A vehicle identification method, system and identification device

Abstract

The application relates to a vehicle identification method, a system and an identification device, and belongs to the technical field of license plate identification.The license plate identification method comprises the following steps: acquiring a vehicle image of a target vehicle in a target area; identifying the license plate number of the target vehicle according to the vehicle image; in response to the existence of a target vehicle whose license plate number is not identified, acquiring a driving parameter of the target vehicle whose license plate number is not identified, and continuously detecting the target vehicle according to the driving parameter; and in response to the plurality of target vehicles being a continuous vehicle team, identifying the license plate number on the side of the body of the target vehicle based on the vehicle image. The driving parameter of the target vehicle whose license plate number is not identified is acquired, if the target vehicles are a continuous vehicle team, it is indicated that the target vehicle whose license plate number is not identified is possibly caused by the shielding of adjacent target vehicles, and the license plate number on the side of the body of the target vehicle in the vehicle image is identified, so that the vehicle in the vehicle team can be identified conveniently.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of license plate recognition, in particular to a vehicle recognition method, system and recognition device. BACKGROUND

[0002] Non-stop over-limit detection is a traffic management technology used to detect whether a vehicle in motion exceeds the prescribed height limit, width limit or weight limit without stopping the vehicle. This detection technology is usually applied to highways, bridges, tunnels and other places where vehicle size or weight needs to be limited. Non-stop over-limit detection mainly includes a visual detection system, an over-limit detection system and a database management system. The visual detection system captures information such as truck license plate numbers and outlines, the detection system detects the height limit, weight limit and other information of the truck, and the database management system records the violation information of the over-limit truck.

[0003] However, when the visual detection system identifies the license plate number of the truck, if the distance between adjacent trucks is too close, or if the over-limit truck passes through the over-limit detection system in the form of a closely arranged vehicle train, the license plate numbers of the trucks located in the middle of the vehicle train are blocked by the adjacent trucks, so the license plate numbers of the trucks located in the middle of the vehicle train cannot be obtained, and the over-limit detection of the trucks cannot be effectively performed, which is not conducive to the traffic management of the trucks. SUMMARY

[0004] In order to facilitate the identification of vehicles in a vehicle train, the present application provides a vehicle recognition method, system and recognition device.

[0005] In a first aspect, the present application provides a vehicle recognition method, which adopts the following technical solution:

[0006] A vehicle recognition method, comprising:

[0007] obtaining a vehicle image of a target vehicle in a target area;

[0008] identifying the license plate number of the target vehicle according to the vehicle image;

[0009] in response to the existence of a target vehicle whose license plate number is not identified, obtaining a driving parameter of the target vehicle whose license plate number is not identified, and performing continuous vehicle train detection on the target vehicle according to the driving parameter;

[0010] in response to the multiple target vehicles being a continuous vehicle train, identifying the license plate number on the side of the body of the target vehicle based on the vehicle image.

[0011] By adopting the technical scheme, the license plate number of the target vehicle is preliminarily identified based on the vehicle image, and the driving parameter of the target vehicle whose license plate number is not identified is acquired to detect whether the target vehicle whose license plate number is not identified is a continuous vehicle team, and if so, it is indicated that the target vehicle whose license plate number is not identified is caused by the shielding of the adjacent target vehicle. The license plate number of the target vehicle at the middle position of the continuous vehicle team is identified by identifying the license plate number of the license plate number of the side of the vehicle body in the vehicle image, so that the license plate number of the target vehicle at the middle position of the continuous vehicle team is acquired, and the effect of identifying the vehicle in the vehicle team is achieved.

[0012] Optionally, the vehicle image includes a first shooting image shot by a first camera and a second shooting image shot by a second camera; the first camera is used to shoot one of the front or the tail of the target vehicle; and the second camera is used to shoot the other of the front or the tail of the target vehicle.

[0013] The identification of the license plate number of the target vehicle according to the vehicle image specifically includes:

[0014] The vehicle image is subjected to morphological processing, and the vehicle image is segmented to obtain a plurality of connected regions;

[0015] The shape features of each connected region are analyzed, and the connected regions are filtered according to the shape features to obtain vehicle component regions; and adjacent or overlapping vehicle component regions are clustered to obtain one or more candidate vehicle regions;

[0016] The edge of each candidate vehicle region is detected to obtain a license plate region, and the license plate region and the candidate vehicle region are combined; and in response to the existence of a candidate vehicle region that is not combined with the license plate region, it is determined that the corresponding candidate vehicle region is a target vehicle whose license plate number is not identified;

[0017] In response to the candidate vehicle region combined with the license plate region, the license plate region is identified to obtain the license plate number of the target vehicle corresponding to the candidate vehicle region.

[0018] By adopting the technical scheme, the vehicle image is segmented to obtain a plurality of connected regions, and the connected regions are filtered to obtain vehicle component regions. The vehicle component regions are clustered to obtain candidate vehicle regions, and the license plate region is identified by edge detection, so that the license plate region and the corresponding candidate vehicle region are combined. When the candidate vehicle region that is not combined with the license plate region appears, it is indicated that the license plate number corresponding to the candidate vehicle region cannot be acquired. The identified license plate region is identified, so that the identification of the license plate number is achieved.

[0019] Optionally, after the identification of the license plate number of the target vehicle according to the vehicle image, the method further includes:

[0020] In response to identifying the target vehicle of the license plate number, the license plate region image is acquired;

[0021] The license plate region image is input into a license plate verification model, the license plate feature information of the license plate region image in a plurality of dimensions is extracted through the license plate verification model, and a verification result of the license plate region image is output according to the license plate feature information in the plurality of dimensions.

[0022] By adopting the above technical solution, for the target vehicle of the identified license plate number, the license plate feature information in a plurality of dimensions is extracted by using the license plate verification model, so as to verify the authenticity of the license plate from the license plate feature information in the plurality of dimensions, and the effectiveness of the acquired license plate number is ensured.

[0023] Optionally, the method further includes a training step of training the preset convolutional neural network to obtain the license plate verification model, and the training step includes:

[0024] A training data set is acquired, and the training data set includes a plurality of sample images, real license plate images and fake license plate images;

[0025] The sample images are labeled with license plate feature information in a plurality of dimensions, and corresponding authenticity labels are labeled for the sample images according to the labeled license plate feature information in the plurality of dimensions;

[0026] The sample images are input into the preset convolutional neural network to output authenticity prediction results, and the model parameters of the preset convolutional neural network are iteratively updated according to the authenticity labels and the authenticity prediction results, to obtain the license plate verification model.

[0027] By adopting the above technical solution, the real license plate images and the fake license plate images in the training data set are labeled with license plate feature information in a plurality of dimensions, and corresponding authenticity labels are labeled for the sample images according to the labeled license plate feature information, the sample images are input into the preset convolutional neural network to obtain authenticity prediction results, and the model parameters of the convolutional neural network are iteratively updated according to the authenticity labels and the authenticity prediction results, so as to generate the license plate verification model having the function of verifying the authenticity of the license plate region.

[0028] Optionally, the method further includes a training step of training the preset convolutional neural network to obtain the license plate verification model, and the training step includes:

[0029] Each target vehicle of which the license plate number is not identified is numbered;

[0030] According to the number and the coordinates of the target vehicle, the driving distance between each target vehicle and each adjacent target vehicle in front and behind is calculated respectively;

[0031] acquire the average speed and the average direction of all the target vehicles whose license plate numbers are not recognized;

[0032] According to the driving distance, the average speed and the average direction, the driving parameters of the target vehicles whose license plate numbers are not recognized are obtained.

[0033] By adopting the technical scheme, each target vehicle whose license plate number is not recognized is numbered so as to be distinguished, and the driving distance between each target vehicle and each target vehicle in front and behind is calculated, and the driving parameters of the target vehicles whose license plate numbers are not recognized are obtained based on the driving distance and the acquired average speed and average direction.

[0034] Optionally, the acquiring the vehicle image of the target vehicle specifically comprises:

[0035] receiving the first shooting signal and the second shooting signal; the first shooting signal and the second shooting signal are output by the geomagnetic ring sensor when the target vehicle passes by;

[0036] controlling the first camera to shoot the first shooting image in response to the first shooting signal and controlling the second camera to shoot the second shooting image in response to the second shooting signal;

[0037] acquiring the vehicle image of the target vehicle according to the first shooting image and the second shooting image.

[0038] By adopting the technical scheme, the first shooting signal or the second shooting signal is output by the geomagnetic ring sensor when the target vehicle passes by, so as to control the first camera and the second camera to shoot the vehicle image when the target vehicle passes by the geomagnetic ring sensor.

[0039] Optionally, the system further comprises:

[0040] storing the vehicle image and sending the vehicle image to the management terminal in response to the target vehicle whose license plate number is not recognized not being a continuous vehicle team or in response to the license plate number of the side of the body of the target vehicle not being recognized.

[0041] By adopting the technical scheme, when the target vehicle whose license plate number is not recognized is not a continuous vehicle team or the target vehicle whose license plate number is not recognized is not a continuous vehicle team, it is regarded as an abnormal situation, and the vehicle image at this time is sent to the management terminal so as to be manually reviewed.

[0042] In a second aspect, the application provides a vehicle identification system, which adopts the following technical scheme:

[0043] A vehicle identification system comprises:

[0044] an image acquisition module, configured to acquire a vehicle image of a target vehicle;

[0045] The first license plate recognition module is configured to recognize the license plate number of the target vehicle according to the vehicle image.

[0046] The vehicle fleet detection module is configured to, in response to the existence of the target vehicle whose license plate number is not recognized, acquire the driving parameter of the target vehicle whose license plate number is not recognized, and perform continuous vehicle fleet detection on the target vehicle according to the driving parameter.

[0047] The second license plate recognition module is configured to, in response to the target vehicle being a continuous vehicle fleet, recognize the license plate number on the side of the vehicle body of the target vehicle based on the vehicle image.

[0048] In a third aspect, the present application provides a computer device, which adopts the following technical solution:

[0049] The computer device comprises a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor executes the computer program of any of the above methods.

[0050] In a fourth aspect, the present application provides a computer readable storage medium, which adopts the following technical solution:

[0051] The computer readable storage medium comprises a computer program capable of being loaded and executed by a processor, and the computer program is any of the above methods. BRIEF DESCRIPTION OF DRAWINGS

[0052] Figure 1 is a flowchart of a license plate recognition method of an embodiment of the present application.

[0053] Figure 2 is a top view of the installation position of the first camera and the second camera of an embodiment of the present application.

[0054] Figure 3 is a side view of the installation position of the first camera and the second camera of an embodiment of the present application.

[0055] Figure 4 is a flowchart of a method for recognizing a license plate number of an embodiment of the present application.

[0056] Figure 5 is a flowchart of a method for checking a license plate number of an embodiment of the present application.

[0057] Figure 6 is a flowchart of a method for training a checking model of an embodiment of the present application.

[0058] Figure 7 is a flowchart of a method for acquiring a driving parameter of an embodiment of the present application.

[0059] Figure 8is a block diagram of a license plate recognition system according to an embodiment of the present application.

[0060] Reference signs: 1, first camera; 2, gantry; 3, second camera; 4, first geomagnetic ring; 5, second geomagnetic ring. DETAILED DESCRIPTION

[0061] In order to make the purpose, technical solutions and advantages of the present application clearer, the present application will be further described in detail below with reference to the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and not to limit the present application.

[0062] The embodiments of the present application disclose a vehicle recognition method. Referring to Figure 1 , Figure 2 , Figure 3 A vehicle recognition method comprises the following steps:

[0063] Step S101: Obtain a vehicle image of a target vehicle in a target area.

[0064] The vehicle image comprises a first captured image captured by the first camera 1 and a second captured image captured by the second camera 3; the first camera 1 is used to capture one of the front or rear view of the target vehicle, and the second camera 3 is used to capture the other of the front or rear view of the target vehicle. Figure 2 , 3 In order to facilitate the installation of the first camera 1 and the second camera 3, the gantry 2 is provided, the two ends of the gantry 2 are horizontally arranged on the two sides of the highway, the first camera 1 is arranged on one side of the gantry 2, and the shooting direction of the first camera 1 is directed away from the gantry 2, the second camera 3 is arranged on the other side of the gantry 2, and the shooting direction of the second camera 3 is directed away from the gantry 2 and opposite to the shooting direction of the first camera 1. And the first camera 1 and the second camera 3 are arranged on two opposite corners at the top of the gantry 2, it should be understood that the first camera 1 can not only capture one of the front or rear view of the target vehicle, but also capture the body of the target vehicle close to the first camera 1, and the second camera 3 can also capture the other of the front or rear view of the target vehicle, and the body of the target vehicle close to the second camera 3, and because the first camera 1 and the second camera 3 are distributed at the top of the gantry 2, the bodies of the target vehicle captured by the first camera 1 and the second camera 3 are not on the same side, so that the first camera 1 and the second camera 3 realize the acquisition of the vehicle image of the front, rear and two sides of the target vehicle.

[0065] The target vehicle comprises a truck that needs to be detected for over-limit.

[0066] Step S102: Identify the license plate number of the target vehicle based on the vehicle image.

[0067] It should be understood that there are two possible results when identifying the license plate number of a target vehicle: one is that the license plate number of the target vehicle is not identified, and the other is that the license plate number of the target vehicle is identified.

[0068] Step S103: In response to the existence of a target vehicle with an unidentified license plate number, obtain the driving parameters of the target vehicle with the unidentified license plate number, and perform continuous platoon detection on the target vehicle based on the driving parameters.

[0069] Among them, a continuous convoy refers to a convoy in which the front and rear of multiple target vehicles are arranged closely in sequence.

[0070] Step S104: In response to multiple target vehicles forming a continuous convoy, identify the license plate number on the side of the target vehicle based on the vehicle image.

[0071] It should be noted that, since the target vehicle in this embodiment is a truck that needs to be inspected for overloading, it is usually required that the license plate number be sprayed on the side of the truck for easier management. Therefore, based on the scenario of this embodiment, the license plate number is sprayed on the side of the target vehicle so that the target vehicle can be identified by the license plate number on the side of the vehicle.

[0072] In the above implementation, the license plate number of the target vehicle is first preliminarily identified based on the vehicle image, and the driving parameters of the target vehicle whose license plate number is not identified are obtained to detect whether the target vehicle whose license plate number is not identified is a continuous convoy. If it is a continuous convoy, it means that the license plate number may not be identified due to the occlusion of adjacent target vehicles. At this time, the license plate number of the target vehicle on the side of the vehicle body in the vehicle image is identified, thereby realizing the acquisition of the license plate number of the target vehicle in the middle of the continuous convoy, thus achieving the effect of vehicle identification.

[0073] As one implementation of step S101, step S101 specifically includes:

[0074] Step S1011: Receive the first shooting signal and the second shooting signal; the first shooting signal and the second shooting signal are output by the geomagnetic ring sensor when the target vehicle passes by.

[0075] Continue to refer to Figure 2The geomagnetic ring sensor is provided with two, including a first geomagnetic ring 4 and a second geomagnetic ring 5, both of which are arranged under the road surface in a buried manner, specifically, a groove with a width of 8mm, a depth of 40mm and a length equal to the width of the road surface is opened on the road surface along the width direction of the road surface for placing the first geomagnetic ring 4 and the second geomagnetic ring 5, and the first geomagnetic ring 4 and the second geomagnetic ring 5 are respectively connected with the electric control box connection line every 1m with 20 turns.

[0076] Step S1012: in response to the first shooting signal to control the first camera 1 to shoot the first shooting image, and in response to the second shooting signal to control the second camera 3 to shoot the second shooting image.

[0077] It should be understood that when the target vehicle passes through the first geomagnetic ring 4 or the second geomagnetic ring 5, the metal accessories of the target vehicle itself will change the value output by the first geomagnetic ring 4 or the second geomagnetic ring 5, thereby realizing the generation of the first shooting signal and the generation of the second shooting signal.

[0078] Step S1013: according to the first shooting image and the second shooting image, the vehicle image of the target vehicle is obtained.

[0079] In the above embodiment, the first shooting signal or the second shooting signal is output by the geomagnetic ring sensor when the target vehicle passes through, so as to realize the control of the first camera 1 and the second camera 3 to shoot the vehicle image when the target vehicle passes through the geomagnetic ring sensor.

[0080] Referring to Figure 4 As an embodiment of step S102, step S102 specifically includes:

[0081] Step S1021: morphological processing is performed on the vehicle image, the vehicle image is segmented to obtain a plurality of connected regions.

[0082] Specifically, the morphological processing includes erosion, dilation, opening operation and closing operation, and the execution order and execution times of the erosion, dilation, opening operation and closing operation are not limited here, and can be adjusted accordingly according to the actual situation.

[0083] Step S1022: analyze the shape features of each connected region, and screen the connected regions according to the shape features to obtain the vehicle component region.

[0084] Among them, the shape features include the area, perimeter, aspect ratio and convex hull feature of the connected region, the convex hull feature refers to a polygon that completely contains the original contour of the connected region and is only composed of points on the contour, and the convex hull feature is mainly used to simplify the approximation polygon.

[0085] Step S1023: clustering the adjacent or overlapping vehicle component regions respectively to obtain one or more candidate vehicle regions.

[0086] Step S1024: performing edge detection on each candidate vehicle region to obtain a license plate region, and combining the license plate region and the candidate vehicle region.

[0087] The algorithm used in the edge detection includes gradient operator, Laplace operator, Canny operator and Hough transform method. The gradient operator represents the rate and direction of change of a function at a given point. Common gradient operators include Sobel operator, Prewitt operator and Roberts operator, etc. The above operators obtain the edge information of the image by calculating the gradient of the image in different directions. The Laplace operator represents the curvature of a function at a given point. The edge of the image is detected by calculating the second derivative of the image. The Canny operator is a combination of the gradient operator and the Laplace operator.

[0088] The Hough transform identifies the features in the object from the image through feature extraction. In this embodiment, the Hough transform accurately finds the license plate region that meets the rectangular feature according to the edge features obtained by the gradient operator, the Laplace operator or the Canny operator.

[0089] Step S1025: in response to the existence of a candidate vehicle region that is not combined with the license plate region, determining that the corresponding candidate vehicle region is a target vehicle whose license plate number is not recognized.

[0090] Step S1026: in response to the candidate vehicle region combined with the license plate region, identifying the license plate region to obtain the license plate number of the target vehicle corresponding to the candidate vehicle region.

[0091] The identification of the license plate region includes number recognition and color recognition of the license plate region. The number recognition can be realized by using the text recognition technology. The color recognition of the license plate region specifically includes extracting color features according to a color space model and classifying according to the color features to realize the classification of the color of the license plate region. Specifically, the color space model can be selected from one of RGB (Red, Green, Blue), HSV (Hue, Saturation, Value) and YCbCr. The classification of the color of the license plate region can be realized by using the color moment method.

[0092] In the above embodiment, the vehicle image is segmented to obtain a plurality of connected regions, the connected regions are screened to obtain vehicle component regions, clustering of the vehicle component regions can obtain candidate vehicle regions, the license plate region is identified through edge detection, so as to combine the license plate region and the candidate vehicle region corresponding to the license plate region, when there is a candidate vehicle region that is not combined with the license plate region, it indicates that the license plate number corresponding to the candidate vehicle region cannot be obtained, and then the obtained license plate region is identified, that is, the identification of the license plate number is realized.

[0093] As a further embodiment of step S102, the pre-processing of the vehicle image is further included before step S1021, and the pre-processing includes image enhancement, denoising and binarization operation.

[0094] Specifically, the image enhancement can be processed by a histogram equalization formula, and the histogram equalization formula is: Q(k) = n k ; Wherein, k = 0, 1, …, L-1;

[0095] Wherein, k represents each gray level of the vehicle image, Q(k) represents the number of each gray level pixel in the vehicle image, n k is the number of pixels with gray level k in the vehicle image, and L is the number of gray level division of the vehicle image; Then the probability of the kth gray division is: The gray value after histogram equalization transformation can be represented as: D is the pixel space depth, usually 255, n i is the number of pixels with gray level i, so the vehicle image after histogram equalization operation is obtained.

[0096] Specifically, the denoising can use the method of median filtering to denoise the image, wherein the specific steps of median filtering are as follows: the candidate vehicle region not combined with the license plate region is represented as vehicle sequence x1, x2,..., x n , taking the window length as m, median filtering is performed on it, that is, m numbers X i -p,...,X i -1,X i ,X i +1,...,X i +p (where xi is the window center point value, ), then the m points are sorted according to their numerical values, and the number with the center point sequence number in the vehicle sequence is taken as the filtering output, and the median filtering formula is: z1=Median{X i -p,...,X i -1,X i ,X i +1,...,X i+ p} ; i e N, wherein p is the median of the vehicle sequence.

[0097] Specifically, the binarization of the vehicle image can adopt an adaptive threshold method, which is calculated by an adaptiveThreshold function.

[0098] The adaptive threshold calculation step is to calculate the threshold value for each pixel point individually, that is, the threshold value of each pixel point is different, that is, the weighted average of the pixels in the surrounding area of the pixel point is obtained, and then a constant o is subtracted, so as to obtain the threshold value of the point. The region is Bx B, wherein B represents the region size, and B>0; the method of the adaptive threshold method is the adaptiveThreshold function, and the thresholdType specifies the threshold type, and the threshold value is equal to the average value or the weighted average value minus the constant.

[0099] Referring to Figure 5 As a further embodiment of the license plate recognition method, after recognizing the license plate number of the target vehicle according to the vehicle image, the method further comprises steps S201-S202, which will be described in detail below.

[0100] Step S201: In response to the target vehicle whose license plate number is recognized, a license plate region image is obtained.

[0101] Step S202: inputting the license plate region image into a license plate verification model, extracting license plate feature information of the license plate region image in a plurality of dimensions through the license plate verification model, and outputting a verification result of the license plate region image according to the license plate feature information in the plurality of dimensions.

[0102] The license plate feature information in the plurality of dimensions includes the color, font, and character spacing characteristics of the license plate.

[0103] The license plate feature information in the plurality of dimensions of the license plate region image is judged by a feature extraction technology using a classification algorithm. Specifically, the feature extraction technology can be extracted in the form of convolution and Zernike moment.

[0104] In the above embodiment, for the target vehicle whose license plate number is recognized, the license plate verification model is used to extract the license plate feature information in the plurality of dimensions, so as to verify the authenticity of the license plate from the license plate feature information in the plurality of dimensions, and ensure the effectiveness of the obtained license plate number.

[0105] As a further embodiment of the license plate recognition method, the license plate recognition method further comprises a training step of training a preset convolutional neural network to obtain the license plate verification model, and the training step comprises:

[0106] Referring to Figure 6, step S301: obtaining a training data set, the training data set including a plurality of sample images, the sample images including real license plate images and fake license plate images.

[0107] The fake license plate images in the training data set are created according to the color, font and character spacing features of the license plate.

[0108] Step S302: labeling the sample images with license plate feature information in multiple dimensions, and labeling the sample images with corresponding true-false labels according to the labeled license plate feature information in multiple dimensions.

[0109] Step S303: inputting the sample images into the preset convolutional neural network to output true-false prediction results, and iteratively updating the model parameters of the preset convolutional neural network according to the true-false labels and the true-false prediction results to obtain a license plate verification model.

[0110] The convolutional neural network can use a CNN model, and the trained weights and offset values are used as the initial values of the model parameters of the preset convolutional neural network through transfer learning.

[0111] In the above embodiment, the real license plate images and the fake license plate images in the training data set are labeled with license plate feature information in multiple dimensions, and the sample images are labeled with corresponding true-false labels according to the labeled license plate feature information. The sample images are input into the preset convolutional neural network to obtain true-false prediction results, and the model parameters of the convolutional neural network are iteratively updated according to the true-false labels and the true-false prediction results, thereby generating a license plate verification model that can identify the authenticity of the license plate region.

[0112] Referring to Figure 7 As an embodiment of obtaining the driving parameters of the target vehicle whose license plate number is not recognized, the driving parameters of the target vehicle whose license plate number is not recognized specifically include:

[0113] Step S1031: numbering each target vehicle whose license plate number is not recognized.

[0114] It should be noted that the numbering of each target vehicle whose license plate number is not recognized is achieved by numbering the candidate vehicle regions that are not combined with the license plate region. Specifically, the gravity center, direction and principal axis features of the candidate vehicle regions that are not combined with the license plate region are calculated and sorted. The sorting can be arranged according to the horizontal coordinates and vertical coordinates of the gravity center in sequence. Further, let Str be an empty string, and c be an initial character. For each candidate vehicle region X iadd c to the end of Str and add one to c, and then assign Str to the i-th candidate vehicle region that is not combined with the license plate region, so as to number each target vehicle whose license plate number is not recognized by the magnitude of the candidate vehicle region that is not combined with the license plate region.

[0115] Step S1032: According to the number and the coordinates of the target vehicles, the driving distance between each target vehicle and the adjacent front and rear target vehicles is calculated respectively.

[0116] Specifically, for each target vehicle i whose license plate number is not recognized, the distances h1 and h2 between the target vehicle and the front and rear target vehicles whose license plate numbers are not recognized are calculated respectively as follows:

[0117] h1 = sqrt((xi-x(i-1)) 2 +(yi-y(i-1)) 2 );

[0118] h1 = sqrt((xi-x(i+1)) 2 +(yi-y(i+1)) 2 );

[0119] where x i is the number of the target vehicle whose license plate number is not recognized, from 1 to u; u is the total number of the target vehicles whose license plate numbers are not recognized, x i and y i are the coordinates of the i-th vehicle.

[0120] Step S1033: The average speed and the average direction of all the target vehicles whose license plate numbers are not recognized are obtained.

[0121] Specifically, the average speed and the average direction of the target vehicles whose license plate numbers are not recognized are calculated as follows: where z is the number of all the target vehicles whose license plate numbers are not recognized, sum represents summation, vi is the speed of the i-th target vehicle whose license plate number is not recognized, and e i is the direction of the i-th target vehicle whose license plate number is not recognized.

[0122] Step S1034: The driving parameters of the target vehicles whose license plate numbers are not recognized are obtained according to the driving distance, the average speed and the average direction.

[0123] It should be noted that when the target vehicle and each adjacent front and rear target vehicle have a driving distance h1 and h2 less than or equal to a preset threshold, and the average speed and the average direction are within a preset range, it is considered that the target vehicle and each adjacent front and rear target vehicle belong to a continuous vehicle fleet; the preset range of the average speed is that the average speed vk is between 0 and 100, and the preset range of the average direction is that the average direction ek is between 0 and 360.

[0124] In the above embodiment, each target vehicle whose license plate number is not recognized is numbered to distinguish each target vehicle whose license plate number is not recognized, so as to calculate the driving distance between each target vehicle and each adjacent front and rear target vehicle, and obtain the driving parameters of the target vehicle whose license plate number is not recognized based on the driving distance and the obtained average speed and average direction.

[0125] As a further embodiment of the license plate recognition method, the license plate recognition method further comprises:

[0126] In response to the target vehicle whose license plate number is not recognized not being a continuous vehicle fleet or in response to the license plate number of the side of the target vehicle body not being recognized, the vehicle image is stored and sent to the management terminal.

[0127] In the above embodiment, when the target vehicle whose license plate number is not recognized is not a continuous vehicle fleet or the target vehicle whose license plate number is not recognized is not a continuous vehicle fleet, it is considered as an abnormal situation, and the vehicle image at this time is sent to the management terminal for manual review.

[0128] In addition, the embodiment of the present application discloses a vehicle recognition system. The license plate recognition system can be applied to a recognition device, and in the embodiment, the recognition device can include a license plate recognition system, a machine readable storage medium and a processor.

[0129] In the embodiment, the machine readable storage medium and the processor can be located in the recognition device and separated. The machine readable storage medium can also be independent of the recognition device and accessed by the processor. The license plate recognition system can include a plurality of functional modules stored in the machine readable storage medium, for example, each software functional module included in the license plate recognition system. When the processor executes the computer program corresponding to the software functional module in the license plate recognition system, the license plate recognition system provided by the foregoing method embodiment is realized.

[0130] In some embodiments, the processor can include one or more processing engines (e.g., single core processors or multi-core processors). For example only, the processor can include one or more of a hardware processor, such as a central processing unit (CPU), an application-specific integrated circuit (ASIC), an application-specific instruction-set processor (ASIP), a graphics processing unit (GPU), a physics processing unit (PPU), a digital signal processor (DSP), a field programmable gate array (FPGA), a programmable logic device (PLD), a controller, a microcontroller unit, a reduced instruction set computer (RISC), a microprocessor, and / or the like, or any combination thereof.

[0131] The machine-readable storage medium can store data and / or instructions. In some embodiments, the machine-readable storage medium can store obtained data or information. In some embodiments, the machine-readable storage medium can store data and / or instructions for execution or use by the identification device, which can implement the example methods described herein by executing or using the data and / or instructions. In some embodiments, the machine-readable storage medium can include a mass storage, a removable storage, a volatile read / write memory, a read-only memory (ROM), and / or the like, or any combination thereof. An example mass storage can include a magnetic disk, an optical disk, a solid-state disk, and / or the like. An example removable storage can include a flash drive, a floppy disk, an optical disk, a memory card, a compact disk, a magnetic tape, and / or the like. An example volatile read / write memory can include a random access memory (RAM). An example random access memory can include a dynamic RAM, a double data rate synchronous dynamic RAM, a static RAM, a thyristor RAM, and a zero capacitor RAM, and / or the like. An example ROM can include a mask ROM, a programmable ROM, an erasable programmable ROM, an electronically erasable programmable ROM, a compact disk ROM, and a digital versatile disk ROM, and / or the like.

[0132] In some embodiments, the vehicle identification system included in the identification device can include one or more software functional modules. The software functional modules can be stored in the machine-readable storage medium as programs or instructions, which, when executed by the corresponding processor, are used to implement the above-described methods, such as when executed by the processor of the unmanned aerial vehicle, are used to implement the method steps performed by the unmanned aerial vehicle, or when executed by the identification device, are used to implement the method steps performed by the identification device.

[0133] In detail, the embodiments of the present application disclose a vehicle identification system. Referring to Figure 8 A vehicle identification system includes an image acquisition module configured to acquire a vehicle image of a target vehicle.

[0134] The first license plate recognition module is configured to recognize the license plate number of the target vehicle according to the vehicle image.

[0135] The vehicle fleet detection module is configured to, in response to the target vehicle having no recognized license plate number, acquire a driving parameter of the target vehicle having no recognized license plate number, and perform continuous vehicle fleet detection on the target vehicle according to the driving parameter.

[0136] The second license plate recognition module is configured to, in response to the target vehicle being a continuous vehicle fleet, recognize the license plate number on the side of the vehicle body of the target vehicle based on the vehicle image.

[0137] The vehicle recognition system provided in the present application can implement the vehicle recognition method described above, and the specific working process of the vehicle recognition system can refer to the corresponding process in the method embodiments described above.

[0138] It should be noted that in the above embodiments, the description of each embodiment has its own emphasis, and the parts not described in detail in a certain embodiment can be referred to the related description of other embodiments.

[0139] Based on the same technical concept, the present application further discloses a computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor executes the computer program of any one of the above methods.

[0140] The present application further discloses a computer readable storage medium, a computer readable storage medium comprising a computer program capable of being loaded and executed by a processor to perform any one of the above methods.

[0141] In the embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are only schematic. For example, the division of the units is only a logical function division, and actual implementation can have another division manner. For example, a plurality of units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the coupling or direct coupling or communication connection between the units shown or discussed can be indirect coupling or communication connection through some communication interfaces, devices or units, and can be electrical, mechanical or other forms.

[0142] In addition, each functional module in each embodiment of the present application can be integrated together to form an independent part, or each module can exist independently, or two or more modules can be integrated to form an independent part.

[0143] The above are only preferred embodiments of the present application, not intended to limit the protection scope of the present application, any feature disclosed in the specification (including the abstract and the drawings) can be replaced by other equivalent or similar features, unless specifically stated otherwise. That is, each feature is only an example of a series of equivalent or similar features, unless specifically stated otherwise.

Claims

1. A vehicle identification method characterized by, The method comprises: acquiring a vehicle image of a target vehicle in a target area; identifying a license plate number of the target vehicle according to the vehicle image; in response to the existence of a target vehicle whose license plate number is not identified, numbering each target vehicle whose license plate number is not identified; calculating a driving distance between each target vehicle and each adjacent target vehicle in front or behind the target vehicle according to the number and coordinates of the target vehicle; acquiring an average speed and an average direction of all target vehicles whose license plate numbers are not identified; obtaining a driving parameter of the target vehicle whose license plate number is not identified according to the driving distance, the average speed and the average direction, and performing continuous vehicle fleet detection on the target vehicle according to the driving parameter; wherein the continuous vehicle fleet refers to a vehicle fleet in which the heads and tails of a plurality of target vehicles are arranged in sequence and closely. in response to the plurality of target vehicles being a continuous vehicle fleet, identifying a license plate number on a side of a vehicle body of the target vehicle based on the vehicle image.

2. The method of claim 1, wherein, The vehicle image comprises a first shooting image shot by a first camera and a second shooting image shot by a second camera; the first camera is used to shoot one of a head or a tail of the target vehicle, and the second camera is used to shoot the other of the head or the tail of the target vehicle. The identification of the license plate number of the target vehicle according to the vehicle image specifically comprises: performing morphological processing on the vehicle image, segmenting the vehicle image to obtain a plurality of connected regions; analyzing shape features of each connected region, and screening the connected regions according to the shape features to obtain vehicle component regions; clustering adjacent or overlapping vehicle component regions to obtain one or more candidate vehicle regions; performing edge detection on each candidate vehicle region to obtain a license plate region, and combining the license plate region and the candidate vehicle region; in response to the existence of a candidate vehicle region that is not combined with the license plate region, determining that the corresponding candidate vehicle region is a target vehicle whose license plate number is not identified; in response to a candidate vehicle region combined with the license plate region, identifying the license plate region to obtain a license plate number of a target vehicle corresponding to the candidate vehicle region.

3. The method of claim 1, wherein, After the identification of the license plate number of the target vehicle according to the vehicle image, the method further comprises: in response to a target vehicle whose license plate number is identified, acquiring a license plate region image; inputting the license plate region image into a license plate verification model, extracting license plate feature information of the license plate region image in a plurality of dimensions through the license plate verification model, and outputting a verification result of the license plate region image according to the license plate feature information in the plurality of dimensions.

4. The method of claim 3, wherein, The method further comprises a training step of training a preset convolutional neural network to obtain the license plate verification model, and the training step comprises: acquiring a training data set, the training data set comprising a plurality of sample images, real license plate images and fake license plate images; annotating license plate feature information in a plurality of dimensions on the sample images, and labeling corresponding true or false labels on the sample images according to the annotated license plate feature information in the plurality of dimensions; The sample image is input into the preset convolutional neural network to output a true or false prediction result, and the model parameters of the preset convolutional neural network are iteratively updated according to the true or false label and the true or false prediction result, so as to obtain the license plate verification model.

5. The method of claim 2, wherein, The vehicle image of the target vehicle in the target area is obtained, and specifically includes: Receiving a first shooting signal and a second shooting signal; The first shooting signal and the second shooting signal are output by the geomagnetic ring sensor when the target vehicle passes through; In response to the first shooting signal, the first camera is controlled to shoot a first shooting image, and in response to the second shooting signal, the second camera is controlled to shoot a second shooting image; According to the first shooting image and the second shooting image, the vehicle image of the target vehicle is obtained.

6. The method of claim 1 or 2, wherein, Also includes: In response to the target vehicle whose license plate number is not recognized not being a continuous vehicle team or in response to the target vehicle whose license plate number is not recognized on the side of the vehicle body, the vehicle image is stored and sent to the management terminal.

7. A vehicle identification system characterized by, Including: An image acquisition module is configured to acquire a vehicle image of a target vehicle; A first license plate recognition module is configured to recognize a license plate number of the target vehicle according to the vehicle image; A vehicle team detection module is configured to number each target vehicle whose license plate number is not recognized in response to the existence of the target vehicle whose license plate number is not recognized; According to the number and coordinates of the target vehicle, the driving distance between each target vehicle and each adjacent target vehicle in front and behind is calculated respectively; The average speed and average direction of all target vehicles whose license plate numbers are not recognized are obtained; According to the driving distance, the average speed and the average direction, the driving parameters of the target vehicle whose license plate number is not recognized are obtained, and the target vehicle is continuously detected according to the driving parameters; wherein the continuous vehicle team refers to a vehicle team in which the heads and tails of a plurality of target vehicles are closely arranged in sequence; A second license plate recognition module is configured to recognize the license plate number on the side of the vehicle body of the target vehicle based on the vehicle image in response to the target vehicle being a continuous vehicle team.

8. An identification device, characterized by: A memory, a processor, and a computer program stored on the memory and executable on the processor are included, and the processor executes the computer program of any one of claims 1-6.

9. A computer readable storage medium comprising a computer program capable of being loaded and executed by a processor to perform any one of the methods of claims 1-6.

Images