Multi-dimensional detection system for tire cracks

Abstract

The invention provides a multi-dimensional detection system for tire cracks, and belongs to the technical field of crack detection. The parameter acquisition module is used for configuring an image system to acquire original image data of a target tire and configuring a sensor system to acquire original physical data of the target tire; the preprocessing module is used for preprocessing the original image data and the original physical data to obtain standard data of all parameter types of the target tire; the feature extraction module is used for obtaining parameter features by combining the standard data of any parameter type according to a feature algorithm; the risk module is used for carrying out feature fusion on the parameter features of all the parameter types to obtain multi-dimensional features, and inputting the multi-dimensional features into a crack risk model to obtain a crack risk score of the tire; and the verification module is used for verifying the accuracy of crack detection of the target tire and optimizing the crack risk model. And through multi-dimensional detection, the tire crack identification precision is improved, and the safety and reliability are ensured.

Description

Technical Field

[0001] The present invention relates to the technical field of crack detection, and in particular to a multi-dimensional detection system for tire cracks. Background Art

[0002] At present, tires are key components that connect vehicles to roads, and their safety directly affects driving safety. With the increase in traffic volume and the complexity of vehicle operating environments, tire wear and aging are becoming more serious, and cracks have become an important issue affecting tire performance. The occurrence of cracks not only reduces the service life of tires, but may also lead to serious accidents such as tire blowouts, endangering the safety of drivers and passengers. Usually, crack detection relies on manual inspection or simple visual inspection, which has disadvantages such as low efficiency, poor accuracy, and susceptibility to external environmental influences.

[0003] Therefore, the present invention proposes a multi-dimensional tire crack detection system. Summary of the invention

[0004] The present invention provides a multi-dimensional detection system for tire cracks, which is used to obtain the original image and physical data of the target tire by configuring the image and sensor modules. The preprocessing module converts the data into a standard format, and the feature extraction module extracts relevant parameters using a feature algorithm. The risk module fuses the multi-dimensional features and inputs them into a crack risk model to generate a risk score. Finally, the verification module evaluates the accuracy of crack detection and optimizes the crack risk model, thereby achieving efficient and accurate detection and risk assessment of tire cracks.

[0005] In one aspect, the present invention provides a multi-dimensional tire crack detection system, comprising: Parameter acquisition module: configure the image system to obtain the original image data of the target tire, configure the sensor system to obtain the original physical data of the target tire; Preprocessing module: preprocess the original image data and original physical data to obtain standard data of all parameter types of the target tire; Feature extraction module: parameter features obtained by combining standard data of any parameter type according to feature algorithms; Risk module: fusion of parameter features of all parameter types to obtain multi-dimensional features, and input them into the crack risk model to obtain the crack risk score of the tire; Verification module: Verifies the accuracy of crack detection of target tires and optimizes the crack risk model.

[0006] On the other hand, the parameter acquisition module includes: Sample preparation unit: obtains a target tire as a test sample, and obtains basic tire information according to the factory information of the target tire; Image system unit: select a high-resolution camera and install and configure it according to the basic information of the tire, initialize and start the high-resolution camera according to the preset angle and preset focal length, capture the image of the target tire, and obtain the original image data; Sensor unit: Select a sensor according to the preset tire parameter type, and obtain the original physical data of the target tire based on the sensor.

[0007] On the other hand, the preprocessing module comprises: Image data unit: for the original image data, firstly, the original image is denoised, and each pixel is gray-scaled to obtain a first gray-scale image; The first grayscale image is mapped to a standard coordinate system to obtain a second image, a Gaussian kernel of a preset size is selected, and a convolution operation is performed on the Gaussian kernel and each pixel of the second image, specifically: ;in Represents the standard point after convolution smoothing of the pixel point (x, y). represents the preset width of the Gaussian function, represents the Gaussian function, Indicates size The convolution function of the Gaussian kernel on the pixel point (x, y); The image formed by all standard points after the convolution operation is the standard image.

[0008] On the other hand, the preprocessing module further includes: Physical data unit: for the original physical data of any parameter type, check the abnormal values ​​and missing values ​​in the original physical data, use interpolation method to supplement, and obtain the first data; Standardizing the first data according to a preset physical quantity standard unit to obtain standard physical data; Standard data unit: Standard physical data and standard image together constitute standard data.

[0009] On the other hand, the feature extraction module comprises: Image feature extraction unit: for a standard image, all standard points of the standard image are obtained, and the differential Gaussian of any standard point is obtained as: ; represents the differential Gaussian of any standard point (x,y), represents the value weight coefficient, represents the directional weight coefficient, represents the Gaussian gradient of the standard point in the y direction, represents the Gaussian gradient of the standard point in the x direction, ( ) represents the direction function; Taking any standard point as the center point and eight surrounding standard points as comparison points, determining the extreme points of the nine points according to differential Gaussian, the extreme points are the edge points of the standard image, and all tire areas formed by all edge points are parameter features of the standard image; Physical feature extraction unit: for standard physical data, obtain standard physical data of any parameter type, process the standard physical data according to a statistical algorithm, and obtain statistical features of the standard physical data; the statistical features are parameter features of the standard physical data.

[0010] On the other hand, the risk module includes: Feature fusion unit: A neural network is constructed according to preset neural network weight parameters, and parameter features of standard physical data and parameter features of standard images are input into the neural network. For the parameter features of the standard image, the convolution layer of the neural network performs convolution processing on them to obtain the first feature, and for the parameter features of the standard physical data, the fully connected layer of the neural network performs connection fusion processing on them to obtain the second feature. Finally, the multi-dimensional features are: ;in, Represents the multi-dimensional features after the neural network, represents the activation function, represents the matching matrix of the first feature, represents the matching matrix of the second feature, Represents the first feature, Represents the second feature, Represents the bias constant.

[0011] On the other hand, the risk module includes: Risk indicator unit: creates risk indicators of crack presence based on standard characteristics of tire cracks; Risk scoring unit: taking the risk index as the input index of the crack risk model, obtaining the weight coefficient of any risk index according to the historical database, taking the weight coefficient as the model parameter, and constructing the crack risk model; The multi-dimensional features are input into the crack risk model, and the crack risk scores of different tire areas in the standard tire image are output.

[0012] On the other hand, the verification module includes: A crack judgment unit: based on the crack risk score of the tire area, obtains the tire area with a risk greater than a preset threshold, and marks the tire area as a crack existing area; Verification unit: According to the actual situation of the target tire, the depth-width ratio of the crack area and the actual area is determined to obtain the harmonic ratio; If the blending ratio is greater than a preset standard value, the determination result is inaccurate, and the model parameters of the crack risk model are subjected to reverse optimization processing; otherwise, the determination result is accurate.

[0013] Compared with the prior art, the present invention has the following beneficial effects: The present invention provides a multi-dimensional detection system for tire cracks, which is used to obtain the original image and physical data of the target tire by configuring the image and sensor modules. The preprocessing module converts the data into a standard format, and the feature extraction module extracts relevant parameters using a feature algorithm. The risk module fuses the multi-dimensional features and inputs them into a crack risk model to generate a risk score. Finally, the verification module evaluates the accuracy of crack detection and optimizes the crack risk model, thereby achieving efficient and accurate detection and risk assessment of tire cracks. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0015] Figure 1 It is a structural schematic diagram of a multi-dimensional tire crack detection system provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0016] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0017] Embodiment 1: like Figure 1 As shown, a multi-dimensional tire crack detection system provided by an embodiment of the present invention includes: Parameter acquisition module: configure the image system to obtain the original image data of the target tire, configure the sensor system to obtain the original physical data of the target tire; Preprocessing module: preprocess the original image data and original physical data to obtain standard data of all parameter types of the target tire; Feature extraction module: parameter features obtained by combining standard data of any parameter type according to feature algorithms; Risk module: fusion of parameter features of all parameter types to obtain multi-dimensional features, and input them into the crack risk model to obtain the crack risk score of the tire; Verification module: Verifies the accuracy of crack detection of target tires and optimizes the crack risk model.

[0018] In this embodiment, the image system is a hardware device used to obtain raw image data of the target tire, and generally includes a camera or other visual sensors.

[0019] In this embodiment, the target tire refers to the tire that is taken as the detection object in the entire detection and evaluation process.

[0020] In this embodiment, the original image data refers to the unprocessed tire surface image directly acquired by the imaging system.

[0021] In this embodiment, the sensor system is a set of devices used to collect raw physical data of the target tire, such as pressure, temperature, vibration, etc.

[0022] In this embodiment, the original physical data is basic data related to the physical state of the target tire collected by the sensor system, such as pressure, temperature, vibration, etc.

[0023] In this embodiment, parameter type refers to different types of data involved in the preprocessing and feature extraction process, such as image type, shape type, pressure type, temperature, etc.

[0024] In this embodiment, standard data refers to data that meets specific format, range, and accuracy requirements after processing and conversion of original data.

[0025] In this embodiment, the feature algorithm is an algorithm used to extract key information and features from raw data.

[0026] In this embodiment, the parameter feature refers to the key attribute information with identifiable features extracted from the original image data and the original physical data.

[0027] In this embodiment, feature fusion refers to the process of integrating feature information from different data sources or different types, using more comprehensive information to perform a more accurate analysis of tire cracks.

[0028] In this embodiment, the multi-dimensional feature refers to a feature containing multiple dimensions formed by fusing data from different types.

[0029] In this embodiment, the crack risk model is a prediction model for evaluating the crack risk of a tire, and the crack risk of the tire is determined by analyzing various parameters of the tire.

[0030] In this embodiment, the crack risk score is a prediction model for evaluating the crack risk of a tire, and the crack risk of the tire is determined by analyzing various parameters of the tire.

[0031] In this embodiment, accuracy refers to the correctness of crack detection, including: prediction accuracy, recall rate, etc.

[0032] The working principle and beneficial effects of the above technical solution are: through multi-dimensional data collection, standardized processing, feature extraction and fusion, combined with the crack risk model, accurate detection and risk assessment of tire cracks can be achieved. The verification module optimizes the model, improves detection accuracy, and ensures tire safety and reliability.

[0033] Embodiment 2: Based on the above embodiment 1, the parameter acquisition module includes: Sample preparation unit: obtains a target tire as a test sample, and obtains basic tire information according to the factory information of the target tire; Image system unit: select a high-resolution camera and install and configure it according to the basic information of the tire, initialize and start the high-resolution camera according to the preset angle and preset focal length, capture the image of the target tire, and obtain the original image data; Sensor unit: Select a sensor according to the preset tire parameter type, and obtain the original physical data of the target tire based on the sensor.

[0034] In this embodiment, the test sample refers to a target tire used to obtain relevant data for analysis and evaluation during the tire crack detection process.

[0035] In this embodiment, the factory information refers to basic and technical data about the target tire recorded during the production process.

[0036] In this embodiment, the basic tire information refers to key parameters and technical specifications related to the target tire, such as tire size, load factor, tread pattern, etc.

[0037] In this embodiment, the high-resolution camera refers to a camera that can capture high-detail and high-precision images.

[0038] In this embodiment, the preset angle refers to a shooting angle that is pre-set by the image system unit according to specific conditions (such as type, size, detection purpose, etc.) of the target tire when capturing images.

[0039] In this embodiment, the preset focal length refers to a fixed lens focal length set in the image system, which is used to control the distance relationship between the camera lens and the target object.

[0040] The working principle and beneficial effects of the above technical solution are: by obtaining target tire samples and selecting appropriate images and sensor systems, the tire images and physical data can be accurately captured. High-resolution cameras and sensors are used to ensure high data quality, provide a reliable basis for subsequent crack detection and risk assessment, and improve detection accuracy.

[0041] Embodiment 3: Based on the above embodiment 1, the preprocessing module includes: Image data unit: for the original image data, firstly, the original image is denoised, and each pixel is gray-scaled to obtain a first gray-scale image; The first grayscale image is mapped to a standard coordinate system to obtain a second image, a Gaussian kernel of a preset size is selected, and a convolution operation is performed on the Gaussian kernel and each pixel of the second image, specifically: ;in Represents the standard point after convolution smoothing of the pixel point (x, y). represents the preset width of the Gaussian function, represents the Gaussian function, Indicates size The convolution function of the Gaussian kernel on the pixel point (x, y); The image formed by all standard points after the convolution operation is the standard image.

[0042] In this embodiment, grayscale processing is a process of converting a color image into a grayscale image.

[0043] In this embodiment, the first grayscale image is the result of grayscale processing on the original image.

[0044] In this embodiment, the standard coordinate system refers to a coordinate system used to describe the position of each pixel in an image.

[0045] In this embodiment, the second image is an image obtained by mapping the first grayscale image to a standard coordinate system.

[0046] In this embodiment, the Gaussian kernel is a convolution kernel commonly used in image processing and is generated based on a Gaussian function.

[0047] In this embodiment, the convolution operation is a common mathematical operation, which operates the convolution kernel with each pixel of the image to generate a new image.

[0048] In this embodiment, the standard point refers to the smoothed value of each pixel in the image obtained after the convolution operation.

[0049] The working principle and beneficial effects of the above technical solution are: improving the quality of the original image through denoising, grayscale processing, standard coordinate mapping and Gaussian convolution. The convolution operation smoothes the image details, reduces noise, generates a standard image, and provides clearer and more accurate image data for subsequent feature extraction and crack detection.

[0050] Embodiment 4: Based on the above embodiment 1, the preprocessing module further includes: Physical data unit: for the original physical data of any parameter type, check the abnormal values ​​and missing values ​​in the original physical data, use interpolation method to supplement, and obtain the first data; Standardizing the first data according to a preset physical quantity standard unit to obtain standard physical data; Standard data unit: Standard physical data and standard image together constitute standard data.

[0051] In this embodiment, an outlier refers to data whose value significantly deviates from the normal range compared with other data points.

[0052] In this embodiment, missing values ​​refer to the lack of values ​​for certain data points in the data set.

[0053] In this embodiment, interpolation is a mathematical method used to estimate or fill in missing data points, especially to infer missing or abnormal data points in the middle when a portion of the data is known.

[0054] In this embodiment, the first data refers to physical data obtained after being processed by the interpolation method.

[0055] In this embodiment, the preset physical quantity standard units refer to a set of standardized units used when processing physical data to ensure that the data conforms to the unit standard requirements widely accepted internationally.

[0056] In this embodiment, the standard physical data is processed and standardized physical data that conforms to a unified preset standard unit.

[0057] In this embodiment, the standardization process refers to converting the data into a unified format.

[0058] The working principle and beneficial effects of the above technical solution are: the original physical data is supplemented by outlier detection and interpolation, and the standard physical data is obtained after standardization. Combined with the standard image, complete standard data is formed to ensure the accuracy and consistency of the data, providing a reliable basis for subsequent analysis and detection.

[0059] Embodiment 5: Based on the above embodiment 1, the feature extraction module includes: Image feature extraction unit: for a standard image, all standard points of the standard image are obtained, and the differential Gaussian of any standard point is obtained as: ; represents the differential Gaussian of any standard point (x,y), represents the value weight coefficient, represents the directional weight coefficient, represents the Gaussian gradient of the standard point in the y direction, represents the Gaussian gradient of the standard point in the x direction, ( ) represents the direction function; Taking any standard point as the center point and eight surrounding standard points as comparison points, determining the extreme points of the nine points according to differential Gaussian, the extreme points are the edge points of the standard image, and all tire areas formed by all edge points are parameter features of the standard image; Physical feature extraction unit: for standard physical data, obtain standard physical data of any parameter type, process the standard physical data according to a statistical algorithm, and obtain statistical features of the standard physical data; the statistical features are parameter features of the standard physical data.

[0060] In this embodiment, the differential Gaussian approximates the Laplacian operator of the image by using the difference between two Gaussian functions with different standard deviations, thereby extracting edge information in the image.

[0061] In this embodiment, Gaussian gradient refers to calculating the gradient of each pixel in the image after applying Gaussian filtering to the image.

[0062] In this embodiment, the direction function refers to a mathematical function that describes the directionality of an image feature point.

[0063] In this embodiment, the comparison points refer to the points that are compared with the center point, and these points are used in the algorithm to help determine the strength of the feature.

[0064] In this embodiment, the extreme point refers to a point with the maximum differential Gaussian in a certain local area.

[0065] In this embodiment, the tire region refers to a region surrounded by points with significant changes extracted from the image by the difference Gaussian method.

[0066] In this embodiment, the statistical algorithm refers to a type of algorithm used to process and analyze data.

[0067] In this embodiment, the statistical features include mean, standard deviation, variance, etc.

[0068] The working principle and beneficial effects of the above technical solution are: extracting the edge points of the standard image through differential Gaussian and extreme point detection to form image parameter features. Combined with statistical algorithms to process standard physical data, extract parameter features, ensure accurate extraction of image and physical data features, and provide efficient and reliable data support for subsequent analysis.

[0069] Embodiment 6: Based on the above embodiment 5, the risk module includes: Feature fusion unit: A neural network is constructed according to preset neural network weight parameters, and parameter features of standard physical data and parameter features of standard images are input into the neural network. For the parameter features of the standard image, the convolution layer of the neural network performs convolution processing on them to obtain the first feature, and for the parameter features of the standard physical data, the fully connected layer of the neural network performs connection fusion processing on them to obtain the second feature. Finally, the multi-dimensional features are: ;in, Represents the multi-dimensional features after the neural network, represents the activation function, represents the matching matrix of the first feature, represents the matching matrix of the second feature, Represents the first feature, Represents the second feature, Represents the bias constant.

[0070] In this embodiment, the preset number of neural network weight parameters refers to the weights and bias values ​​in all layers (such as convolutional layers, fully connected layers, etc.) when the neural network model is constructed.

[0071] In this embodiment, the neural network is a computational model that simulates information processing in a biological nervous system and processes input data through hierarchical connections of neurons (also called nodes).

[0072] In this embodiment, convolutional layers are used to automatically extract features from input data (usually images).

[0073] In this embodiment, the first feature is a result obtained by convolutionally processing the parameter feature of the standard image through the convolution layer of the neural network.

[0074] In this embodiment, the fully connected layer achieves feature fusion and conversion by connecting all neurons in the previous layer with each neuron in the current layer.

[0075] In this embodiment, the second feature refers to a feature vector obtained by processing the parameter features of the standard physical data through the fully connected layer of the neural network.

[0076] In this embodiment, the activation function is a function used to introduce nonlinear transformation in the neural network.

[0077] In this embodiment, the matching matrix refers to a matrix used in a neural network to combine different types of features (such as image features and physical data features).

[0078] The working principle and beneficial effects of the above technical solution are: the parameter features of the standard image and standard physical data are fused through a neural network, the convolution layer extracts the image features, and the fully connected layer fuses the physical data features, and finally multi-dimensional features are obtained, which improves the comprehensive expressiveness of the data and provides accurate information support for subsequent decision-making and analysis.

[0079] Embodiment 7: Based on the above embodiment 6, the risk module includes: Risk indicator unit: creates risk indicators of crack presence based on standard characteristics of tire cracks; Risk scoring unit: taking the risk index as the input index of the crack risk model, obtaining the weight coefficient of any risk index according to the historical database, taking the weight coefficient as the model parameter, and constructing the crack risk model; The multi-dimensional features are input into the crack risk model, and the crack risk scores of different tire areas in the standard tire image are output.

[0080] In this embodiment, the standard features refer to key features used to indicate tire cracks.

[0081] In this embodiment, the risk index is a quantitative standard used to quantify the crack risk of different areas of the tire during the crack risk assessment process, including: crack length, crack width, number of cracks, tire pressure, etc.

[0082] In this embodiment, the input indicators are those standard features that are directly used to construct the crack risk score.

[0083] In this embodiment, the history library refers to a database containing past tire data, which records the occurrence of cracks in different tires during actual use.

[0084] In this embodiment, the model parameters refer to values ​​that need to be determined based on historical data and analysis results when constructing the crack risk model.

[0085] In this embodiment, the crack risk score is a numerical indicator used to evaluate the risk of cracks occurring in a tire under specific conditions.

[0086] The working principle and beneficial effects of the above technical solution are: by creating crack risk indicators and combining the weight coefficients of the historical library, a crack risk model is constructed, multi-dimensional features are input into the model, and crack risk scores for different tire areas are output, so as to effectively evaluate tire crack risks and provide accurate basis for maintenance decisions.

[0087] Embodiment 8: Based on the above embodiment 7, the verification module includes: A crack judgment unit: based on the crack risk score of the tire area, obtains the tire area with a risk greater than a preset threshold, and marks the tire area as a crack existing area; Verification unit: According to the actual situation of the target tire, the depth-width ratio of the crack area and the actual area is determined to obtain the harmonic ratio; If the blending ratio is greater than a preset standard value, the determination result is inaccurate, and the model parameters of the crack risk model are subjected to reverse optimization processing; otherwise, the determination result is accurate.

[0088] In this embodiment, the crack existing area refers to a tire area identified as having possible cracks after analysis based on the tire crack risk scoring model.

[0089] In this embodiment, the harmonic ratio is a ratio used to measure the depth-to-width ratio of the crack existing area to the actual crack area.

[0090] In this embodiment, the preset standard value is a threshold value used to evaluate the prediction accuracy of the model.

[0091] The working principle and beneficial effects of the above technical solution are: by verifying the accuracy of tire crack detection, combining the depth-width ratio to adjust the risk model, and optimizing the model parameters if the blending ratio exceeds the standard, the accuracy of crack risk assessment is ensured, the detection accuracy is improved, and a more reliable decision-making basis is provided for tire maintenance.

[0092] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A multi-dimensional tire crack detection system, characterized in that: include: Parameter acquisition module: configure the image system to obtain the original image data of the target tire, configure the sensor system to obtain the original physical data of the target tire; Preprocessing module: preprocessing the original image data and original physical data to obtain standard data of all parameter types of the target tire; Feature extraction module: parameter features obtained by combining standard data of any parameter type according to feature algorithms; Risk module: fusion of parameter features of all parameter types to obtain multi-dimensional features, and input them into the crack risk model to obtain the crack risk score of the tire; Verification module: Verifies the accuracy of crack detection of target tires and optimizes the crack risk model.

2. The multi-dimensional tire crack detection system according to claim 1, characterized in that: The parameter acquisition module includes: Sample preparation unit: obtains a target tire as a test sample, and obtains basic tire information according to the factory information of the target tire; Image system unit: select a high-resolution camera and install and configure it according to the basic information of the tire, initialize and start the high-resolution camera according to the preset angle and preset focal length, capture the image of the target tire, and obtain the original image data; Sensor unit: Select a sensor according to the preset tire parameter type, and obtain the original physical data of the target tire based on the sensor.

3. The multi-dimensional tire crack detection system according to claim 1, characterized in that: The preprocessing module comprises: Image data unit: for the original image data, firstly, the original image is denoised, and each pixel is gray-scaled to obtain a first gray-scale image; The first grayscale image is mapped to a standard coordinate system to obtain a second image, a Gaussian kernel of a preset size is selected, and a convolution operation is performed on the Gaussian kernel and each pixel of the second image, specifically: ;in Represents the standard point after convolution smoothing of the pixel point (x, y). represents the preset width of the Gaussian function, represents the Gaussian function, Indicates size The convolution function of the Gaussian kernel on the pixel point (x, y); The image formed by all standard points after the convolution operation is the standard image.

4. The multi-dimensional tire crack detection system according to claim 1, characterized in that: The preprocessing module further includes: Physical data unit: for the original physical data of any parameter type, check the abnormal values ​​and missing values ​​in the original physical data, use interpolation method to supplement, and obtain the first data; Standardizing the first data according to a preset physical quantity standard unit to obtain standard physical data; Standard data unit: Standard physical data and standard image together constitute standard data.

5. The multi-dimensional tire crack detection system according to claim 1, characterized in that: The feature extraction module comprises: Image feature extraction unit: for a standard image, all standard points of the standard image are obtained, and the differential Gaussian of any standard point is obtained as: ; represents the differential Gaussian of any standard point (x,y), represents the value weight coefficient, represents the directional weight coefficient, represents the Gaussian gradient of the standard point in the y direction, represents the Gaussian gradient of the standard point in the x direction, ( ) represents the direction function; Taking any standard point as the center point and eight surrounding standard points as comparison points, determining the extreme points of the nine points according to differential Gaussian, the extreme points are the edge points of the standard image, and all tire areas formed by all edge points are parameter features of the standard image; Physical feature extraction unit: for standard physical data, obtain standard physical data of any parameter type, process the standard physical data according to a statistical algorithm, and obtain statistical features of the standard physical data; the statistical features are parameter features of the standard physical data.

6. The multi-dimensional tire crack detection system according to claim 5, characterized in that: The risk module includes: Feature fusion unit: A neural network is constructed according to preset neural network weight parameters, and parameter features of standard physical data and parameter features of standard images are input into the neural network. For the parameter features of the standard image, the convolution layer of the neural network performs convolution processing on them to obtain the first feature, and for the parameter features of the standard physical data, the fully connected layer of the neural network performs connection fusion processing on them to obtain the second feature. Finally, the multi-dimensional features are: ;in, Represents the multi-dimensional features after the neural network, represents the activation function, represents the matching matrix of the first feature, represents the matching matrix of the second feature, Represents the first feature, Represents the second feature, Represents the bias constant.

7. The multi-dimensional tire crack detection system according to claim 6, characterized in that: The risk module includes: Risk indicator unit: creates risk indicators of crack presence based on standard characteristics of tire cracks; Risk scoring unit: taking the risk index as the input index of the crack risk model, obtaining the weight coefficient of any risk index according to the historical database, taking the weight coefficient as the model parameter, and constructing the crack risk model; The multi-dimensional features are input into the crack risk model, and the crack risk scores of different tire areas in the standard tire image are output.

8. The multi-dimensional tire crack detection system according to claim 7, characterized in that: The verification module comprises: A crack judgment unit: based on the crack risk score of the tire area, obtains the tire area with a risk greater than a preset threshold, and marks the tire area as a crack existing area; Verification unit: According to the actual situation of the target tire, the depth-width ratio of the crack area and the actual area is determined to obtain the harmonic ratio; If the blending ratio is greater than a preset standard value, the determination result is inaccurate, and the model parameters of the crack risk model are subjected to inverse optimization processing; otherwise, the determination result is accurate.

Images