminal equipment |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN104463220A (en) | License plate detection method and system | |
| CN111368687B (en) | Sidewalk vehicle illegal parking detection method based on target detection and semantic segmentation | |
| CN109919072B (en) | Fine vehicle type recognition and flow statistics method based on deep learning and trajectory tracking | |
| CN109657716B (en) | Vehicle appearance damage identification method based on deep learning | |
| CN104239867B (en) | License plate locating method and system | |
| CN103366602B (en) | Method of determining parking lot occupancy from digital camera images | |
| Zheng et al. | A novel vehicle detection method with high resolution highway aerial image | |
| US20180165552A1 (en) | All-weather thermal-image pedestrian detection method | |
| CN103942560B (en) | A kind of high-resolution video vehicle checking method in intelligent traffic monitoring system | |
| CN105260749B (en) | Real-time target detection method based on direction gradient binary pattern and soft cascade SVM | |
| CN104318225A (en) | License plate detection method and device | |
| CN104361343A (en) | Method and device for identifying vehicle types | |
| CN101930549B (en) | Second generation curvelet transform-based static human detection method | |
| Zheng et al. | Vehicle detection based on morphology from highway aerial images | |
| CN110659550A (en) | Traffic sign recognition method, traffic sign recognition device, computer equipment and storage medium | |
| CN103745197B (en) | A kind of detection method of license plate and device | |
| CN103065163B (en) | A kind of fast target based on static images detects recognition system and method | |
| CN106845458B (en) | Rapid traffic sign detection method based on nuclear overrun learning machine | |
| CN105184291A (en) | Method and system for detecting multiple types of license plates | |
| CN104376303A (en) | Vehicle low-resolution imaging method | |
| CN111898491A (en) | Method and device for identifying reverse driving of vehicle and electronic equipment | |
| CN104268595A (en) | General object detecting method and system | |
| CN104463238B (en) | A kind of automobile logo identification method and system | |
| Lin et al. | Improved traffic sign recognition for in-car cameras | |
| CN106778777B (en) | Vehicle matching method and system |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20150325 |