CN102637257B - Video-based detection and recognition system and method of vehicles - Google Patents
Video-based detection and recognition system and method of vehicles Download PDFInfo
- Publication number
- CN102637257B CN102637257B CN201210078707.5A CN201210078707A CN102637257B CN 102637257 B CN102637257 B CN 102637257B CN 201210078707 A CN201210078707 A CN 201210078707A CN 102637257 B CN102637257 B CN 102637257B
- Authority
- CN
- China
- Prior art keywords
- vehicular traffic
- vehicle
- image
- sample image
- search window
- 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
Landscapes
- Image Analysis (AREA)
- Traffic Control Systems (AREA)
Abstract
The invention discloses a video-based detection and recognition system and a method of vehicles. The system comprises a target characteristic database module, a moving target detection module and a vehicle recognition module. A vehicle sample target characteristic database is built to store data by virtue of the target characteristic database module. After a moving target zone is detected by the moving target detection module, the vehicle recognition module builds a search window and recognizes vehicles. The system and the method resolve the technical problems of poor vehicle distinguishing capability and low distinguishing accuracy of the video-based vehicle detection under the condition of obvious background movement, vehicle adhesion, vehicle occlusion and the like in the prior art. The video-based detection and recognition system and the method of the vehicles have the advantages of excellent vehicle distinguishing capability and high accuracy.
Description
Technical field
The present invention relates to image and process and mode identification technology, specifically a kind of vehicular traffic based on video detects recognition system and method.
Background technology
In recent years, along with the fast development of electronic industry especially novel sensor and mass storage devices, video monitoring technology is with its dirigibility and the comprehensive favor that is more and more subject to people, it is take real-time dynamic information as core, the modern high technologies such as integrated use computing machine, control technology, fast video image is processed in real time flexibly, reached real-time monitoring.Vehicle detection is the important step in traffic system, and for traffic monitoring, traffic control provide information, conventional vehicle checking method comprises coil detection, detections of radar, laser detection etc.Installation and maintenance engineering amount that coil detects are large, destroy road surface, affect life-span of road, and laser detection and detections of radar not only cost is high, and easily human body is worked the mischief.
Vehicle detection recognition technology based on video can extract information of vehicles from video image, not only flexible, does not destroy road surface, and can provide a large amount of detection information for traffic monitoring, for traffic administration provides visual information.Vehicle detection recognition technology based on video, as emerging vehicle checking method, receives people's concern day by day.
Chinese patent literature CN101719217A has proposed a kind of model recognition system and method based on relaxation algorithm, this system comprises the video camera connected with data line, image pick-up card, image processing and display device, the method is first to use camera acquisition automobile video frequency, the image gathering is set up based on Gauss's average background model, using this as detecting foreground area; By the poor method of operating of frame, the foreground area at moving target place is cut apart and is extracted as foreground image from background again, with relaxation algorithm to foreground image processing obtain vehicle commander and overall height, then judge vehicle result according to national vehicle criterion of identification.Chinese patent literature CN102332167A discloses the object detection method of vehicle and pedestrian in a kind of intelligent traffic monitoring, sequence of frames of video is carried out to the initialization of background model, independently set up the mixed Gaussian background component model of saturation degree component and luminance component and get component average; Present frame in sequence of frames of video, with background frames phase difference, is carried out removing shade and noise again and carrying out morphologic filtering after binary conversion treatment to prospect frame; Upgrade weights, average and the variance of the component of the mixture Gaussian background model obtaining by the renewal factor; The Jeffrey value of each distribution in the mixture Gaussian background model after the moving target pixel point value that will mate and renewal compares, and utilizes Jeffrey value to judge whether moving target pixel belongs to foreground point.Chinese patent literature CN102222346A discloses a kind of vehicle detection and tracking, first each two field picture in video is set up to Gaussian Background model; Utilize frame difference method to do difference processing to adjacent two frames, obtain rough moving region and stagnant zone; The stagnant zone obtaining is carried out to context update, and moving region is not upgraded; Background image after the renewal of current frame image and acquisition is done to difference, obtain accurate moving region; Utilize each pixel matching process to find out overlapping region to the adjacent two frame moving region images that obtain, and compare overlapping region and given threshold size; If overlapping region is greater than given threshold value, judges whether to occur target and overlap; If so, calculate the length breadth ratio of the first frame moving region in adjacent two frames, by this this moving vehicle of ratio detection and tracking; If not, be judged as same vehicle; If overlapping region is less than given threshold value, the minimum boundary rectangle of obtaining multiple target frames comes correctly to vehicle detection and tracking.
Above-mentioned three pieces of patent documentations, in vehicle detection process, substantially all single pixel modelings based on image, identification vehicle, what utilize is image single-point information, judge that current pixel point is foreground pixel (vehicle pixel) or background pixel (non-vehicle pixel), do not consider the relativity problem between pixel, it is the surface information of image, for example, so in vehicle detection process, there are the following problems in meeting:, slightly rock or video camera slight jitter in the situation that at branch, above-mentioned patent documentation can meet vehicle detection identification, but rock obvious at branch, when the obvious situation of DE Camera Shake, vehicle detection identification based on single pixel modeling can produce a lot of wrong reports, especially abundanter in the marginal information of image, the wrong report producing is disturbed just more, in the time of the situation of vehicle adhesion and occlusion, can not effectively each independent vehicle be distinguished, several vehicle identification can be become to a vehicle and a vehicle identification is become to several vehicles etc., thereby affect the tracking of each independent vehicle, and then it is (illegally retrograde to affect the behavioural analysis of vehicle below, make a dash across the red light, illegal lane change etc.), in addition, for the pedestrian target in traffic image and vehicle target, can not effectively distinguish.
In video identification technology, need to carry out feature extraction to image information, the feature of extracting comprises Harr feature, integration feature and angle character, and Haar feature is divided three classes: edge feature, linear feature, central feature and diagonal line feature, be combined into feature templates.Adularescent and two kinds of rectangles of black in feature templates, and the eigenwert that defines this template be white rectangle pixel and deduct black rectangle pixel and.Integration feature refers to that in rectangular area that the numerical value of any point in image equals to form to this point from the upper left corner of image, there is a pixel sum in institute.Gradient Features has been described the information of the regional area such as edge, the angle point variation of image, has stronger robustness for the variation of illumination, is widely used in target signature description, images match and target detection.
Summary of the invention
For this reason, to be solved by this invention is the existing vehicle detection based on video at background motion significantly, distinguish the technical matters that vehicular traffic ability, resolving accuracy are not high in the situation such as vehicle adhesion and occlusion, provides a kind of and has good vehicle separating capacity, the high-precision vehicular traffic based on video detects recognition system and method.
For solving the problems of the technologies described above, the technical solution used in the present invention is as follows:
Vehicular traffic based on video detects a recognition system, comprising: comprising:
(1) target signature database module, store vehicular traffic sample image is carried out after pattern-recognition training, the vehicular traffic sample object property data base obtaining according to normalization type, angular type and type of vehicle, described target signature database module comprises
Sample image pretreatment unit, from traffic video, cut out non-vehicular traffic sample image and described vehicular traffic sample image, described vehicular traffic sample image and described non-vehicular traffic sample image are carried out to image pre-service, and described image pre-service comprises normalized and gray processing processing;
Vehicular traffic sample image taxon, is connected with described sample image pretreatment unit, and the described vehicular traffic sample image after normalized is carried out to angle classification and vehicle class classification;
Sample image feature extraction unit, be connected with described vehicular traffic sample image taxon, described non-vehicular traffic sample image and sorted described vehicular traffic sample image are extracted to its gray processing characteristics of image after treatment, and described characteristics of image comprises integration feature, Haar feature and Gradient Features;
Sample image pattern-recognition training unit, be connected with described sample image feature extraction unit, described integration feature to described vehicular traffic sample image and described non-vehicular traffic sample image, described Haar feature and described Gradient Features carry out pattern-recognition training, and described pattern-recognition training adopts Adaboost algorithm or SVM algorithm;
Target signature database storage unit, is connected with described sample image pattern-recognition training unit, stores described sample image pattern-recognition training unit and obtains vehicular traffic sample object property data base according to normalization type, angular type and type of vehicle;
(2) moving object detection module with traffic video equipment connection, detects motion target area from the traffic video image sequence obtaining;
(3) vehicular traffic identification module, carry out data transmission with described moving object detection module, load described vehicular traffic sample object property data base, the described motion target area that described moving object detection Module recognition is gone out carries out vehicle identification, and described vehicular traffic identification module comprises
Vehicle search window unit arranges a vehicle search window in described motion target area;
Search window feature extraction unit, is connected with described vehicle search window unit, obtains integration feature, Haar feature and the Gradient Features of the gray level image in described vehicle search window;
Comparing unit, be connected with described search window feature extraction unit and described target signature database storage unit, described vehicular traffic sample object property data base in described integration feature, described Haar feature and described Gradient Features and the described target signature database storage unit that described search window feature extraction unit is obtained is compared, and the vehicle in described vehicle search window is identified.
The vehicular traffic that uses the described vehicular traffic based on video to detect recognition system detects a recognition methods, comprises the steps:
1. set up vehicular traffic sample object property data base
First, from traffic video, cut out vehicular traffic sample image and non-vehicular traffic sample image, by sample image pretreatment unit, described vehicular traffic sample image and described non-vehicular traffic sample image are carried out to image pre-service, described image pre-service comprises normalized and gray processing processing;
Then, described vehicular traffic sample image by vehicular traffic sample image taxon after to normalized carries out angle classification and vehicle class is classified, and by sample image feature extraction unit, described non-vehicular traffic sample image and sorted described vehicular traffic sample image being extracted to its gray processing characteristics of image after treatment, described characteristics of image comprises integration feature, Haar feature and Gradient Features;
Finally, call described integration feature, described Haar feature and the described Gradient Features of sample image pattern-recognition training unit to described vehicular traffic sample image and described non-vehicular traffic sample image and carry out pattern-recognition training, described pattern-recognition training adopts Adaboos t algorithm or SVM algorithm, obtain vehicular traffic sample object property data base according to normalization type, angular type and type of vehicle, and the described vehicular traffic sample object property data base obtaining is stored in to described target signature database storage unit;
2. read traffic video image sequence, from described traffic video image sequence, detect motion target area;
3. load described vehicular traffic sample object property data base, described motion target area is carried out to vehicle identification: first described in each, in motion target area, define a vehicle search window by vehicle search window unit, then call integration feature, Haar feature and Gradient Features that search window feature extraction unit obtains the gray level image in described vehicle search window; Finally call comparing unit described integration feature, Haa r feature and Gradient Features and described vehicular traffic sample object property data base are compared, the vehicle in described vehicle search window is identified.
Described vehicular traffic detects recognition methods, described step 2. in, described in detect motion target area process comprise the steps:
(1) read traffic video image sequence, adopt mixed Gauss model, described traffic video image sequence is carried out to background modeling, obtain the bianry image to described traffic video image sequence;
(2) described bianry image is carried out to morphologic filtering, then use connected domain analysis to carry out target area demarcation, obtain some motion target areas;
(3), for described motion target area, in the time that the single pixel of traffic video image is in described motion target area, described in this, pixel does not carry out context update process, otherwise carries out context update process.
Described vehicular traffic detects recognition methods, described step 3. in, when described vehicle search window and described vehicular traffic sample object property data base are compared, according to the sample image size of the described vehicular traffic sample image after normalized and described non-vehicular traffic sample image, described vehicle search window is carried out to proportional zoom, to the integration feature of the gray level image in described vehicle search window, Haar feature and Gradient Features mate with described vehicular traffic sample object property data base, if matching degree is less than matching threshold, think that described vehicle search window is non-vehicle region, otherwise, think that described vehicle search window is vehicle region, obtain angular type and the type of vehicle of coupling.
Described vehicular traffic detects recognition methods, also comprises the step that described traffic video image sequence is carried out to embossment computing in described step (1).
Described vehicular traffic detects recognition methods, described step 1. described in vehicular traffic sample image and described non-vehicular traffic sample image normalized obtain 24 × 24,32 × 32,48 × 48 3 kinds of big or small sample images.
Described vehicular traffic detects recognition methods, described step 1. described in angle classification refer to according to body sway angle in described vehicular traffic sample image and classify, be divided into :-30 ° to+30 ° ,-25 ° to-55 ° and+25 ° to+55 ° ,-50 ° to-80 ° and+50 ° to+80 ° three classes.
Described vehicular traffic detects recognition methods, described step 1. described in vehicle class classification refer to described vehicular traffic sample image to be divided into compact car, the large car that does not comprise public transit vehicle and public transit vehicle three classes.Technique scheme of the present invention has the following advantages compared to existing technology:
(1) vehicular traffic based on video of the present invention detects recognition system and method, comprise target signature database module, moving object detection module and vehicular traffic identification module, in described target signature database module, by sample image pretreatment unit, vehicular traffic sample image and non-vehicular traffic sample image are normalized and gray processing processing, by vehicular traffic sample image taxon, pretreated vehicle sample image is carried out to angle classification and vehicle class classification, by sample image feature extraction unit, non-vehicular traffic sample image and sorted vehicular traffic sample image are carried out to feature extraction, then by sample image pattern-recognition training unit, the described feature of extracting is trained and pattern-recognition, set up vehicular traffic sample object property data base and store.When moving object detection module detects after motion target area, set up vehicle search window and carry out vehicle identification by vehicular traffic identification module.Setting up in vehicular traffic sample object property data base process, vehicular traffic sample image is classified according to normalization kind, angle kind and vehicle class, then carry out feature extraction, improve accuracy of identification and operation time, in the training process of vehicle sample, the present invention utilizes Adaboost algorithm for pattern recognition or SVM algorithm identified vehicular traffic, this algorithm in vehicular traffic image between each pixel correlativity take into account, what Adaboost algorithm for pattern recognition was identified utilization to vehicle is the surface information of image, dot information and surface information are organically combined, can overcome branch rocks, the interference that the situations such as DE Camera Shake and vehicle adhesion are brought, can distinguish for the non-vehicle target (as pedestrian) in traffic image, can be better, more effectively carry out vehicle detection identification.Carrying out in the process of vehicle identification, by definition vehicle search window, described motion target area is searched for, can carry out setting search step-length according to the size of motion target area, guarantee speed and the precision of search, improve search efficiency and search quality.Vehicular traffic based on video of the present invention detects recognition system and method, by setting up vehicular traffic target signature mathematical model, research vehicle target Feature Extraction Technology, set up vehicular traffic sample object property data base, the automatic renewal technology of research background, the pure video of realizing traffic events and traffic behavior detects, it calculates accurately, judges by accident that few feature provides accurately, practicality, technological means intuitively, the road conditions that can adapt to well urban transportation more complicated, have high practicality.
(2) vehicular traffic of the present invention detects recognition methods, in the described process that detects motion target area, adopt mixed Gauss model to carry out background modeling to traffic video image sequence, the bianry image obtaining is carried out to morphologic filtering and connected domain analysis, for described motion target area, in the time that the single pixel of traffic video image is in described motion target area, described in this, pixel does not carry out context update process, otherwise carry out context update process, improved the accuracy of detection of motion target area.Carry out morphologic filtering by the bianry image to after Gaussian modeling, can reduce picture noise, contribute to obtain motion target area.Adopt the region of this detection moving target, the method that background modeling and Adaboost pattern-recognition can be combined, Adaboost pattern-recognition does not need entire image to search for, only need after mixed Gaussian background modeling, extract in area-of-interest and search for identification vehicle, greatly improve arithmetic speed; The area-of-interest that extracts at Gaussian modeling is smaller, when Adaboost identification, can reduce the step-length of search, improve precision; The area-of-interest that extracts at Gaussian modeling is larger, Adaboost can expand step-size in search when identification a little, realize the adjustment of News Search step-length, and no longer adopt single fixing scaling, in the situation that guaranteeing arithmetic speed, improve precision, simultaneously in vehicular traffic target in situation at a distance, still can detect accurately identification, can overcome the unfavorable factors such as illumination, shade, leaf are rocked, DE Camera Shake, just can solve preferably vehicle adhesion and occlusion.
(3) vehicular traffic of the present invention detects recognition methods, when described vehicle search window and described vehicular traffic sample object property data base are compared, according to the sample image size of the described vehicular traffic sample image after normalized and described non-vehicular traffic sample image, described vehicle search window is carried out to proportional zoom.First, choose less vehicle search window and carry out feature extraction and vehicle identification, then progressively expand vehicle search window and carry out again feature extraction and vehicle identification, vehicle search window is until expand motion target area size to, vehicle identification finishes, and finally obtains the angular type of vehicle identification and the kind type of vehicle.In this vehicle identification process, in the time of the distant situation of video scene, can choose less step-size in search, when the closer situation of video scene, can choose the larger step-length of searching, the accuracy that has improved vehicle identification like this in improving the recognition efficiency of vehicle.
(4) vehicular traffic of the present invention detects recognition methods, and traffic video image sequence is carried out to embossment computing, makes the dark area in image increase some brightness, has reduced the shading value of image, contributes to Gaussian modeling.
(5) vehicular traffic of the present invention detects recognition methods, described step 1. described in vehicular traffic sample image and described non-vehicular traffic sample image normalized obtain 24 × 24,32 × 32,48 × 48 3 kinds of big or small sample images, these three kinds of image patterns are evenly distributed in the size of the video image obtaining, and contribute to later vehicle sample training and video image is carried out to vehicle identification.
(6) vehicular traffic of the present invention detects recognition methods, described step 1. described in angle classification refer to according to body sway angle in described vehicular traffic sample image and classify, be divided into :-30 ° to+30 ° ,-25 ° to-55 ° and+25 ° to+55 ° ,-50 ° to-80 ° and+50 ° to+80 ° three classes.Adopt this mode classification, both can comprehensive vehicle obliquity, react the characteristics of image of all angles comprehensively, reduce again the kind of angle as far as possible, thereby reduce the data type in vehicular traffic sample object property data base, improve the matching speed in identifying.
(7) vehicular traffic of the present invention detects recognition methods, described step 1. described in vehicle class classification refer to described vehicular traffic sample image to be divided into compact car, the large car that does not comprise public transit vehicle and public transit vehicle three classes, because this three classes vehicle has not only been summarized the type of most vehicles, and the data message of these vehicles detects the more information that provides for traffic administration and road conditions, therefore vehicle class is divided to three classes for this reason, the classification while being convenient to vehicle identification.
(8) vehicular traffic of the present invention detects recognition methods, the training process of vehicle sample training need to be collected a large amount of vehicular traffic sample images and non-vehicular traffic sample image, non-vehicular traffic sample image major part derives from the non-vehicular traffic part in traffic video image, vehicular traffic sample image is as positive sample, non-vehicular traffic sample image is as negative sample, positive sample is included in the training aids of vehicle detection together with negative sample, obtain the car modal of stronger feature by training, can effectively must distinguish vehicle image and non-vehicle image, improve the precision of vehicle identification.
Accompanying drawing explanation
For content of the present invention is more likely to be clearly understood, below according to a particular embodiment of the invention and by reference to the accompanying drawings, the present invention is further detailed explanation, wherein
Fig. 1 is that the vehicular traffic based on video of one embodiment of the invention detects recognition system structured flowchart;
Fig. 2 be shown in Fig. 1 based on the vehicular traffic of video detect recognition system recognition methods set up vehicular traffic sample object property data base process flow diagram;
Fig. 3 is the moving object detection process flow diagram that detects the recognition methods of recognition system shown in Fig. 1 based on the vehicular traffic of video;
Fig. 4 be shown in Fig. 1 based on the vehicular traffic of video detect recognition system recognition methods motion target area is carried out to vehicle identification process flow diagram;
Fig. 5 and Fig. 6 are the simulated effect figure that uses system embodiment illustrated in fig. 1 one two field picture to be carried out to moving region detection;
Fig. 7 and Fig. 8 are the simulated effect figure that uses system embodiment illustrated in fig. 1 to identify image.
In figure, Reference numeral is expressed as: 1-moving object detection module; 2-target signature database module; 21-sample image pretreatment unit; 22-vehicular traffic sample image taxon; 23-sample image feature extraction unit; 24-sample image pattern-recognition training unit; 25-target signature database storage unit; 3-vehicular traffic identification module; 31-vehicle search window unit; 32-search window feature extraction unit; 33-comparing unit.
Embodiment
Embodiment 1:
Provide the concrete embodiment that a kind of vehicular traffic based on video of the present invention detects recognition system below, shown in Figure 1, the vehicular traffic based on video in the present embodiment detects recognition system, comprising:
(1) target signature database module 2, store vehicular traffic sample image is carried out after pattern-recognition training, the vehicular traffic sample object property data base obtaining according to normalization type, angular type and type of vehicle, comprise sample image pretreatment unit 21, vehicular traffic sample image taxon 22, sample image feature extraction unit 23, sample image pattern-recognition training unit 24 and target signature database storage unit 25, respectively said units is described in detail below.
Sample image pretreatment unit 21, from traffic video, cut out non-vehicular traffic sample image and described vehicular traffic sample image, described vehicular traffic sample image and described non-vehicular traffic sample image are carried out to image pre-service, and described image pre-service comprises normalized and gray processing processing.In the present embodiment, take to cut out 8000 vehicular traffic sample images and 12000 non-vehicular traffic sample images from traffic video, 8000 described vehicular traffic sample images and 12000 described non-vehicular traffic sample images are normalized and gray processing processing, and described normalized obtains 24 × 24,32 × 32,48 × 48 3 kinds of big or small sample images.
Vehicular traffic sample image taxon 22, is connected with described sample image pretreatment unit 21, and the described vehicular traffic sample image after normalized is carried out to angle classification and vehicle class classification.The classification of described angle refers to according to body sway angle in described vehicular traffic sample image classifies (take scene visual direction as 0 ° of direction), in the present embodiment, is divided into :-30 ° to+30 ° ,-25 ° to-55 ° and+25 ° to+55 ° ,-50 ° to-80 ° and+50 ° to+80 ° three classes.Described vehicle class classification is divided into described vehicular traffic sample image compact car, the large car that does not comprise public transit vehicle and public transit vehicle three classes in the present embodiment.
Sample image feature extraction unit 23, be connected with described vehicular traffic sample image taxon 22, described non-vehicular traffic sample image and sorted described vehicular traffic sample image are extracted to its gray processing integration feature after treatment, Haar feature and Gradient Features.
Integration feature refer to institute in the rectangular area that the numerical value of any point in image equals to form to this point from the upper left corner of image have a pixel with.Haar feature is divided three classes: edge feature, linear feature, central feature and diagonal line feature, be combined into feature templates.Adularescent and two kinds of rectangles of black in feature templates, and the eigenwert that defines this template be white rectangle pixel and deduct black rectangle pixel and.Determining that the quantity of Harr-like feature after characteristic formp just depends on the size of training sample image matrix, feature templates is placed arbitrarily in subwindow, a kind of form is called a kind of feature, and the feature of finding out all subwindows is the basis of carrying out weak typing training.Gradient Features has been described the information of the regional area such as edge, the angle point variation of image, has stronger robustness for the variation of illumination, is widely used in target signature description, images match and target detection.The integration feature of image, Haar feature, Gradient Features have detailed explanation and concrete preparation method in recognition of face, the identification of head shoulder of OpenCV the inside, adopt method of the prior art can obtain integration feature, Haar feature and the Gradient Features of vehicular traffic sample image.The full name of OpenCV is Open Source Computer Vision Library, and OpenCV was set up by Intel in 1999, was provided support by Willow Garage now.OpenCV is a cross-platform computer vision storehouse based on BSD licence mandate (increasing income) distribution, may operate in Linux, Windows and Mac OS operating system.Its lightweight and efficient---formed by a series of C functions and a small amount of C++ class, the interface of the language such as Python, Ruby, MATLAB is provided simultaneously, realized a lot of general-purpose algorithms of image processing and computer vision aspect.Integration feature, Haar feature, the Gradient Features of the method providing in OpenCV in can Direct Recognition image.
Sample image pattern-recognition training unit 24, be connected with described sample image feature extraction unit 23, the described integration feature to described vehicular traffic sample image and described non-vehicular traffic sample image, described Haar feature and described Gradient Features carry out Adaboost algorithm pattern recognition training.Adaboost is a kind of general-purpose algorithm that the Weak Classifier that is better than arbitrarily random conjecture is combined into strong classifier.Viola combines with Haar rectangular characteristic theoretical Adaboost at first, being applied to face detects, realized real-time face detection, the method is that multiple Weak Classifiers based on single feature are cascaded into a strong classifier, then multiple strong classifiers is cascaded into an object detector completing.The training process of AdaBoost vehicle sample is substantially consistent with the face training process of OpenCV the inside in the present embodiment.
Target signature database storage unit 25, be connected with described sample image pattern-recognition training unit 24, store described sample image pattern-recognition training unit 24 and obtain 27 kinds of vehicular traffic sample object property data bases according to normalization type, angular type and type of vehicle.
(2) moving object detection module 1, with traffic video equipment connection, traffic video equipment herein can be for being arranged on video capture device on crossing or section as camera, video camera etc., also can, for the video equipment of recording being collected, from the traffic video image sequence obtaining, detect motion target area.
(3) vehicular traffic identification module 3, carry out data transmission with described moving object detection module 1, load 27 kinds of described vehicular traffic sample object property data bases that described target signature database storage unit 25 is stored, the described motion target area that described moving object detection module 1 is identified carries out vehicle identification.
Described vehicular traffic identification module 3 comprises: vehicle search window unit 31 arranges a vehicle search window in described motion target area.
Search window feature extraction unit 32, is connected with described vehicle search window unit 31, obtains integration feature, Haar feature and the Gradient Features of the gray level image in described vehicle search window.
Comparing unit 33, be connected with described search window feature extraction unit 32 and described target signature database storage unit 25, described vehicular traffic sample object property data base in the described integration feature that described search window feature extraction unit 32 is obtained, described Haar feature and described Gradient Features and described target signature database storage unit 25 is compared, and the vehicle in described vehicle search window is identified.
The vehicular traffic that uses the vehicular traffic based on video described in the present embodiment to detect recognition system detects a recognition methods, comprises the steps:
1. set up vehicular traffic sample object property data base, as shown in Figure 2:
S01: cut out 8000 vehicular traffic sample images and 12000 non-vehicular traffic sample images from traffic video, by sample image pretreatment unit 21, described vehicular traffic sample image and described non-vehicular traffic sample image are normalized and gray processing processing, described normalized obtains 24 × 24,32 × 32,48 × 48 3 kinds of big or small sample images.
S02: the described vehicular traffic sample image after normalized is carried out to angle classification and vehicle class classification, calling vehicular traffic sample image taxon 22 classifies, described angle classification refers to according to body sway angle in described vehicular traffic sample image classifies, be divided into :-30 ° to+30 ° ,-25 ° to-55 ° and+25 ° to+55 ° ,-50 ° to-80 ° and+50 ° to+80 ° three classes, the classification of described vehicle class refers to described vehicular traffic sample image to be divided into compact car, the large car that does not comprise public transit vehicle and public transit vehicle three classes.After classification, by sample image feature extraction unit 23, described non-vehicular traffic sample image and sorted described vehicular traffic sample image are extracted to its gray processing integration feature after treatment, Haar feature and Gradient Features.
S03: integration feature, Haar feature and Gradient Features to described vehicular traffic sample image and described non-vehicular traffic sample image carry out Adaboost algorithm pattern recognition training, obtain 27 kinds of vehicular traffic sample object property data bases according to normalization type, angular type and type of vehicle, herein 27 kinds is to combine 27 kinds that (3 × 3 × 3) obtain according to 3 kinds of normalization types, 3 kinds of angular type and 3 kinds of type of vehicle.
2. read traffic video image sequence, from described traffic video image sequence, detect motion target area.Can adopt moving target detecting method in prior art to realize herein, for example movement technique based on code book.
3. load described vehicular traffic sample object property data base, described motion target area is carried out to vehicle identification, as shown in Figure 4:
D01: define a vehicle search window described in each in motion target area;
D02: the integration feature, Haar feature and the Gradient Features that obtain the gray level image in described vehicle search window;
D03: described integration feature, Haar feature and Gradient Features in described vehicle search window and described vehicular traffic sample object property data base are compared, vehicle in described vehicle search window is identified, obtain angle and the type of vehicle of the auto model that matching degree is the highest, as vehicle angles and type of vehicle in described vehicle search window.
In the present embodiment, wherein the training of vehicular traffic sample is according to reducing vehicular traffic sample image normalization size 24 × 24, 32 × 32, 48 × 48 types, angular type-30 ° are to+30 ° of one class,-25 ° to-55 ° and+25 ° to+55 ° one classes,-50 ° to-80 ° and+50 ° to+80 ° one classes, type of vehicle compact car, large car (not comprising public transit vehicle), public transit vehicle carries out classification based training jointly, according to normalization size, angle, type of vehicle is divided into altogether 27 types, then train respectively, obtain altogether 27 kinds of vehicular traffic target signature database models.The embodiment that can convert the most, normalization type can be set to other types according to conventional image size, several as in the types such as 24 × 24,28 × 28,32 × 32,36 × 36,48 × 48,56 × 56.Angular type also can be set to as required more multiclass and segment, as-20 ° to+20 ° one classes ,-20 ° to-50 ° and+20 ° to+50 ° one classes ,-50 ° to-80 ° and+50 ° to+80 ° one classes, be divided into 3-6 class can, arrange as required.Type of vehicle also can be divided into other some types as types such as increase shipping is called a taxi, minibuses, and the type of vehicle separating as required arranges.
Embodiment 2:
On the basis of above-described embodiment, the step that described vehicular traffic detects recognition methods can also adopt specific method to detect motion target area in 2..The size that described step can be approached by expanding successively the mode of search window motion target area in 3..Concrete step is as follows:
2. read traffic video image sequence, the process that detects motion target area from described traffic video image sequence is as follows, shown in Figure 3:
T01: read traffic video image sequence, carry out embossment computing, the principle of embossment computing is a pixel of image to be adjacent to upper left pixel carry out difference computing and add a constant (in order to keep the gray scale of image, the constant here can be set to 128), make dark area increase some brightness.Carry out then adopting mixed Gauss model after embossment computing, described traffic video image sequence is carried out to background modeling, obtain the bianry image to described traffic video image sequence.
Mixed Gaussian background modeling basic thought is: first define the color of K each pixel of state representation, wherein the value of K is generally between 3～5, and K value is larger, processes fluctuation ability stronger, and speed is slower.Each state in K represents with a gaussian kernel function, the pixel value of these state part expression backgrounds, and remainder represents the pixel value of sport foreground.As each pixel color value variable X
trepresent, its probability density function can be represented by following K three-dimensional Gaussian function:
η (x, μ
i, t, ∑
i, t) be i the Gaussian distribution in t moment, its average is μ
i, t, covariance matrix is ∑
i, t=(σ
i, t) 2, ω
i, tbe the weight of i Gaussian distribution, and have
wherein
i＝1，2，...，K (2)
In formula, n represents X
tdimension.
Along with the continuous variation of scene, the mixed Gauss model of each pixel needs constantly to learn to upgrade, and method is by the K in mixed Gauss model gauss component ω
i, t/ σ
i, tdescending sequence, then uses current pixel value X
tcompare one by one with K gauss component in mixed Gauss model, if current pixel value X
twith the average μ in i gauss component
i, tbetween difference be less than this gauss component standard deviation sigma
i, t2.5 to 3.5 times, mixed Gauss model is upgraded, renewal equation is as follows:
ω
i，t+1＝(1-α)ω
i，t-αM
i，t (3)
μ
i，t+1＝(1-ρ)μ
i，t-ρx
i (4)
(σ
i.t+1)
2＝(1-ρ)(σ
i，t)
2-ρ(x
i-μ
i，t)
T(x
i-μ
i，t) (5)
ρ＝α/ω
i，t (6)
Wherein α is the learning rate of mixed Gauss model, as i gauss component and X
twhen coupling, M
i, tbe 1, otherwise be 0; If current pixel value X
twhile coupling with K gauss component in mixed Gauss model, in this pixel mixed Gauss model, coming last gauss component is replaced by new gauss component, the average of new gauss component replaces with current pixel value, poor and the weight of primary standard uses as default, after renewal completes, the weights of each gauss component are normalized again.
In to the process of traffic video image Gauss modeling, for mixed Gauss model, carry out discrete digital processing, so greatly optimize the arithmetic speed of mixed Gaussian, according to the emulation of actual scene, in the present embodiment, set up the mixture Gaussian background model of K=3.
T02: described bianry image is carried out to morphologic filtering, then use connected domain analysis to carry out target area demarcation, obtain some motion target areas;
T03: for described motion target area, in the time that the single pixel of traffic video image is in described motion target area, described in this, pixel does not carry out context update process, otherwise carries out context update process.
3. load described vehicular traffic sample object property data base, described motion target area carried out to vehicle identification:
D01: define a vehicle search window described in each in motion target area;
D02: the integration feature, Haar feature and the Gradient Features that obtain the gray level image in described vehicle search window;
D03: when described vehicle search window and described vehicular traffic sample object property data base are compared, according to the sample image size of the described vehicular traffic sample image after normalized and described non-vehicular traffic sample image, described vehicle search window is carried out to proportional zoom, integration feature, Haar feature and the Gradient Features to the gray level image in described vehicle search window mates with described vehicular traffic sample object property data base.That is to say and in each motion target area, define a vehicle search window, described vehicle search window increases automatically according to certain step-length, according to the size of described vehicle search window, automatically with 27 vehicular traffic sample object property data bases in sample type be that 24 × 24,32 × 32,48 × 48 target signature data contrast identification.If the match is successful in this process, degree is less than matching threshold, thinks that described vehicle search window is non-vehicle region, otherwise, think that described vehicle search window is vehicle region, obtain angular type and the type of vehicle of coupling.
Vehicular traffic of the present invention detects recognition methods, to adopt the method that background modeling and pattern-recognition are combined, the vehicle that detection can be recognized is classified according to angle and type of vehicle, vehicular traffic target is in remote situation simultaneously, still can detect and recognize accurately, improve accuracy of identification, can overcome the unfavorable factors such as illumination, shade, leaf are rocked, DE Camera Shake simultaneously, just can solve preferably vehicle adhesion and occlusion.
Fig. 5 and Fig. 6 are that the vehicular traffic described in use above-described embodiment detects recognition system is carried out moving region detection simulated effect figure to a two field picture, wherein in the wire frame in Fig. 6, represent it is motion detection regional extent, Fig. 5 uses mixed Gauss model to carry out the two-value classification results that modeling obtains to traffic video image sequence, Fig. 6 carries out target to Fig. 5 to demarcate two rectangle frames that processing obtains, be respectively vehicle and pedestrian, also just had two motion target areas---vehicle region and pedestrian region here.
Fig. 7 and Fig. 8 are that the vehicular traffic described in use above-described embodiment detects the simulated effect figure that recognition system is identified image, Fig. 7 is that the motion target area that Fig. 5 is obtained carries out vehicle identification, here face has filtered out pedestrian, is only left our interested vehicular traffic; Fig. 8 is that another one time point recognizes vehicular traffic, has identified two vehicular traffic targets, and one is compact car, and another one is large car.
The embodiment that can convert as other, described Adaboost algorithm can replace with SVM algorithm.SVM (support vector machine support vector machine) is a kind of trainable machine learning method, the main thought of SVM may be summarized to be 2 points: it is to analyze for linear separability situation for (1), for the situation of linearly inseparable, make its linear separability by using non-linear map that the sample of low-dimensional input space linearly inseparable is converted into high-dimensional feature space, become possibility thereby make high-dimensional feature space adopt linear algorithm to carry out linear analysis to the nonlinear characteristic of sample; (2) it based on structural risk minimization theory in feature space construction optimum segmentation lineoid, make learner obtain global optimization, and meet certain upper bound in the expected risk of whole sample space with certain probability.Because SVM Algorithm for Training speed is fast, but vehicle identification speed is slow, now needs a special hardware accelerator support; Other same above-described embodiment.Can realize equally object of the present invention, belong to protection scope of the present invention.
Obviously, above-described embodiment is only for example is clearly described, and the not restriction to embodiment.For those of ordinary skill in the field, can also make other changes in different forms on the basis of the above description.Here without also giving exhaustive to all embodiments.And the apparent variation of being extended out thus or variation are still among the protection domain in the invention.
Claims (1)
1. the vehicular traffic based on video detects a recognition system, it is characterized in that, comprising:
(1) target signature database module, store vehicular traffic sample image is carried out after pattern-recognition training, the vehicular traffic sample object property data base obtaining according to normalization type, angular type and type of vehicle, described target signature database module comprises
Sample image pretreatment unit, from traffic video, cut out non-vehicular traffic sample image and described vehicular traffic sample image, described vehicular traffic sample image and described non-vehicular traffic sample image are carried out to image pre-service, and described image pre-service comprises normalized and gray processing processing;
Vehicular traffic sample image taxon, be connected with described sample image pretreatment unit, described vehicular traffic sample image after normalized is carried out to angle classification and vehicle class classification, and described vehicle class classification refers to described vehicular traffic sample image to be divided into compact car, the large car that does not comprise public transit vehicle and public transit vehicle three classes;
Sample image feature extraction unit, be connected with described vehicular traffic sample image taxon, described non-vehicular traffic sample image and sorted described vehicular traffic sample image are extracted to its gray processing characteristics of image after treatment, and described characteristics of image comprises integration feature, Haar feature and Gradient Features;
Sample image pattern-recognition training unit, be connected with described sample image feature extraction unit, described integration feature to described vehicular traffic sample image and described non-vehicular traffic sample image, described Haar feature and described Gradient Features carry out pattern-recognition training, and described pattern-recognition training adopts Adaboost algorithm or SVM algorithm;
Target signature database storage unit, is connected with described sample image pattern-recognition training unit, stores described sample image pattern-recognition training unit and obtains vehicular traffic sample object property data base according to normalization type, angular type and type of vehicle;
(2) moving object detection module with traffic video equipment connection, detects motion target area from the traffic video image sequence obtaining, and process comprises:
Read traffic video image sequence, adopt mixed Gauss model, described traffic video image sequence is carried out to background modeling, obtain the bianry image to described traffic video image sequence;
Described bianry image is carried out to morphologic filtering, then use connected domain analysis to carry out target area demarcation, obtain some motion target areas;
For described motion target area, in the time that the single pixel of traffic video image is in described motion target area, described in this, pixel does not carry out context update process, otherwise carries out context update process;
(3) vehicular traffic identification module, carry out data transmission with described moving object detection module, load described vehicular traffic sample object property data base, the described motion target area that described moving object detection Module recognition is gone out carries out vehicle identification, and described vehicular traffic identification module comprises
Vehicle search window unit arranges a vehicle search window in described motion target area;
Search window feature extraction unit, is connected with described vehicle search window unit, obtains integration feature, Haar feature and the Gradient Features of the gray level image in described vehicle search window;
Comparing unit, be connected with described search window feature extraction unit and described target signature database storage unit, described vehicular traffic sample object property data base in described integration feature, described Haar feature and described Gradient Features and the described target signature database storage unit that described search window feature extraction unit is obtained is compared, and the vehicle in described vehicle search window is identified.
2. right to use requires the vehicular traffic of the detection of the vehicular traffic based on the video recognition system described in 1 to detect a recognition methods, it is characterized in that, comprises the steps:
First, from traffic video, cut out vehicular traffic sample image and non-vehicular traffic sample image, by sample image pretreatment unit, described vehicular traffic sample image and described non-vehicular traffic sample image are carried out to image pre-service, described image pre-service comprises normalized and gray processing processing;
Then, described vehicular traffic sample image by vehicular traffic sample image taxon after to normalized carries out angle classification and vehicle class is classified, described vehicle class classification refers to described vehicular traffic sample image to be divided into compact car, the large car that does not comprise public transit vehicle and public transit vehicle three classes, and by sample image feature extraction unit, described non-vehicular traffic sample image and sorted described vehicular traffic sample image being extracted to its gray processing characteristics of image after treatment, described characteristics of image comprises integration feature, Haar feature and Gradient Features;
Finally, call described integration feature, described Haar feature and the described Gradient Features of sample image pattern-recognition training unit to described vehicular traffic sample image and described non-vehicular traffic sample image and carry out pattern-recognition training, described pattern-recognition training adopts Adaboost algorithm or SVM algorithm, obtain vehicular traffic sample object property data base according to normalization type, angular type and type of vehicle, and the described vehicular traffic sample object property data base obtaining is stored in to described target signature database storage unit;
2. read traffic video image sequence, from described traffic video image sequence, detect motion target area, described in detect motion target area process comprise the steps:
(1) read traffic video image sequence, adopt mixed Gauss model, described traffic video image sequence is carried out to background modeling, obtain the bianry image to described traffic video image sequence;
(2) described bianry image is carried out to morphologic filtering, then use connected domain analysis to carry out target area demarcation, obtain some motion target areas;
(3), for described motion target area, in the time that the single pixel of traffic video image is in described motion target area, described in this, pixel does not carry out context update process, otherwise carries out context update process;
3. load described vehicular traffic sample object property data base, described motion target area is carried out to vehicle identification: first described in each, in motion target area, define a vehicle search window by vehicle search window unit, then call integration feature, Haar feature and Gradient Features that search window feature extraction unit obtains the gray level image in described vehicle search window; Finally call comparing unit described integration feature, Haar feature and Gradient Features and described vehicular traffic sample object property data base are compared, the vehicle in described vehicle search window is identified.
3. vehicular traffic according to claim 2 detects recognition methods, it is characterized in that: described step 3. in, when described vehicle search window and described vehicular traffic sample object property data base are compared, according to the sample image size of the described vehicular traffic sample image after normalized and described non-vehicular traffic sample image, described vehicle search window is carried out to proportional zoom, to the integration feature of the gray level image in described vehicle search window, Haar feature and Gradient Features mate with described vehicular traffic sample object property data base, if matching degree is less than matching threshold, think that described vehicle search window is non-vehicle region, otherwise, think that described vehicle search window is vehicle region, obtain angular type and the type of vehicle of coupling.
4. vehicular traffic according to claim 3 detects recognition methods, it is characterized in that: in described step (1), also comprise the step that described traffic video image sequence is carried out to embossment computing.
5. detect recognition methods according to the vehicular traffic described in any one in claim 2-4, it is characterized in that: described step
described in vehicular traffic sample image and described non-vehicular traffic sample image normalized obtain 24 × 24,32 × 32,48 × 48 3 kinds of big or small sample images.
6. vehicular traffic according to claim 5 detects recognition methods, it is characterized in that: described step
described in angle classification refer to according to body sway angle in described vehicular traffic sample image and classify, be divided into :-30
0to+30
0,-25
0to-55
0and+25
0to+55
0,-50
0to-80
0and+50
0to+80
0three classes.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201210078707.5A CN102637257B (en) | 2012-03-22 | 2012-03-22 | Video-based detection and recognition system and method of vehicles |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201210078707.5A CN102637257B (en) | 2012-03-22 | 2012-03-22 | Video-based detection and recognition system and method of vehicles |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN102637257A CN102637257A (en) | 2012-08-15 |
| CN102637257B true CN102637257B (en) | 2014-07-02 |
Family
ID=46621645
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201210078707.5A Expired - Fee Related CN102637257B (en) | 2012-03-22 | 2012-03-22 | Video-based detection and recognition system and method of vehicles |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN102637257B (en) |
Families Citing this family (49)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102902955B (en) * | 2012-08-30 | 2016-10-19 | 中国科学技术大学 | The intelligent analysis method of a kind of vehicle behavior and system |
| CN103116986B (en) * | 2013-01-21 | 2014-12-10 | 信帧电子技术（北京）有限公司 | Vehicle identification method |
| CN104123532B (en) * | 2013-04-28 | 2017-05-10 | 浙江大华技术股份有限公司 | Target object detection and target object quantity confirming method and device |
| CN103390145B (en) * | 2013-06-30 | 2016-05-25 | 安科智慧城市技术（中国）有限公司 | A kind of target area vehicle checking method and system |
| CN103310207B (en) * | 2013-07-06 | 2016-03-09 | 中国科学技术大学 | A kind of vehicle using motor detection method based on many Gauss models |
| CN103455795B (en) * | 2013-08-27 | 2017-03-29 | 西北工业大学 | A kind of method of the determination traffic target region based on traffic video data image |
| CN103530366B (en) * | 2013-10-12 | 2016-11-23 | 湖北微模式科技发展有限公司 | A kind of vehicle search method and system based on user-defined feature |
| CN103646544B (en) * | 2013-11-15 | 2016-03-09 | 天津天地伟业数码科技有限公司 | Based on the vehicle behavioural analysis recognition methods of The Cloud Terrace and camera apparatus |
| CN103714697B (en) * | 2013-12-20 | 2015-10-28 | 华南理工大学 | A kind of method of recognition and tracking criminal vehicle |
| CN103810498B (en) * | 2014-02-14 | 2018-04-13 | 广州供电局有限公司 | The monitoring method and system of excavator in scene of constructing |
| CN104268573B (en) * | 2014-09-24 | 2017-12-26 | 深圳市华尊科技股份有限公司 | Vehicle checking method and device |
| CN104463104B (en) * | 2014-11-14 | 2018-06-08 | 武汉工程大学 | A kind of stationary vehicle target rapid detection method and device |
| CN104506800B (en) * | 2014-12-15 | 2018-11-13 | 浙江宇视科技有限公司 | The alert camera scene synthesis of the multi-direction electricity of one kind and comprehensive monitoring and controlling method and device |
| CN105809708A (en) * | 2014-12-31 | 2016-07-27 | 北京智诺英特科技有限公司 | Image tracking method and device |
| CN104951784B (en) * | 2015-06-03 | 2018-10-12 | 贵州警官职业学院 | A kind of vehicle is unlicensed and license plate shading real-time detection method |
| CN104875681A (en) * | 2015-06-16 | 2015-09-02 | 四川长虹佳华信息产品有限责任公司 | Dynamic vehicle-mounted camera control method based on application scenarios |
| US10372977B2 (en) * | 2015-07-09 | 2019-08-06 | Analog Devices Gloval Unlimited Company | Video processing for human occupancy detection |
| CN105844909A (en) * | 2016-02-04 | 2016-08-10 | 彭冬青 | Electric shock prevention distribution box |
| CN105844666B (en) * | 2016-03-25 | 2019-07-12 | 深圳大学 | One kind being based on the matched vehicle speed measuring method of single camera video sequence and its system |
| CN105930803A (en) * | 2016-04-22 | 2016-09-07 | 北京智芯原动科技有限公司 | Preceding vehicle detection method based on Edge Boxes and preceding vehicle detection device thereof |
| CN106295541A (en) * | 2016-08-03 | 2017-01-04 | 乐视控股（北京）有限公司 | Vehicle type recognition method and system |
| CN106384089A (en) * | 2016-09-07 | 2017-02-08 | 中国人民解放军国防科学技术大学 | Human body reliable detection method based on lifelong learning |
| CN106446150B (en) * | 2016-09-21 | 2019-10-29 | 北京数字智通科技有限公司 | A kind of method and device of vehicle precise search |
| CN106846374B (en) * | 2016-12-21 | 2020-10-09 | 大连海事大学 | Method for calculating track of vehicle in multi-camera scene |
| CN106940784A (en) * | 2016-12-26 | 2017-07-11 | 无锡高新兴智能交通技术有限公司 | A kind of bus detection and recognition methods and system based on video |
| CN106846813A (en) * | 2017-03-17 | 2017-06-13 | 西安电子科技大学 | The method for building urban road vehicle image data base |
| CN107730972A (en) * | 2017-05-10 | 2018-02-23 | 西安艾润物联网技术服务有限责任公司 | The method and apparatus that video identification controls banister |
| JP6514736B2 (en) * | 2017-05-17 | 2019-05-15 | 株式会社Ｓｕｂａｒｕ | Outside environment recognition device |
| CN108932857B (en) * | 2017-05-27 | 2021-07-27 | 西门子（中国）有限公司 | Method and device for controlling traffic signal lamp |
| CN108038423B (en) * | 2017-11-22 | 2022-03-04 | 广东数相智能科技有限公司 | Automobile type identification method and device based on image identification |
| CN108108750A (en) * | 2017-12-12 | 2018-06-01 | 清华大学苏州汽车研究院（吴江） | Metric space method for reconstructing based on deep learning and monocular vision |
| CN108009513A (en) * | 2017-12-14 | 2018-05-08 | 合肥寰景信息技术有限公司 | A kind of image processing method of vehicle dynamic behaviour analysis |
| CN108022433A (en) * | 2017-12-14 | 2018-05-11 | 合肥寰景信息技术有限公司 | Road traffic condition analysis system based on cloud database |
| CN110232300A (en) * | 2018-03-05 | 2019-09-13 | 上海汽车集团股份有限公司 | Lane vehicle lane-changing intension recognizing method and system by a kind of |
| CN108537125B (en) * | 2018-03-13 | 2019-02-12 | 北京影谱科技股份有限公司 | Automobile search system based on amendment type image procossing |
| CN108645625B (en) * | 2018-03-21 | 2022-02-18 | 北京纵目安驰智能科技有限公司 | Tail end and side surface combined 3D vehicle detection method, system, terminal and storage medium |
| CN109031262B (en) * | 2018-06-05 | 2023-05-05 | 鲁忠 | Positioning vehicle searching system and method thereof |
| CN109191841A (en) * | 2018-09-17 | 2019-01-11 | 天津中德应用技术大学 | A kind of urban transportation intelligent management system based on raspberry pie |
| WO2020061766A1 (en) * | 2018-09-25 | 2020-04-02 | 西门子股份公司 | Vehicle event detection apparatus and method, and computer program product and computer-readable medium |
| CN109410582B (en) * | 2018-11-27 | 2021-11-16 | 易念科技（深圳）有限公司 | Traffic condition analysis method and terminal equipment |
| CN109556625A (en) * | 2018-11-30 | 2019-04-02 | 努比亚技术有限公司 | Air navigation aid, device, navigation equipment and storage medium based on front windshield |
| CN109559518A (en) * | 2018-12-10 | 2019-04-02 | 安徽四创电子股份有限公司 | A kind of novel intelligent traffic block port based on structured image recognizer |
| CN111915878B (en) * | 2019-05-09 | 2022-11-29 | 奥迪股份公司 | Method and device for predicting road traffic state, computer device and storage medium |
| CN110111577B (en) * | 2019-05-15 | 2020-11-27 | 武汉纵横智慧城市股份有限公司 | Non-motor vehicle identification method, device, equipment and storage medium based on big data |
| CN110188693B (en) * | 2019-05-30 | 2023-04-07 | 重庆大学 | Improved complex environment vehicle feature extraction and parking discrimination method |
| CN113095289A (en) * | 2020-10-28 | 2021-07-09 | 重庆电政信息科技有限公司 | Massive image preprocessing network method based on urban complex scene |
| CN113326786B (en) * | 2021-06-01 | 2024-07-05 | 百度在线网络技术（北京）有限公司 | Data processing method, device, equipment, vehicle and storage medium |
| CN115797897A (en) * | 2023-02-03 | 2023-03-14 | 广州斯沃德科技有限公司 | Vehicle collision recognition method and system based on image processing |
| CN117237396B (en) * | 2023-11-16 | 2024-02-06 | 山东华盛中天工程机械有限责任公司 | Rail bolt rust area segmentation method based on image characteristics |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102087790A (en) * | 2011-03-07 | 2011-06-08 | 中国科学技术大学 | Method and system for low-altitude ground vehicle detection and motion analysis |
| CN102194122A (en) * | 2010-03-05 | 2011-09-21 | 索尼公司 | Method and equipment for classifying images |
| CN202563526U (en) * | 2012-03-22 | 2012-11-28 | 北京尚易德科技有限公司 | Transportation vehicle detection and recognition system based on video |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7853072B2 (en) * | 2006-07-20 | 2010-12-14 | Sarnoff Corporation | System and method for detecting still objects in images |
-
2012
- 2012-03-22 CN CN201210078707.5A patent/CN102637257B/en not_active Expired - Fee Related
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102194122A (en) * | 2010-03-05 | 2011-09-21 | 索尼公司 | Method and equipment for classifying images |
| CN102087790A (en) * | 2011-03-07 | 2011-06-08 | 中国科学技术大学 | Method and system for low-altitude ground vehicle detection and motion analysis |
| CN202563526U (en) * | 2012-03-22 | 2012-11-28 | 北京尚易德科技有限公司 | Transportation vehicle detection and recognition system based on video |
Non-Patent Citations (4)
| Title |
|---|
| Attribute-based Vehicle Search in Crowded Surveillance Videos;Rogerio Feris等;《ACM international Conference on Multimedia Retrieval(ICMR2011)》;20111231;全文 * |
| Rogerio Feris等.Attribute-based Vehicle Search in Crowded Surveillance Videos.《ACM international Conference on Multimedia Retrieval(ICMR2011)》.2011,全文. |
| 一种基于类Haar特征和改进Adaboost分类器的车辆识别算法;文学志等;《电子学报》;20110531(第5期);第1121-1126页 * |
| 文学志等.一种基于类Haar特征和改进Adaboost分类器的车辆识别算法.《电子学报》.2011,(第5期),第1121-1126页. |
Also Published As
| Publication number | Publication date |
|---|---|
| CN102637257A (en) | 2012-08-15 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN102637257B (en) | Video-based detection and recognition system and method of vehicles | |
| CN202563526U (en) | Transportation vehicle detection and recognition system based on video | |
| Alefs et al. | Road sign detection from edge orientation histograms | |
| Lin et al. | A license plate recognition system for severe tilt angles using mask R-CNN | |
| Romdhane et al. | An improved traffic signs recognition and tracking method for driver assistance system | |
| CN102289686A (en) | Method for identifying classes of moving targets based on transfer learning | |
| CN103279756A (en) | Vehicle detecting analysis system and detecting analysis method thereof based on integrated classifier | |
| CN103268489A (en) | Motor vehicle plate identification method based on sliding window searching | |
| CN102682304A (en) | Multi-feature integrated passer-by detection method and device | |
| CN102902983B (en) | A kind of taxi identification method based on support vector machine | |
| Hechri et al. | Robust road lanes and traffic signs recognition for driver assistance system | |
| CN104615986A (en) | Method for utilizing multiple detectors to conduct pedestrian detection on video images of scene change | |
| CN105260749A (en) | Real-time target detection method based on oriented gradient two-value mode and soft cascade SVM | |
| Kim et al. | Effective traffic lights recognition method for real time driving assistance systemin the daytime | |
| CN103310006A (en) | ROI extraction method in auxiliary vehicle driving system | |
| CN104463232A (en) | Density crowd counting method based on HOG characteristic and color histogram characteristic | |
| Ketcham et al. | Recognizing the Illegal Parking Patterns of Cars on the Road in Front of the Bus Stop Using the Support Vector Machine | |
| Chen et al. | Robust and real-time traffic light recognition based on hierarchical vision architecture | |
| CN108520528B (en) | Mobile vehicle tracking method based on improved difference threshold and displacement matching model | |
| Pavaskar et al. | Real-time vehicle-type categorization and character extraction from the license plates | |
| KR100942409B1 (en) | Method for detecting a moving vehicle at a high speed | |
| Kang | Dynamic programming-based method for extraction of license plate numbers of speeding vehicles on the highway | |
| Emami et al. | Real time vehicle make and model recognition based on hierarchical classification | |
| Zhou et al. | Real-time traffic light recognition based on c-hog features | |
| Promlainak et al. | Thai traffic sign detection and recognition for driver assistance |
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: 20140702 Termination date: 20200322 |
|
| CF01 | Termination of patent right due to non-payment of annual fee |