A novel motor vehicle license plate detection and recognition method, system and readable storage medium

Abstract

The present invention discloses a novel motor vehicle license plate detection and recognition method, system, and readable storage medium. The method comprises: acquiring detection image data; inputting the detection image data into a preset motor vehicle license plate detection and recognition model; analyzing the detection image data to obtain license plate area image data; analyzing the license plate area image data to obtain predicted binary code data and license plate color data; verifying the predicted binary code data using a preset method; if the verification is successful, decoding the predicted binary code data to obtain license plate data; and analyzing the license plate data and license plate color data to obtain motor vehicle model data. Based on deep learning-based image recognition technology, the present invention introduces a priori rules for motor vehicle coding and combines verification rules for data packet transmission in communication technology with motor vehicle license plate recognition, effectively reducing the error rate of motor vehicle license plate recognition.

Description

Technical Field

[0001] The present application relates to the field of data processing and data transmission, and more specifically, to a novel motor vehicle license plate detection and recognition method, system, and readable storage medium. Background Art

[0002] The detection and recognition of motor vehicle license plates is a crucial component of parking gate systems and traffic violation detection systems, and research on this topic holds significant application value. Before the advent of deep learning-based image processing technology, license plate detection primarily relied on character segmentation coupled with machine learning classifiers. The primary machine learning classifier used was the Support Vector Machine (SVM). However, the accuracy of this approach is significantly affected by environmental factors, lighting, and occlusion.

[0003] After achieving promising results with deep learning-based image processing technology, it has also been applied to the detection and recognition of motor vehicle license plates, achieving good results. However, the current main recognition algorithm classifies all characters on a license plate as a detection class, without considering the actual rules of license plate encoding.

[0004] Therefore, the prior art has defects and is in urgent need of improvement. Summary of the Invention

[0005] In view of the above problems, the purpose of the present invention is to provide a new motor vehicle license plate detection and recognition method, system and readable storage medium, which can improve the recognition rate of motor vehicle license plates.

[0006] The first aspect of the present invention provides a novel method for detecting and identifying motor vehicle license plates, comprising:

[0007] Acquire detection image data;

[0008] Inputting the detection image data into a preset motor vehicle license plate detection and recognition model;

[0009] The preset motor vehicle license plate detection and recognition model analyzes the detection image data to obtain license plate area image data;

[0010] The preset motor vehicle license plate detection and recognition model analyzes the license plate area image data to obtain predicted binary code data and license plate color data;

[0011] Verifying the predicted binary coded data by a preset method;

[0012] If the verification is successful, the predicted binary coded data is decoded to obtain the license plate data;

[0013] The license plate data and the license plate color data are analyzed to obtain the motor vehicle model data.

[0014] In this solution, the verification of the predicted binary coded data by a preset method includes:

[0015] Analyze the predicted binary coded data to obtain license plate binary coded data and parity bit coded data;

[0016] Performing parity statistics on the license plate binary coded data, and comparing the obtained parity statistics with the parity check bit coded data to determine whether the parity statistics and the parity check bit coded data are consistent;

[0017] If they are consistent, it means the prediction is correct and the license plate data is output;

[0018] If they are inconsistent, it means the prediction is wrong and the license plate data will not be output.

[0019] This plan also includes:

[0020] Obtain sample data of motor vehicle license plates;

[0021] Dividing the motor vehicle license plate data in the motor vehicle license plate sample data into a provincial administrative region abbreviation portion and a mixed portion of Arabic numerals, English letters, and special Chinese characters;

[0022] The provincial administrative region abbreviation part and the mixed part of Arabic numerals, English letters and special Chinese characters are binary-encoded respectively to establish a motor vehicle license plate binary coding database.

[0023] This plan also includes:

[0024] Obtain license plate sample image data;

[0025] Analyzing and training the license plate sample image data to establish a preset motor vehicle license plate detection and recognition model;

[0026] The preset motor vehicle license plate detection and recognition model is optimized through a loss function.

[0027] In this solution, the preset motor vehicle license plate detection and recognition model is optimized by the loss function, specifically:

[0028] The loss function is expressed as follows:

[0029] ;

[0030] Among them, L represents the loss function, and is the loss function of the detection part, is the loss function of the binary coding part, is the loss function of the license plate color classification part; α, β, γ, and λ are all influence coefficients, and i is the number of binary codes.

[0031] In this solution, the analysis of the detected image data to obtain the license plate area image data includes:

[0032] Analyze the detection image data to determine whether there is a motor vehicle license plate;

[0033] If it exists, the motor vehicle license plate is located and the motor vehicle license plate area is intercepted to obtain license plate area image data; if it does not exist, filtering is performed.

[0034] In this solution, decoding the predicted binary coded data to obtain license plate data includes:

[0035] Segmenting the predicted binary code data according to a preset template to obtain a plurality of sub-character code data;

[0036] Comparing the plurality of sub-character coded data with the standard data of the motor vehicle license plate binary code database respectively, decoding the plurality of sub-character coded data to obtain a plurality of sub-character data;

[0037] The plurality of sub-character data are integrated to obtain license plate data.

[0038] A second aspect of the present invention provides a novel motor vehicle license plate detection and recognition system, comprising a memory and a processor, wherein the memory includes a novel motor vehicle license plate detection and recognition method program, and when the novel motor vehicle license plate detection and recognition method program is executed by the processor, the following steps are implemented:

[0039] Acquire detection image data;

[0040] Inputting the detection image data into a preset motor vehicle license plate detection and recognition model;

[0041] The preset motor vehicle license plate detection and recognition model analyzes the detection image data to obtain license plate area image data;

[0042] The preset motor vehicle license plate detection and recognition model analyzes the license plate area image data to obtain predicted binary code data and license plate color data;

[0043] Verifying the predicted binary coded data by a preset method;

[0044] If the verification is successful, the predicted binary coded data is decoded to obtain the license plate data;

[0045] The license plate data and the license plate color data are analyzed to obtain the motor vehicle model data.

[0046] In this solution, the verification of the predicted binary coded data by a preset method includes:

[0047] Analyze the predicted binary coded data to obtain license plate binary coded data and parity bit coded data;

[0048] Performing parity statistics on the license plate binary coded data, and comparing the obtained parity statistics with the parity check bit coded data to determine whether the parity statistics and the parity check bit coded data are consistent;

[0049] If they are consistent, it means the prediction is correct and the license plate data is output;

[0050] If they are inconsistent, it means the prediction is wrong and the license plate data will not be output.

[0051] This plan also includes:

[0052] Obtain sample data of motor vehicle license plates;

[0053] Dividing the motor vehicle license plate data in the motor vehicle license plate sample data into a provincial administrative region abbreviation portion and a mixed portion of Arabic numerals, English letters, and special Chinese characters;

[0054] The provincial administrative region abbreviation part and the mixed part of Arabic numerals, English letters and special Chinese characters are binary-encoded respectively to establish a motor vehicle license plate binary coding database.

[0055] This plan also includes:

[0056] Obtain license plate sample image data;

[0057] Analyzing and training the license plate sample image data to establish a preset motor vehicle license plate detection and recognition model;

[0058] The preset motor vehicle license plate detection and recognition model is optimized through a loss function.

[0059] In this solution, the preset motor vehicle license plate detection and recognition model is optimized by the loss function, specifically:

[0060] The loss function is expressed as follows:

[0061] ;

[0062] Among them, L represents the loss function, and is the loss function of the detection part, is the loss function of the binary coding part, is the loss function of the license plate color classification part; α, β, γ, and λ are all influence coefficients, and i is the number of binary codes.

[0063] In this solution, the analysis of the detected image data to obtain the license plate area image data includes:

[0064] Analyze the detection image data to determine whether there is a motor vehicle license plate;

[0065] If it exists, the motor vehicle license plate is located and the motor vehicle license plate area is intercepted to obtain license plate area image data; if it does not exist, filtering is performed.

[0066] In this solution, decoding the predicted binary coded data to obtain license plate data includes:

[0067] Segmenting the predicted binary code data according to a preset template to obtain a plurality of sub-character code data;

[0068] Comparing the plurality of sub-character coded data with the standard data of the motor vehicle license plate binary code database respectively, decoding the plurality of sub-character coded data to obtain a plurality of sub-character data;

[0069] The plurality of sub-character data are integrated to obtain license plate data.

[0070] The third aspect of the present invention provides a computer-readable storage medium, which includes a new motor vehicle license plate detection and recognition method program. When the new motor vehicle license plate detection and recognition method program is executed by a processor, the steps of the new motor vehicle license plate detection and recognition method as described in any one of the above items are implemented.

[0071] The present invention discloses a novel motor vehicle license plate detection and recognition method, system, and readable storage medium. The method comprises: acquiring detection image data; inputting the detection image data into a preset motor vehicle license plate detection and recognition model; analyzing the detection image data to obtain license plate area image data; analyzing the license plate area image data to obtain predicted binary code data and license plate color data; verifying the predicted binary code data using a preset method; if the verification is successful, decoding the predicted binary code data to obtain license plate data; and analyzing the license plate data and license plate color data to obtain motor vehicle model data. Based on deep learning-based image recognition technology, the present invention introduces a priori rules for motor vehicle coding and combines verification rules for data packet transmission in communication technology with motor vehicle license plate recognition, effectively reducing the error rate of motor vehicle license plate recognition. BRIEF DESCRIPTION OF THE DRAWINGS

[0072] Figure 1 A flow chart showing a novel motor vehicle license plate detection and recognition method of the present invention is shown;

[0073] Figure 2 A flowchart of a binary coded data verification method according to the present invention is shown;

[0074] Figure 3 A flow chart showing a method for training a preset motor vehicle license plate detection and recognition model according to the present invention is shown;

[0075] Figure 4 A block diagram of a novel motor vehicle license plate detection and recognition system according to the present invention is shown;

[0076] Figure 5 A flowchart of network detection and recognition of motor vehicle license plates according to the present invention is shown. DETAILED DESCRIPTION

[0077] In order to more clearly understand the above-mentioned objects, features and advantages of the present invention, the present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be noted that, in the absence of conflict, the embodiments of the present application and the features therein can be combined with each other.

[0078] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Therefore, the scope of protection of the present invention is not limited to the specific embodiments disclosed below.

[0079] Figure 1 The flowchart of a novel motor vehicle license plate detection and recognition method of the present invention is shown.

[0080] like Figure 1 As shown, the present invention discloses a novel motor vehicle license plate detection and recognition method, comprising:

[0081] S102, acquiring detection image data;

[0082] S104, inputting the detection image data into a preset motor vehicle license plate detection and recognition model;

[0083] S106, a preset motor vehicle license plate detection and recognition model analyzes the detection image data to obtain license plate area image data;

[0084] S108, using a preset motor vehicle license plate detection and recognition model to analyze the license plate area image data to obtain predicted binary code data and license plate color data;

[0085] S110, verifying the predicted binary coded data using a preset method;

[0086] S112, if the verification is successful, decoding the predicted binary coded data to obtain license plate data;

[0087] S114: Analyze the license plate data and the license plate color data to obtain vehicle model data.

[0088] According to an embodiment of the present invention, a motor vehicle license plate binary encoding database of motor vehicle license plate binary encoding is first established based on the motor vehicle license plate encoding rules. Then, based on the established motor vehicle license plate binary encoding database, this scheme designs a convolutional neural network Alpha-LPDRNet (Alpha-License Plate Detection Recognition Network, Alpha-LPDRNet), such as Figure 5 As shown, the detection and recognition of motor vehicle license plates is achieved through three aspects: first, the position of the motor vehicle license plate is located in the image; second, the color data of the motor vehicle license plate is output; and third, the binary code of the motor vehicle license plate is output. The output motor vehicle license plate binary code is then decoded based on a motor vehicle license plate binary code database to obtain the motor vehicle license plate data. Combined with the motor vehicle license plate color data, the motor vehicle license plate is recognized and the vehicle model is determined. In addition, the present invention also adds a parity bit at the end of the character code to verify the prediction results output by the model, thereby improving the model recognition accuracy and reducing the error rate of motor vehicle license plate recognition.

[0089] Figure 2 A flow chart of a binary coded data verification method of the present invention is shown.

[0090] like Figure 2 As shown, according to an embodiment of the present invention, verifying the predicted binary coded data by a preset method includes:

[0091] S202, analyzing the predicted binary coded data to obtain license plate binary coded data and parity bit coded data;

[0092] S204, performing parity statistics on the license plate binary coded data, and comparing the obtained parity statistics with the parity check bit coded data to determine whether the parity statistics and the parity check bit coded data are consistent;

[0093] S206, if they are consistent, it means the prediction is correct, and the license plate data is output;

[0094] S208: If they are inconsistent, it means the prediction is wrong and the license plate data is not output.

[0095] It should be noted that the binary-encoded data packet consists of two parts: the binary-encoded data of the number plate and the parity-encoded data. The parity-encoded data is obtained by counting the number of 1s in the binary number of each binary-encoded license plate. If the number of 1s is odd, the parity bit is 1; if the number of 1s is even, the parity bit is 0. During network training, this parity bit is represented by a 1x2 tensor. If the parity bit is 1, its label is (0, 1); if the parity bit is 0, its label is (1, 0). Cross entropy is used as the loss function. During inference, if the tensor outputs (1, 0), it indicates that the predicted parity-encoded data is 0; if the tensor outputs (0, 1), it indicates that the predicted parity-encoded data is 1. Then, a parity count is performed on the output license plate binary coded data. The number of 1s in the license plate binary coded data is counted. If the number of 1s is odd, the parity count is 1; if the number of 1s is even, the parity count is 0. The parity count is compared with the parity check bit coded data. If they are consistent, it means that the license plate prediction is correct and the prediction result is output; if they are inconsistent, it means that the license plate prediction is incorrect and the prediction result is not output.

[0096] According to an embodiment of the present invention, the further embodiment includes:

[0097] Obtain sample data of motor vehicle license plates;

[0098] Dividing the motor vehicle license plate data in the motor vehicle license plate sample data into a provincial administrative region abbreviation portion and a mixed portion of Arabic numerals, English letters, and special Chinese characters;

[0099] The provincial administrative region abbreviation part and the mixed part of Arabic numerals, English letters and special Chinese characters are binary-encoded respectively to establish a motor vehicle license plate binary coding database.

[0100] It should be noted that motor vehicle license plates generally consist of three parts: the abbreviation of the provincial administrative region, the code of the prefecture-level administrative region, and a random code of Arabic numerals and English letters. Figure 3 A flow chart of a method for training a preset motor vehicle license plate detection and recognition model according to the present invention is shown.

[0101] like Figure 3 As shown, according to an embodiment of the present invention, it also includes:

[0102] S302, obtaining license plate sample image data;

[0103] S304, analyzing and training the license plate sample image data to establish a preset motor vehicle license plate detection and recognition model;

[0104] S306: Optimize the preset motor vehicle license plate detection and recognition model through a loss function.

[0105] It should be noted that a license plate detection and recognition network requires three functions: first, locating the license plate in an image; second, outputting the license plate's color; and third, outputting the license plate's binary code. To achieve these functions, the present invention designed a convolutional neural network, Alpha-LPDRNet (Alpha-LicensePlate Detection Recognition Network), with specific parameters shown in Table 1.

[0106] The convolutional neural network Alpha-LPDRNet is mainly composed of convolutional layer (Convolutional Layer), normalization layer (Batch Normalization), non-linear activation layer (RELU), CSP module (cross stage Partial), fully connected layer (Fully Connected Layer) and other parts.

[0107] During the training process, Alpha-LPDRNet performs random contrast, brightness, and saturation adjustments on the original image, as well as image enhancement techniques such as small-angle rotation and random occlusion. After the Alpha-LPDRNet network detection portion completes positioning, the corresponding area does not require regional correction like traditional license plate recognition algorithms, nor does it require character segmentation of the feature map. Instead, each layer directly outputs the binary code, parity bit code, and license plate color of the corresponding position. This not only incorporates prior knowledge of license plate codes but also improves the execution efficiency of edge devices. While improving recognition accuracy, it also reduces user waiting time and enhances the user experience.

[0108] During the training process, it is necessary to mark the position and color of the motor vehicle license plate in the image and perform binary encoding on the motor vehicle license plate. The marked file and binary encoding are then used as supervision data for network training.

[0109] In Alpha-LPDRNet, Conv1 through Out3 layers are the feature extraction and detection components, locating the license plate in the image and acquiring high-dimensional features for the corresponding region. The Out4 layer outputs a 6-bit binary code representing the provincial administrative region. The Out5 through Out11 layers output a mixed character region code consisting of English letters, numbers, and special Chinese characters, with each layer outputting a 6-bit binary code. The Out12 layer outputs a 1-bit binary code for the parity check bit. The Out13 layer outputs the class probabilities for five colors (blue - 0, green - 1, yellow - 2, white - 3, black - 4). Focal Loss is used to supervise the probabilities of each color during training.

[0110] Table 1: Motor vehicle license plate detection and recognition network parameters

[0111]

[0112] In Table 1, the Conv classification layer parameters (e.g., 24, 3x3, 2, 1) represent the number of output channels, kernel size, kernel stride, and padding of the convolution layer. The input size (e.g., 3x288x512) represents the number of channels, height, and width of the input feature map, respectively. The output size (e.g., 24x144x256) represents the number of channels, height, and width of the output feature map, respectively.

[0113] During the model training process, the nonlinear optimization method is combined with the back-propagation algorithm to minimize the loss function and achieve the purpose of converging the network model.

[0114] According to an embodiment of the present invention, the preset motor vehicle license plate detection and recognition model is optimized by using a loss function, specifically:

[0115] The loss function is expressed as follows:

[0116] ;

[0117] Among them, L represents the loss function, and is the loss function of the detection part, is the loss function of the binary coding part, is the loss function of the license plate color classification part; α, β, γ, and λ are all influence coefficients, and i is the number of binary codes.

[0118] It should be noted that because the convolutional neural network Alpha-LPDRNet consists of three distinct functional components, the network's loss function during training also consists of multiple components: the detection loss function, the binary encoding loss function, and the license plate color classification loss function. Furthermore, in this invention, the influence coefficients are α = 1, β = 10, γ = 10, and λ = 10, respectively.

[0119] According to an embodiment of the present invention, analyzing the detected image data to obtain license plate area image data includes:

[0120] Analyze the detection image data to determine whether there is a motor vehicle license plate;

[0121] If it exists, the motor vehicle license plate is located and the motor vehicle license plate area is intercepted to obtain license plate area image data; if it does not exist, filtering is performed.

[0122] It should be noted that after inputting the detection image data into the model, it first determines whether the vehicle license plate (license plate) is present in the detection image data. If so, the license plate area is located and captured within the feature map. The model then uses feature extraction and detection to locate the license plate in the image and generate the license plate area image data.

[0123] According to an embodiment of the present invention, decoding the predicted binary coded data to obtain license plate data includes:

[0124] Segmenting the predicted binary code data according to a preset template to obtain a plurality of sub-character code data;

[0125] Comparing the plurality of sub-character coded data with the standard data of the motor vehicle license plate binary code database respectively, decoding the plurality of sub-character coded data to obtain a plurality of sub-character data;

[0126] The plurality of sub-character data are integrated to obtain license plate data.

[0127] It should be noted that the system sets a preset template according to the motor vehicle license plate coding rules to facilitate the segmentation of binary coded data. The setting method of the preset template is specifically to segment the predicted binary coded data into multiple sub-character coded data with 6 characters as an interval, and divide the last character into a sub-character, namely the parity bit coded data.

[0128] Afterwards, the obtained multiple sub-character encoding data are decoded respectively according to the standard data in the motor vehicle license plate binary encoding database, and the decoded character data are integrated to obtain the license plate data.

[0129] Figure 4 A block diagram of a novel motor vehicle license plate detection and recognition system according to the present invention is shown.

[0130] like Figure 4 As shown, the second aspect of the present invention provides a novel motor vehicle license plate detection and recognition system 4, comprising a memory 41 and a processor 42. The memory includes a novel motor vehicle license plate detection and recognition method program, and when the novel motor vehicle license plate detection and recognition method program is executed by the processor, the following steps are implemented:

[0131] Acquire detection image data;

[0132] Inputting the detection image data into a preset motor vehicle license plate detection and recognition model;

[0133] The preset motor vehicle license plate detection and recognition model analyzes the detection image data to obtain license plate area image data;

[0134] The preset motor vehicle license plate detection and recognition model analyzes the license plate area image data to obtain predicted binary code data and license plate color data;

[0135] Verifying the predicted binary coded data by a preset method;

[0136] If the verification is successful, the predicted binary coded data is decoded to obtain the license plate data;

[0137] The license plate data and the license plate color data are analyzed to obtain the motor vehicle model data.

[0138] According to an embodiment of the present invention, a license plate binary code database is first established based on the license plate encoding rules. Then, based on the established license plate binary code database, a convolutional neural network (Alpha-LPDRNet) is designed to detect and recognize license plates through three aspects: first, locating the license plate in the image; second, outputting the license plate color data; and third, outputting the license plate binary code. The output license plate binary code is then decoded based on the license plate binary code database to obtain the license plate data. Combined with the license plate color data, the license plate is recognized and the vehicle model is determined. In addition, the present invention also adds a parity bit at the end of the character code to verify the prediction results output by the model, thereby improving the model recognition accuracy and reducing the error rate of license plate recognition.

[0139] According to an embodiment of the present invention, verifying the predicted binary coded data by a preset method includes:

[0140] Analyze the predicted binary coded data to obtain license plate binary coded data and parity bit coded data;

[0141] Performing parity statistics on the license plate binary coded data, and comparing the obtained parity statistics with the parity check bit coded data to determine whether the parity statistics and the parity check bit coded data are consistent;

[0142] If they are consistent, it means the prediction is correct and the license plate data is output;

[0143] If they are inconsistent, it means the prediction is wrong and the license plate data will not be output.

[0144] It should be noted that the binary-encoded data packet consists of two parts: the binary-encoded data of the number plate and the parity-encoded data. The parity-encoded data is obtained by counting the number of 1s in the binary number of each binary-encoded license plate. If the number of 1s is odd, the parity bit is 1; if the number of 1s is even, the parity bit is 0. During network training, this parity bit is represented by a 1x2 tensor. If the parity bit is 1, its label is (0, 1); if the parity bit is 0, its label is (1, 0). Cross entropy is used as the loss function. During inference, if the tensor outputs (1, 0), it indicates that the predicted parity-encoded data is 0; if the tensor outputs (0, 1), it indicates that the predicted parity-encoded data is 1. Then, a parity count is performed on the output license plate binary coded data. The number of 1s in the license plate binary coded data is counted. If the number of 1s is odd, the parity count is 1; if the number of 1s is even, the parity count is 0. The parity count is compared with the parity check bit coded data. If they are consistent, it means that the license plate prediction is correct and the prediction result is output; if they are inconsistent, it means that the license plate prediction is incorrect and the prediction result is not output.

[0145] According to an embodiment of the present invention, the further embodiment includes:

[0146] Obtain sample data of motor vehicle license plates;

[0147] Dividing the motor vehicle license plate data in the motor vehicle license plate sample data into a provincial administrative region abbreviation portion and a mixed portion of Arabic numerals, English letters, and special Chinese characters;

[0148] The provincial administrative region abbreviation part and the mixed part of Arabic numerals, English letters and special Chinese characters are binary-encoded respectively to establish a motor vehicle license plate binary coding database.

[0149] It should be noted that motor vehicle license plates generally consist of three parts: the abbreviation of the provincial administrative region, the code of the prefecture-level administrative region, and a random code of Arabic numerals and English letters.

[0150] According to an embodiment of the present invention, the further embodiment includes:

[0151] Obtain license plate sample image data;

[0152] Analyzing and training the license plate sample image data to establish a preset motor vehicle license plate detection and recognition model;

[0153] The preset motor vehicle license plate detection and recognition model is optimized through a loss function.

[0154] It should be noted that a license plate detection and recognition network requires three functions: first, locating the license plate in an image; second, outputting the license plate's color; and third, outputting the license plate's binary code. To achieve these functions, the present invention designed a convolutional neural network, Alpha-LPDRNet (Alpha-LicensePlate Detection Recognition Network), with specific parameters shown in Table 1.

[0155] The convolutional neural network Alpha-LPDRNet is mainly composed of convolutional layer (Convolutional Layer), normalization layer (Batch Normalization), non-linear activation layer (RELU), CSP module (cross stage Partial), fully connected layer (Fully Connected Layer) and other parts.

[0156] During the training process, Alpha-LPDRNet performs random contrast, brightness, and saturation adjustments on the original image, as well as image enhancement techniques such as small-angle rotation and random occlusion. After the Alpha-LPDRNet network detection portion completes positioning, the corresponding area does not require regional correction like traditional license plate recognition algorithms, nor does it require character segmentation of the feature map. Instead, each layer directly outputs the binary code, parity bit code, and license plate color of the corresponding position. This not only incorporates prior knowledge of license plate codes but also improves the execution efficiency of edge devices. While improving recognition accuracy, it also reduces user waiting time and enhances the user experience.

[0157] During the training process, it is necessary to mark the position and color of the motor vehicle license plate in the image and perform binary encoding on the motor vehicle license plate. The marked file and binary encoding are then used as supervision data for network training.

[0158] In Alpha-LPDRNet, Conv1 through Out3 layers are the feature extraction and detection components, locating the license plate in the image and acquiring high-dimensional features for the corresponding region. The Out4 layer outputs a 6-bit binary code representing the provincial administrative region. The Out5 through Out11 layers output a mixed character region code consisting of English letters, numbers, and special Chinese characters, with each layer outputting a 6-bit binary code. The Out12 layer outputs a 1-bit binary code for the parity check bit. The Out13 layer outputs the class probabilities for five colors (blue - 0, green - 1, yellow - 2, white - 3, black - 4). Focal Loss is used to supervise the probabilities of each color during training.

[0159] In Table 1, the Conv classification layer parameters (e.g., 24, 3x3, 2, 1) represent the number of output channels, kernel size, kernel stride, and padding of the convolution layer. The input size (e.g., 3x288x512) represents the number of channels, height, and width of the input feature map, respectively. The output size (e.g., 24x144x256) represents the number of channels, height, and width of the output feature map, respectively.

[0160] During the model training process, the nonlinear optimization method is combined with the back-propagation algorithm to minimize the loss function and achieve the purpose of converging the network model.

[0161] According to an embodiment of the present invention, the preset motor vehicle license plate detection and recognition model is optimized by using a loss function, specifically:

[0162] The loss function is expressed as follows:

[0163] ;

[0164] Among them, L represents the loss function, and is the loss function of the detection part, is the loss function of the binary coding part, is the loss function of the license plate color classification part; α, β, γ, and λ are all influence coefficients, and i is the number of binary codes.

[0165] It should be noted that because the convolutional neural network Alpha-LPDRNet consists of three distinct functional components, the network's loss function during training also consists of multiple components: the detection loss function, the binary encoding loss function, and the license plate color classification loss function. Furthermore, in this invention, the influence coefficients are α = 1, β = 10, γ = 10, and λ = 10, respectively.

[0166] According to an embodiment of the present invention, analyzing the detected image data to obtain license plate area image data includes:

[0167] Analyze the detection image data to determine whether there is a motor vehicle license plate;

[0168] If it exists, the motor vehicle license plate is located and the motor vehicle license plate area is intercepted to obtain license plate area image data; if it does not exist, filtering is performed.

[0169] It should be noted that after inputting the detection image data into the model, it first determines whether the vehicle license plate (license plate) is present in the detection image data. If so, the license plate area is located and captured within the feature map. The model then uses feature extraction and detection to locate the license plate in the image and generate the license plate area image data.

[0170] According to an embodiment of the present invention, decoding the predicted binary coded data to obtain license plate data includes:

[0171] Segmenting the predicted binary code data according to a preset template to obtain a plurality of sub-character code data;

[0172] Comparing the plurality of sub-character coded data with the standard data of the motor vehicle license plate binary code database respectively, decoding the plurality of sub-character coded data to obtain a plurality of sub-character data;

[0173] The plurality of sub-character data are integrated to obtain license plate data.

[0174] It should be noted that the system sets a preset template according to the motor vehicle license plate coding rules to facilitate the segmentation of binary coded data. The setting method of the preset template is specifically to segment the predicted binary coded data into multiple sub-character coded data with 6 characters as an interval, and divide the last character into a sub-character, namely the parity bit coded data.

[0175] Afterwards, the obtained multiple sub-character encoding data are decoded respectively according to the standard data in the motor vehicle license plate binary encoding database, and the decoded character data are integrated to obtain the license plate data.

[0176] The third aspect of the present invention provides a computer-readable storage medium, which includes a new motor vehicle license plate detection and recognition method program. When the new motor vehicle license plate detection and recognition method program is executed by a processor, the steps of the new motor vehicle license plate detection and recognition method as described in any one of the above items are implemented.

[0177] The present invention discloses a novel motor vehicle license plate detection and recognition method, system, and readable storage medium. The method comprises: acquiring detection image data; inputting the detection image data into a preset motor vehicle license plate detection and recognition model; analyzing the detection image data to obtain license plate area image data; analyzing the license plate area image data to obtain predicted binary code data and license plate color data; verifying the predicted binary code data using a preset method; if the verification is successful, decoding the predicted binary code data to obtain license plate data; and analyzing the license plate data and license plate color data to obtain motor vehicle model data. Based on deep learning-based image recognition technology, the present invention introduces a priori rules for motor vehicle coding and combines verification rules for data packet transmission in communication technology with motor vehicle license plate recognition, effectively reducing the error rate of motor vehicle license plate recognition.

[0178] In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are merely schematic. For example, the division of the units is merely a logical function division. In actual implementation, there may be other division methods, such as: multiple units or components can be combined, or can be integrated into another system, or some features can be ignored or not executed. In addition, the coupling, direct coupling, or communication connection between the components shown or discussed can be through some interfaces, and the indirect coupling or communication connection of the devices or units can be electrical, mechanical or other forms.

[0179] The units described above as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units; they may be located in one place or distributed across multiple network units; some or all of the units may be selected according to actual needs to achieve the purpose of the scheme of this embodiment.

[0180] In addition, all functional units in the embodiments of the present invention may be integrated into one processing unit, or each unit may be separately used as a unit, or two or more units may be integrated into one unit; the above-mentioned integrated units may be implemented in the form of hardware or in the form of hardware plus software functional units.

[0181] Those skilled in the art will appreciate that all or part of the steps of the above-mentioned method embodiments may be implemented by hardware associated with program instructions, and the aforementioned program may be stored in a computer-readable storage medium. When the program is executed, the program executes the steps of the above-mentioned method embodiments. The aforementioned storage medium includes various media that can store program codes, such as mobile storage devices, read-only memories (ROMs), random access memories (RAMs), magnetic disks, or optical disks.

[0182] Alternatively, if the integrated units described above are implemented as software modules and sold or used as standalone products, they can also be stored on a computer-readable storage medium. Based on this understanding, the technical solutions of the embodiments of the present invention, or the portion that contributes to the prior art, can be embodied in the form of a software product. This computer software product, stored on a storage medium, includes instructions for enabling a computer device (such as a personal computer, server, or network device) to execute all or part of the methods described in various embodiments of the present invention. The aforementioned storage media include various media capable of storing program code, such as removable storage devices, ROM, RAM, magnetic disks, or optical disks.

Claims

1. A new motor vehicle license plate detection and recognition method, characterized in that: include: Acquire detection image data; Inputting the detection image data into a preset motor vehicle license plate detection and recognition model; The preset motor vehicle license plate detection and recognition model analyzes the detection image data to obtain license plate area image data; The preset motor vehicle license plate detection and recognition model analyzes the license plate area image data to obtain predicted binary code data and license plate color data; Verifying the predicted binary coded data by a preset method; If the verification is successful, the predicted binary coded data is decoded to obtain the license plate data; Analyze the license plate data and the license plate color data to obtain the vehicle model data; The verifying the predicted binary coded data by a preset method includes: Analyze the predicted binary coded data to obtain license plate binary coded data and parity bit coded data; Performing parity statistics on the license plate binary coded data, and comparing the obtained parity statistics with the parity check bit coded data to determine whether the parity statistics and the parity check bit coded data are consistent; If they are consistent, it means the prediction is correct and the license plate data is output; If they are inconsistent, it means the prediction is wrong and the license plate data will not be output.

2. The novel motor vehicle license plate detection and recognition method according to claim 1, characterized in that: Also includes: Obtain sample data of motor vehicle license plates; Dividing the motor vehicle license plate data in the motor vehicle license plate sample data into a provincial administrative region abbreviation portion and a mixed portion of Arabic numerals, English letters, and special Chinese characters; The provincial administrative region abbreviation part and the mixed part of Arabic numerals, English letters and special Chinese characters are binary-encoded respectively to establish a motor vehicle license plate binary coding database.

3. The novel motor vehicle license plate detection and recognition method according to claim 1, characterized in that: Also includes: Obtain license plate sample image data; Analyzing and training the license plate sample image data to establish a preset motor vehicle license plate detection and recognition model; The preset motor vehicle license plate detection and recognition model is optimized through a loss function.

4. The novel motor vehicle license plate detection and recognition method according to claim 3, characterized in that: The preset motor vehicle license plate detection and recognition model is optimized by using a loss function, specifically: The loss function is expressed as follows: ； Among them, L represents the loss function, and is the loss function of the detection part, is the loss function of the binary coding part, is the loss function of the license plate color classification part; α, β, γ, and λ are all influence coefficients, and i is the number of binary codes.

5. The novel motor vehicle license plate detection and recognition method according to claim 1, characterized in that: The analyzing the detected image data to obtain license plate area image data includes: Analyze the detection image data to determine whether there is a motor vehicle license plate; If it exists, the motor vehicle license plate is located and the motor vehicle license plate area is intercepted to obtain license plate area image data; if it does not exist, filtering is performed.

6. The novel motor vehicle license plate detection and recognition method according to claim 1, characterized in that: The step of decoding the predicted binary coded data to obtain license plate data includes: Segmenting the predicted binary code data according to a preset template to obtain a plurality of sub-character code data; Comparing the plurality of sub-character coded data with the standard data of the motor vehicle license plate binary code database respectively, decoding the plurality of sub-character coded data to obtain a plurality of sub-character data; The plurality of sub-character data are integrated to obtain license plate data.

7. A new motor vehicle license plate detection and recognition system, characterized in that: The system comprises a memory and a processor, wherein the memory comprises a program for detecting and identifying a new motor vehicle license plate, and when the program is executed by the processor, the following steps are implemented: Acquire detection image data; Inputting the detection image data into a preset motor vehicle license plate detection and recognition model; The preset motor vehicle license plate detection and recognition model analyzes the detection image data to obtain license plate area image data; The preset motor vehicle license plate detection and recognition model analyzes the license plate area image data to obtain predicted binary code data and license plate color data; Verifying the predicted binary coded data by a preset method; If the verification is successful, the predicted binary coded data is decoded to obtain the license plate data; Analyze the license plate data and the license plate color data to obtain the vehicle model data; The verifying the predicted binary coded data by a preset method includes: Analyze the predicted binary coded data to obtain license plate binary coded data and parity bit coded data; Performing parity statistics on the license plate binary coded data, and comparing the obtained parity statistics with the parity check bit coded data to determine whether the parity statistics and the parity check bit coded data are consistent; If they are consistent, it means the prediction is correct and the license plate data is output; If they are inconsistent, it means the prediction is wrong and the license plate data will not be output.

8. A computer-readable storage medium, characterized in that The computer-readable storage medium includes a new motor vehicle license plate detection and recognition method program. When the new motor vehicle license plate detection and recognition method program is executed by a processor, the steps of the new motor vehicle license plate detection and recognition method as described in any one of claims 1 to 6 are implemented.

Images