CN101630361A - Plate number, body color and mark identification-based equipment and plate number, body color and mark identification-based method for identifying fake plate vehicles - Google Patents

Abstract

Translated fromChinese

本发明公开了一种基于车身颜色、车牌和车标识别的套牌车辆自动识别设备及其方法。所述的设备包括：视频检测器、车牌定位器、车牌识别器、车身颜色识别器、车标定位器、车标识别器、数据库查询与报警装置等。根据车牌图像的局部边缘信息丰富的特点，该设备可以在捕获的图像中准确的定位车牌，利用车牌的位置，又提取车身区域和车标的粗略位置，根据提取的粗略位置再准确的提取车身颜色和车标；接下来是车牌字符识别、车身颜色判断和车标识别，根据此三项识别结果就和数据库里的资料进行比对，可以查出是否为套牌车辆，如果是套牌车辆，可以自动报警，由执法人员拦截进行进一步核实。该设备结构化好，操作简单，判别精度高，而且人工设定参数少。该设备还可以应用在欠费车辆、被盗车辆或违章车辆抓逃等。The invention discloses an automatic identification device and a method for identifying a set plate vehicle based on the color of a vehicle body, a license plate and a vehicle logo. The equipment includes: a video detector, a license plate locator, a license plate recognizer, a vehicle body color recognizer, a vehicle logo locator, a vehicle logo recognizer, a database query and alarm device, and the like. According to the rich local edge information of the license plate image, the device can accurately locate the license plate in the captured image, use the position of the license plate, extract the rough position of the body area and the car logo, and then accurately extract the color of the car body according to the extracted rough position and the car logo; the next step is license plate character recognition, body color judgment and car logo recognition. According to the three recognition results, it is compared with the data in the database to find out whether it is a fake car. If it is a fake car, It can automatically call the police and be intercepted by law enforcement officers for further verification. The device has the advantages of good structure, simple operation, high discrimination accuracy, and few manual setting parameters. The device can also be used in the capture and escape of arrears vehicles, stolen vehicles or illegal vehicles.

Description

A kind ofly identify other fake-licensed car identification equipment and method based on car plate, body color and car

Technical field

The present invention relates to the system of video image acquisition, Flame Image Process, image recognition, belong to computer vision and mode identification technology and intelligent transportation application, be specially and realize that vehicle management intellectuality, vehicles peccancy are handled real time implementation automatically and the traffic video automated processing system that designs.

Background technology

Fake-licensed car is exactly an automobile license plate of illegally applying mechanically other people, car plate is the I.D. of vehicle, in recent years, the fake-licensed car phenomenon spreads the whole nation, so that accident takes place frequently, crime increases severely, make a very bad impression, the people are boiled with resentment, upset normal civil order, how to hit these type of illegal activities effectively, become a difficult problem of vehicle supervision department.

At present, discern based on car plate " electronic police " every field and different regions popularization and applications such as technology is paid dues in urban highway traffic, expressway, public security bayonet socket, parking lot, the most frequently used way of tracing fake-licensed car is to go out license plate number by Computer Automatic Recognition, and whether Query Database is cover board suspected vehicles then.But car plate is changed when waiting factor affecting greatly by partial occlusion, stain, distortion and ambient lighting, accuracy of identification reduces greatly, error is big, brought inconvenience for traffic police and car owner, therefore only relying on the identification vehicle license to trace fake-licensed car is not enough, the judgement of testing vehicle register should need more feature, as: body color, car mark, vehicle or the like, for accurately judging testing vehicle register, effectively hit cover board phenomenon, the present invention proposes a fake-licensed car recognition methods that does not combine based on car plate identification, body color and car sign.

Summary of the invention

The objective of the invention is to adopt the automatic collection and the Intelligent treatment of traffic video completely, realization is to the location and the identification of vehicle license, body color, vehicle mark, the information of vehicles of collection and the information of vehicles of vehicle data storehouse record are compared, suspicious fake-licensed car is carried out automatic alarm, prompting traffic-police further evidence obtaining examines, thereby effectively solve and hit the cover board phenomenon of spreading unchecked at present, finish the intellectuality of traffic administration, the real time implementation of processing.

Equipment of the present invention realizes that by the following technical solutions described equipment comprises: the video camera that the video camera of video image acquisition and pattern process computer, video acquisition are used links to each other with pattern process computer by image pick-up card.The video image acquisition video camera can be a ccd video camera, also can be Digital Video, and video camera can use A-frame to be erected at by the road, also can be erected in the vehicle by fixed support.For obtaining license plate image clearly, video camera is monitored two tracks at most.Pattern process computer can be portable Industrial Control Computer, also can be common notebook computer.Pattern process computer has the function of radio communication, can realize being connected with the network of traffic control center by wireless communication module or wireless network card.

A fake-licensed car of the present invention recognition system can be used in warning with automobile, video camera can be placed in the shield glass back by fixed support, the power supply of video camera can provide by vehicle-mounted 12V power supply, the output of video camera is input to computer equipment by video frequency collection card, computer equipment comprises notebook computer or portable industrial pc, computer equipment among the present invention is equipped with sound card, the audible alarm equipment that audio amplifier is formed, car plate steady arm positioning car memorial tablet is put, the car plate recognizer identifies characters on license plate, the body color recognizer is used to discern body color, car mark steady arm is according to the accurate positioning car mark in the position of car plate, the car sign does not identify the car mark according to positioning result, the data base querying device can be according to the car plate result who identifies, body color and car mark result and database are compared, if the vehicle that computer Recognition goes out and the information of vehicles of actual enrollment is not inconsistent or " blacklist " on vehicle, computing machine can be reported to the police by sound and software interface, and the prompting law enfrocement official further examines.Database can be the local data base that has downloaded to computing machine, also can be to connect by wireless network card and command centre by computing machine, supports the long-range data transmission, and the real time access central database is inquired about.

This illustrates that a described fake-licensed car recognition system also can be in the other use of road, video camera can be erected at by the road by A-frame, it is approaching with the car plate height that video camera sets up height, and taking direction can be consistent with vehicle heading, also can be towards vehicle heading.Shooting angle is defined as vehicle heading and the acute angle portion of taking direction, generally between 20 ~ 45 degree, can monitor equidirectional two tracks.

A fake-licensed car of the present invention recognition system adopts the high-definition picture acquisition technique, clear picture, have continuous candid photograph ability, image processing speed is fast, can guarantee can both capture under the different speed of a motor vehicle vehicle and identify information such as car plate, body color and Che Biao.The information that computing machine is preserved comprises vehicle monitoring place, vehicle heading, vehicle time of occurrence, car plate, body color, Che Biao etc., is convenient to the analysis retrieval of lot of data file.

The present invention's video detector is mainly finished the vehicle detection in the video image, and whether vehicle is arranged in the detected image, if discovery has vehicle to occur then starts memory image, and starts recognizer, location and identification car plate, vehicle color and car mark.The work distance of video detector is as follows:

(4) video image obtains: the video image of shot by camera is caught by image pick-up card.

(5) background study: background image is used for the detection of vehicle target, obtains background image by several two field pictures.If I_tBe t image constantly, B_tIt is the background image of t$B_t=\frac{1}{t}\underset{i=1}{\overset{t}{\mathrm{\Σ}}}{I}_i,$Then next background constantly has following formula to produce B_T+1=α I_T+1+ (1-α) B_t, α is a weighting factor, strengthens the influence of current frame image to background.

(6) target detection: target detection mainly is to detect vehicle target, subtract each other by present frame and background image, if gray scale difference greater than preset threshold, is then thought background, for reducing The noise, because car plate part edge abundant information is carried out smoothly with a rectangle frame, if the part greater than threshold value is arranged in the image, then think and include vehicle in this two field picture, trigger, otherwise think do not have vehicle in the image, do not handle.

The present invention's car plate steady arm and recognizer, body color recognizer and car mark steady arm, recognizer have constituted a fake-licensed car core that automatic identification is handled, and the workflow of car plate steady arm is as follows:

(5) difference image that obtains: after video detector detects the image that contains vehicle, trigger the car plate steady arm at once, the car plate steady arm is according to the image detection car plate of catching, and the car plate steady arm carries out difference with image with the Sobel operator, the image high fdrequency component that obtains earlier.

(6) seek the candidate region: the width of general license plate area accounts for 1/5 to 1/7 of picture traverse, with the rectangle frame similar to car plate, its length is respectively from 1/5 to 1/7 of image length, and in entire image, slide, add up its high fdrequency component and, and, count the maximum value of horizontal direction to the horizontal direction projection; Add up the high fdrequency component sum on the maximum position again, to the vertical direction projection, get the maximum value of several vertical direction, the maximum position that level that obtains and vertical direction obtain has constituted the position candidate of car plate.

(7) judgement of license plate area: the candidate region has a plurality of, and car plate has only one or two in the image, the judgement of license plate area is to carry out according to the characteristics of the standard car plate of existing Ministry of Public Security promulgation, present car plate mainly is divided into the civilian car plate of 92 formulas (wrongly written or mispronounced character of the blue end), double cart car plate (yellow end surplus), police car, military vehicle, People's Armed Police's vehicle (white gravoply, with black engraved characters) and part 02 formula car plate.Judge whether the candidate region meets the character queueing discipline to register the license, if, then confidence level is improved, if not, then confidence level reduces, and coming top after the ordering is to be the zone of car plate most possibly.

(8) character cutting: carry out character cutting according to the car plate type of judging, the cutting cost in the process of cost function record cutting is used to weigh the size into the possibility of car plate.

The present invention's car plate recognizer has been finished the identification of characters on license plate and the affirmation of car plate final position, and its workflow is as follows:

(5) character is regular: the single character of car plate that cutting obtains not of uniform size, for ease of identification, to image adopt carry out based on the method for character center of gravity regular, unified regular be 40 * 20 sizes.

(6) feature extraction: only get edge of image, again image be divided into the sub-piece of 8 * 4 sizes, extract the direction character of each sub-piece, with the level of 3 * 3 each sub-piece of template extraction, vertical ,-45 °, the feature of+45 ° of four directions.

(7) character recognition: the feature and the training characteristics that are drawn into are mated, according to the principle output result of arest neighbors.The usefulness because car plate diverse location character rule differs, the scope during coupling is also different, generally speaking, first all is writing a Chinese character in simplified form and military Chinese character of each province, and second is English character, and third and fourth both might be English alphabet, also might be numeral, back three be had only numeral.

(8) confidence level is estimated: confidence level is that the matching degree of character is calculated during according to the cost function of character cutting and character recognition, according to the size of confidence level, and final decision result output, and provide the position of final car plate.

The present invention's body color recognizer is mainly discerned the automobile body color, its workflow is as follows: (4) body color location: the position of car plate is relatively accurate, according to the position of car plate, seeking large stretch of difference value above it is the zone of body color near zero zone.

(5) color conversion: what the spatial model of the coloured image that video camera obtains adopted is the RGB color model, and each pixel has R, G, and the combination of three channel value of B is represented jointly.The HIS space proposes based on the human visual system, has higher precision when being used for identification colors, and calculates simple.So body color RGB is transformed to the HSI space by computing formula.

(6) color identification: in the HSI space, the variation of statistics tone utilizes information such as brightness and saturation degree simultaneously, and vehicle color is classified, and can be divided into: black, white, red, yellow, green, blue etc.

The present invention's car is demarcated the position and has been finished car target Intelligent treatment with recognizer, and its treatment scheme is as follows:

(4) car mark coarse positioning: car mark coarse positioning at first according to the accurate position of car plate, is carried out in certain zone in car target location directly over car plate, seek the general position of car mark.

(5) the car mark is carefully located: may contain unnecessary noises such as the horizontal grid of headstock in the zone of coarse positioning, utilize image enhancement technique, Mathematical Morphology technology that noise is removed, accurately the positioning car mark.

(6) the car sign is not: the car sign does not adopt based on the template matches technology, with binaryzation and regular after car mark on a map the picture mate with known car target image library, the similarity between computed image, the most similar person exports as car mark result, if similarity is less than preset threshold, and refuse to know.

The present invention's data base querying and warning device are the front-end processing parts of a fake-licensed car identification, and its workflow is as follows:

(5) data base querying: by the recognition result of car plate recognizer, body color recognizer and the output of car mark recognizer, with the license plate recognition result is the master index inquiry, inquiry local data base or connect the database of command centre by wireless network is checked in conjunction with body color and car mark again.

(6) automatically report to the police: be registered fake-licensed car if inquire this with the car plate result in database, equipment is directly reported to the police; If not registered fake-licensed car, but the information of vehicles that identifies information of vehicles and record is not inconsistent, and equipment is with alarm.

(7) artificial check and correction: recognition result always has certain error, if recognition result is with a low credibility, but the vehicle of inquiry has cover board suspicion, and device prompts has the artificial nucleus right, and can make amendment to the result.

(8) image is preserved: for preserving the punishment evidence, equipment can adopt the form of .jpg to preserve to the image of capturing.

Below in conjunction with the detailed description of accompanying drawing, the flesh and blood of technical scheme that the present invention may be better understood, identical mark is represented identical device in the accompanying drawing.

Description of drawings

Fig. 1 is based on car plate, a car mark and body color fake-licensed car identification equipment block diagram;

Fig. 2 is the synoptic diagram of video detector;

Fig. 3 is a fake-licensed car identification processing procedure synoptic diagram;

Fig. 4 is data base querying and warning processing flow chart.

Embodiment

An a kind of fake-licensed car identification equipment and method thereof based on car plate, car mark and body color information that the present invention proposes realizes by the following technical solutions.

Fig. 1 is based on a fake-licensed car identification equipment block diagram of car plate, car mark and body color information.As shown in Figure 1, described equipment comprises:

Video detector is used for the obtaining of video image, background study and target detection, occurs if video detector has detected vehicle target, then image is sent into the car plate steady arm, and starts car plate identification, body color identification and car sign not.The car plate steady arm is mainly oriented a plurality of candidate regions of car plate, and the car plate recognizer carries out character cutting to each candidate region, identifies characters on license plate, each zone is provided a confidence level judge the output as a result of that confidence level is the highest.The body color recognizer is according to the accurate location positioning body color of car plate, and the identification body color, and the car plate position that car mark steady arm also provides according to the car plate steady arm is the positioning car mark accurately, and car mark recognizer is discerned the car mark.The data base querying device can be compared according to the car plate result who identifies, body color and car mark result and database, at first inquire about a local fake-licensed car database with the number-plate number of identification, an if fake-licensed car then direct sound and light alarm, if not in a fake-licensed car database, then inquire about command centre vehicle data storehouse, the information of vehicles of the car plate that identifies, car mark and body color and actual enrollment is not inconsistent, computing machine can be reported to the police by sound and software interface, and the prompting law enfrocement official further examines.Database can be the local data base that has downloaded to computing machine, also can be to connect by wireless network card and command centre by computing machine, supports the long-range data transmission, and the real time access central database is inquired about.

The process flow diagram of Fig. 2 video detector.Its concrete steps are as follows:

Video detector obtains video image from input media, and input media can also can be a Digital Video by ccd video camera.Video camera can be erected in the car by fixed support, also can be erected at by the road by A-frame, and the power supply of video camera is provided by vehicle power.Background study is used for the detection of target, and the order of accuarcy of background has influence on the accuracy of target detection, so background study adopts the method for weighting to upgrade.The detection of target can obtain by the mode of frame difference, if region of variation has surpassed preset threshold, then thinking has target to occur, and is for further processing, if there is not target to occur, continues target detection, and upgrades background.

Fig. 3 is an information of vehicles identification processing flow chart.Its step is as follows:

Car plate steady arm and recognizer, body color recognizer and car mark steady arm, recognizer has constituted the core that automatic vehicle identification is handled, the car plate steady arm is the basis of entire process, after video detector detects vehicle, trigger the car plate steady arm at once, the car plate steady arm is according to the informative characteristics of car plate part edge, seek the candidate region, the candidate region of being found has a plurality of, and car plate has only one or two in the image, the judgement of ensuing license plate candidate area is then according to existing car plate standard, judge the whether character queueing discipline of conformance with standard car plate of candidate region, give different confidence levels according to matching degree.Character cutting carries out cutting according to the car plate type of judging, the cutting cost in the process of cost function record cutting also is the size that is used to weigh into the possibility of car plate.The car plate recognizer has been finished the affirmation to the identification of characters on license plate and car plate final position, feature extraction has partly been finished the character feature after the marginalisation has been extracted, the feature and the training characteristics that are drawn into are mated in character recognition, according to the principle output result of arest neighbors.Confidence level is that the matching degree of character is calculated during according to the cost function of character cutting and character recognition, according to the size of confidence level, and final decision result output, and provide the position of final car plate.The position of car plate is relatively accurate, and according to the position of car plate, seeking large stretch of difference value above it is the zone of body color near zero zone.By computing formula body color RGB is transformed to the HSI space.In the HSI space, the variation of statistics tone utilizes information such as brightness and saturation degree simultaneously, and vehicle color is classified, and can be divided into: black, white, red, yellow, green, blue etc.Car is demarcated the position and has been finished car target Intelligent treatment with recognizer, and car target location also is the accurate position according to car plate, carries out car mark coarse positioning in certain zone directly over car plate, seeks the general position of car mark.In the zone of coarse positioning, may contain unnecessary noises such as the horizontal grid of headstock, utilize image enhancement technique, Mathematical Morphology technology that noise is removed, accurately the positioning car mark.Car sign does not adopt based on the template matches technology, with binaryzation and regular after the car picture of marking on a map mate with known car target image library, the similarity between computed image, the most similar person exports as car mark result.

Fig. 4 is data query and warning device.Its workflow is as follows:

Recognition result by car plate recognizer, body color recognizer and the output of car mark recognizer, with the license plate recognition result is the master index inquiry, inquiry local data base or connect the database of command centre by wireless network is checked in conjunction with body color and car mark again.Be registered fake-licensed car if inquire this with the car plate result in database, equipment is directly reported to the police; If not registered fake-licensed car, but the information of vehicles that identifies information of vehicles and record is not inconsistent, and equipment is with alarm.Recognition result always has certain error, if recognition result is with a low credibility, but the vehicle of inquiry has cover board suspicion, and device prompts has the artificial nucleus right, and can make amendment to the result.For preserving the punishment evidence, equipment can adopt the form of .jpg to preserve to the image of capturing.

More than describe and only provide implementation method of the present invention by means of embodiment.For those skilled in the art is conspicuous, the implementation detail that the invention is not restricted to provide above, can realize with additional embodiments under the situation that does not break away from feature of the present invention that some parts among the embodiment decompose, merge or use processor to realize.Therefore, it is illustrative that the embodiment that provides should be considered to, rather than restrictive.Therefore, it is defined by the appended claims realizing and use possibility of the present invention.Thereby the realization various selections of being determined by claim of the present invention comprise that equivalent embodiment also belongs to scope of the present invention.

Claims (9)

Translated fromChinese

1、一种基于车牌、车身颜色和车标识别的套牌车辆识别设备及方法，其特征在于：所述的设备包括：视频检测器、车牌定位器、车牌识别器、车身颜色识别器、车标定位器、车标识别器、数据库查询与报警装置等，其中视频检测器可以捕获视频，并检测车辆，将含有车辆的图像传送给车牌定位器，车牌识别器给出识别结果。以车牌位置为相对坐标，车身颜色识别器和车标识别器识别车身颜色和车标，首先以识别的车牌号码查询本地套牌车辆数据库，如果是套牌车辆报警装置则直接声光报警，如果不在套牌车辆数据库中，则查询指挥中心车辆数据库，识别出的车牌、车标和车身颜色和实际登记的车辆信息不符，报警装置可以通过声音和软件界面报警。1. A license plate vehicle recognition device and method based on license plate, vehicle body color and vehicle logo identification, characterized in that: the device includes: a video detector, a license plate locator, a license plate recognizer, a vehicle body color recognizer, a vehicle Mark locator, car logo recognizer, database query and alarm device, etc., where the video detector can capture video, detect vehicles, and send the image containing the vehicle to the license plate locator, and the license plate recognizer will give the recognition result. With the position of the license plate as the relative coordinate, the body color recognizer and the car logo recognizer recognize the body color and the car logo, first query the local license plate vehicle database with the recognized license plate number, if it is a license plate vehicle alarm device, it will directly sound and light alarm, if If it is not in the licensed vehicle database, query the vehicle database of the command center. The identified license plate, vehicle logo and body color do not match the actual registered vehicle information. The alarm device can alarm through sound and software interface.2、根据权利要求1所述的套牌车辆识别设备，其特征在于设备使用灵活，可以架设在警车内，也可以架设在道路旁，操作简单，安装方便。2. The license plate vehicle identification device according to claim 1, characterized in that the device is flexible in use and can be installed in a police car or beside a road, and is easy to operate and install.3、根据权利要求1所述的套牌车辆识别设备及方法，其特征在于可以识别车辆牌照、车身颜色以及车标等车辆信息，并记录时间、地点等相关信息。3. The device and method for identifying vehicles with license plates according to claim 1, characterized in that it can identify vehicle information such as vehicle license plate, body color, and vehicle logo, and record relevant information such as time and location.4、根据权利要求3所述的识别套牌车辆识别设备，其特征在于视频检测器的工作流程如下：4. The vehicle recognition device for identifying decks according to claim 3, wherein the workflow of the video detector is as follows:(1)视频图像获取：将摄像机所拍摄的视频图像通过图像采集卡进行捕获。(1) Video image acquisition: capture the video image captured by the camera through the image acquisition card.(2)背景学习：通过加权的方式学习背景，加大当前帧图像对背景的影响。(2) Background learning: learn the background by weighting, and increase the influence of the current frame image on the background.(3)目标检测：通过当前帧与背景图像相减，如果灰度差大于设定的阈值，则认为是背景，为减少噪声的影响，因为车牌部分边缘信息丰富，用一个矩形框进行平滑，如果图像中有大于阈值的部分，则认为此帧图像中包含有车辆，进行触发，否则认为图像中没有车辆，不进行处理。(3) Target detection: Subtract the current frame from the background image. If the gray level difference is greater than the set threshold, it is considered to be the background. In order to reduce the impact of noise, because the edge information of the license plate part is rich, a rectangular frame is used for smoothing. If there is a part larger than the threshold in the image, it is considered that the frame image contains a vehicle, and the trigger is performed; otherwise, it is considered that there is no vehicle in the image, and no processing is performed.5、根据权利要求3所述的识别套牌车辆识别设备，其特征在于车牌定位器的工作流程如下：5. The vehicle identification device for identifying a set of license plates according to claim 3, wherein the working process of the license plate locator is as follows:(1)获的差分图像：将图像用Sobel算子进行差分，获的图像高频分量。(1) Obtained differential image: the image is differentiated by the Sobel operator to obtain the high-frequency component of the image.(2)寻找候选区域：统计其高频分量的和，并向水平和垂直方向投影，获得多个车牌的候选位置。(2) Find candidate areas: count the sum of its high-frequency components, and project them in the horizontal and vertical directions to obtain multiple candidate positions of license plates.(3)车牌区域的判定：判断候选区域是否符合现有车牌的字符排列规则，如果是，则将可信度提高，如果否，则可信度降低，排序后排在最前面的为最有可能是车牌的区域。(3) Judgment of the license plate area: judge whether the candidate area conforms to the character arrangement rules of the existing license plate, if yes, the credibility will be increased, if not, the credibility will be reduced, and the one ranked at the top after sorting is the most reliable Probably the area of the license plate.(4)字符切分：根据判断的车牌类型进行字符切分，代价函数记录切分的过程中的切分代价，用于衡量为车牌的可能性的大小。(4) Character segmentation: perform character segmentation according to the type of license plate judged, and the cost function records the segmentation cost in the process of segmentation, which is used to measure the possibility of being a license plate.6、根据权利要求3所述的识别套牌车辆识别设备，其特征在于车牌识别器的工作流程如下：6. The vehicle recognition device for identifying a set of license plates according to claim 3, wherein the working process of the license plate recognizer is as follows:(1)字符规整：对图像采用基于字符重心的方法进行规整，统一规整为40×20大小。(1) Character regularization: The image is regularized based on the center of gravity of the character, and the size is uniformly regularized to 40×20.(2)特征抽取：抽取字符的方向线素特征。(2) Feature extraction: extract the direction line feature of the character.(3)字符识别：将抽取到的特征与训练特征进行匹配，根据最近邻的原则输出结果。(3) Character recognition: match the extracted features with the training features, and output the results according to the principle of nearest neighbor.(4)可信度估计：可信度是根据字符切分的代价函数和字符识别时字符的匹配程度来计算的，根据可信度的大小，最终决定结果输出，并给出最终的车牌的位置。(4) Credibility estimation: the credibility is calculated according to the cost function of character segmentation and the matching degree of characters during character recognition. According to the size of the credibility, the final result output is determined, and the final license plate is given. Location.7、根据权利要求3所述的识别套牌车辆识别设备，其特征在于车身颜色识别器的工作流程如下：7. The vehicle identification device for identifying a license plate according to claim 3, wherein the working process of the vehicle body color identifier is as follows:(1)车身颜色定位：车牌的位置相对比较精确，根据车牌的位置，在其上方寻找大片差分值接近零的区域为车身颜色的区域。(1) Body color positioning: The position of the license plate is relatively accurate. According to the position of the license plate, a large area with a difference value close to zero is found above it as the area of the body color.(2)颜色转换：把车身颜色RGB变换到HSI空间。(2) Color conversion: Transform the body color RGB to HSI space.(3)颜色识别：在HSI空间，统计色调的变化，同时利用亮度和饱和度等信息，将车辆颜色进行分类。(3) Color recognition: In the HSI space, the change of the hue is counted, and the vehicle color is classified by using information such as brightness and saturation.8、根据权利要求3所述的识别套牌车辆识别设备，其特征在于车标定位与识别器的工作流程如下：8. The vehicle identification device for identifying a license plate according to claim 3, wherein the working process of the vehicle logo positioning and the identifier is as follows:(1)车标粗定位：车标的定位首先根据车牌的准确位置，在车牌的正上方一定区域内进行车标粗定位，寻找车标大概的位置。(1) Coarse positioning of the car logo: The positioning of the car logo is firstly based on the accurate position of the license plate, and the rough positioning of the car logo is carried out in a certain area directly above the license plate to find the approximate position of the car logo.(2)车标细定位：利用图像增强技术、数学形态学技术对噪声进行去除，准确定位车标。(2) Fine positioning of car logos: use image enhancement technology and mathematical morphology technology to remove noise and accurately locate car logos.(3)车标识别：车标识别采用基于模板匹配技术，计算图像间的相似度，最相似者作为车标结果输出，若相似度小于设定的阈值，而拒识。(3) Car logo recognition: The car logo recognition uses template matching technology to calculate the similarity between images, and the most similar one is output as the car logo result. If the similarity is less than the set threshold, it will be rejected.9、根据权利要求3所述的识别套牌车辆识别设备，其特征在于车牌识别器的工作流程如下：9. The vehicle recognition device for identifying sets of license plates according to claim 3, wherein the working process of the license plate recognizer is as follows:(1)数据库查询：由车牌识别器、车身颜色识别器和车标识别器输出的识别结果，以车牌识别结果为主索引查询，查询本地数据库或通过无线网络连接指挥中心的数据库，再结合车身颜色和车标进行核对。(1) Database query: the recognition results output by the license plate recognizer, body color recognizer and car logo recognizer, the license plate recognition result is the main index query, query the local database or the database connected to the command center through the wireless network, and then combine the vehicle body Check the color and logo.(2)自动报警：如果以车牌结果在数据库中查询到该为已登记的套牌车辆，设备直接报警；如果不是已登记的套牌车辆，但识别出车辆信息与记录的车辆信息不符，设备将报警提示。(2) Automatic alarm: If the license plate result is found in the database as a registered licensed vehicle, the device will directly alarm; if it is not a registered licensed vehicle, but the identified vehicle information does not match the recorded vehicle information, the device will Alarm will be prompted.(3)人工校对：识别结果总有一定误差，如果识别结果可信度低，但查询的车辆有套牌嫌疑，设备提示有人工核对，并可对结果进行修改。(3) Manual proofreading: There are always certain errors in the recognition results. If the reliability of the recognition results is low, but the queried vehicle is suspected of being fake, the device prompts that there is a manual verification, and the results can be modified.(4)图像保存：为保存处罚证据，设备可对抓拍的图像采用.jpg的格式进行保存。(4) Image storage: In order to preserve the punishment evidence, the device can save the captured image in .jpg format.

Images