Vehicle defect comprehensive detection method and system based on part collaboration

Abstract

The present invention provides a vehicle defect comprehensive detection method and system based on part coordination, which belongs to the detection field and solves the problem of low vehicle detection efficiency. The specific method is as follows: Step S1: Acquire a side image, driving record and driver age of the vehicle; Step S2: Mark and extract the side image to obtain a defect image set; Calculate the weighted defect depth of different surfaces of the vehicle according to the defect image set; Step S3: Acquire a defect data set; Construct relationship equations 1 and 2 between defects and road sections and speeds according to the driving record and the defect data set; Analyze and construct relationship equation 3 between defects and driver age according to the driver age and the defect data set; Summarize relationship equations 1 to 3 as probability occurrence equations of vehicle defects; Step S4: Update the probability occurrence equation. The present invention detects defects and improves the efficiency of defect detection by acquiring, analyzing and processing relevant information of defective vehicles.

Description

Technical Field

[0001] The present invention discloses a vehicle defect comprehensive detection method and system based on part coordination, and relates to the detection field. Background Art

[0002] The existing vehicle defect detection methods and systems have the following deficiencies:

[0003] Technical complexity and high cost: Existing detection systems integrate multiple technologies such as sensors, image processing, and artificial intelligence. The research and development and application of these technologies require a lot of capital and human resources, and the construction cost is relatively high.

[0004] Dependency and stability issues: The existing system is highly dependent on various sensors and data processing equipment. Once these devices fail or their performance degrades, the stability and reliability of the entire detection system will be affected.

[0005] Limitations of detection range: The existing detection system has high detection accuracy and efficiency, but it is still affected by factors such as vehicle type, brand, and model, resulting in certain limitations on the accuracy and reliability of the detection results. Summary of the invention

[0006] In view of the shortcomings of the prior art, the purpose of the present invention is to provide a comprehensive vehicle defect detection method and system based on part collaboration, aiming to solve the problem of low vehicle detection efficiency.

[0007] In order to achieve the above object, the present invention is implemented through the following technical solutions: A vehicle defect comprehensive detection method based on part coordination includes:

[0008] Step S1: Obtain the side image of the vehicle, driving record and driver age;

[0009] Step S2: obtaining a test picture set, and constructing a vehicle defect recognition model based on the test picture set; using the vehicle defect recognition model, marking and extracting pictures of defective areas in the side pictures to obtain a defect picture set; and calculating weighted defect depths of different surfaces of the vehicle based on the defect picture set;

[0010] Step S3: Obtain side images, driving records and driver ages corresponding to all vehicles in the target area to obtain mixed information; use the vehicle defect recognition model to extract and count the total number of defects on different surfaces and weighted defect depths on different surfaces corresponding to all vehicles in the mixed information to obtain a defect data set;

[0011] According to the driving records and defect data set, the relationship equations 1 and 2 between defects, road sections and speed are analyzed and constructed;

[0012] Based on the driver age and defect data set, analyze and construct the relationship equation between defect and driver age3;

[0013] Summarize the relationship equations 1 to 3 as the probability occurrence equation of vehicle defects;

[0014] Step S4: Continuously obtain the side images, driving records and driver ages of all vehicles in the target area, and update the probability occurrence equation.

[0015] Furthermore, the specific steps of step S2 are as follows:

[0016] Step S21: obtaining a test picture set, and constructing a vehicle defect recognition model according to the test picture set;

[0017] Step S22: Substituting the side image of the vehicle into the vehicle defect recognition model, marking and extracting the images of the areas where defects occur on the A side, B side, C side and D side of the vehicle in sequence, and obtaining a defect image set;

[0018] Count the number of area images on the vehicle's A side, denoted as pa;

[0019] Count the number of area images on the vehicle's B side, denoted as pb;

[0020] Count the number of area images on the C side of the vehicle, denoted as pc;

[0021] Count the number of area images of vehicle surface D, recorded as pd;

[0022] pn=pa+pb+pc+pd; where pn represents the number of images in the defect image set;

[0023] Step S23: binarize the images in the defect image set to obtain the grayscale value of each defect image;

[0024] Step S24: repeat the same steps of binarizing the first regional image, and binarize the second to pn-th images to obtain the to-be-analyzed images 2 to pn;

[0025] Step S25: defining calculation formulas 1 to 6 for calculating the defect depth according to the grayscale of the defect image and the defect width;

[0026] Step S26: Obtain the grayscale value of each image to be analyzed and the width of each defect, and calculate the depth of each defect on each surface of the vehicle according to the calculation formulas 1 to 6 defined in step S25 to obtain:

[0027] The depth of each defect on the vehicle surface A is denoted as h(A) 1 、h(A) 2 ~h(A) pa ;

[0028] Where h(A) 1 Indicates the depth of the first defect on the A side of the vehicle; h (A) 2 Indicates the depth of the second defect on the A side of the vehicle; and so on, h(A) pa Indicates the depth of the pa-th defect on the vehicle surface A;

[0029] The depth of each defect on the B side of the vehicle, denoted as h(B) 1 、h(B) 2 ~h(B) pb ;

[0030] Among them, h(B) 1 Indicates the depth of the first defect on the B side of the vehicle; h(B) 2 Indicates the depth of the second defect on the B side of the vehicle; and so on, h(B) pb Indicates the depth of the pbth defect on the B side of the vehicle;

[0031] The depth of each defect on the C surface of the vehicle, denoted as h(C) 1 、h(C) 2 ~h(C) pc ;

[0032] Among them, h (C) 1 Indicates the depth of the first defect on the C surface of the vehicle; h(C) 2 Indicates the depth of the second defect on the C surface of the vehicle; and so on, h(C) pc Indicates the depth of the pcth defect on the C surface of the vehicle;

[0033] The depth of each defect on the vehicle surface D is denoted as h(D) 1 、h(D) 2 ~h(D) pd ;

[0034] Among them, h (D) 1 Indicates the depth of the first defect on the A side of the vehicle; h (D) 2 Indicates the depth of the second defect on the A side of the vehicle; and so on, h(D) pd Indicates the depth of the pdth defect on the D surface of the vehicle;

[0035] Step S27: Extract the maximum value, minimum value and mode of the defect depth of each surface of the vehicle to obtain:

[0036] For the A side of the vehicle, the maximum defect depth is denoted as ha(A), the minimum defect depth is denoted as hb(A), and the mode defect depth is denoted as hc(A);

[0037] For vehicle side B, the maximum defect depth is denoted as ha(B), the minimum defect depth is denoted as hb(B), and the mode defect depth is denoted as hc(B);

[0038] For the C side of the vehicle, the maximum defect depth is recorded as ha(C), the minimum defect depth is recorded as hb(C), and the mode defect depth is recorded as hc(C);

[0039] For the vehicle D surface, the maximum defect depth is denoted as ha(D), the minimum defect depth is denoted as hb(D), and the mode defect depth is denoted as hc(D);

[0040] Step S28: Calculate the weighted defect depth of each surface of the vehicle and proceed to step S3.

[0041] Furthermore, the specific steps of step S21 are as follows:

[0042] Step S211: using the torchvision.models module in the PyTorch library, importing the AlexNet model, the SqueezeNet model, and the GoogLeNet model as the basic models of the vehicle defect recognition model;

[0043] The activation function selects the softmax function;

[0044] Step S212: define a BN operation; constrain the underlying eigenvalues ​​of the basic model to a distribution with a mean equal to 0 and a variance equal to 1;

[0045] For any input B{x 1 、x 2 ……x m The process of performing BN operation is as follows:

[0046] Calculate the mean μB of the elements in set B, ;

[0047] Calculate the variance σB of the elements in set B 2 , ; Standardize the elements in set B and get ;

[0048] Calculate x 1 ~x m get of The calculation formula is as follows:

[0049] ; Wherein, input B refers to the image in the test image set; i represents 1 to m, and m represents the pixel value of the image in input B; x i represents the gray value corresponding to the i-th pixel in image B, and m is the maximum number of pixels in image B;

[0050] Step S213: creating two pooling layers, Pool 1 and Pool 2, for the vehicle defect recognition model;

[0051] Create 5 convolutional layers, starting with the convolutional layer Con1 and the subsequent convolutional layers Conv2 to Conv5;

[0052] Create a connection layer Concat; create two fully connected layers Fc6~Fc8;

[0053] Step S214: upload the image of the test picture set as a primary input to the Con1 layer and the Pool1 layer, perform bottom-level feature extraction, and obtain extracted feature 1;

[0054] Take the extracted feature 1 as the secondary input and upload it to the Conv2 layer and the Pool2 layer to obtain the extracted feature 2;

[0055] Backup extracted feature 2 as the underlying feature;

[0056] Step S125: repeat step S24, upload the extracted feature 2 to the Conv3 layer for processing, and obtain the extracted feature 3; upload the extracted feature 3 to the Conv4 layer for processing, and obtain the extracted feature 4; upload the extracted feature 4 to the Conv5 layer for processing, and obtain the extracted feature 5 as the high-level feature;

[0057] Step S216: perform BN operation on the bottom-level features and high-level features respectively to obtain fused features; use the fused features as the input of the fully connected layer fc6;

[0058] The cross entropy loss function is used as the loss function of the vehicle defect recognition model;

[0059] Step S217: defining the output of the vehicle defect recognition model;

[0060] If the side image is associated with the test image set, the output is defective; if the side image is not associated with the test image set, no output is performed.

[0061] Furthermore, the specific steps of step S23 are as follows:

[0062] Step S231: record the first region image as image p, and perform binarization processing on image p;

[0063] Construct a histogram of the image p with the pixel brightness value as the horizontal coordinate and the frequency of the pixel brightness value in the image as the vertical coordinate;

[0064] Step S232: import the OpenCV library; use the cv2.equalizeHist histogram equalization function in the OpenCV library to perform image enhancement processing on the image p to obtain the image p1;

[0065] Step S233: Use the cv2.adaptiveThreshold() function in the OpenCV library to process the image A2 and obtain an adaptive threshold;

[0066] Use the cv2.threshold() function in the OpenCV library to binarize the image p2 according to the adaptive threshold to obtain the image 1 to be analyzed;

[0067] Step S24: Repeat the same steps of binarizing the first regional image, and binarize the second to pn-th images to obtain the image 2 to the image pn to be analyzed.

[0068] Furthermore, the specific steps of step S25 are as follows:

[0069] Step S251: set the depth of the depression to h and the width of the depression to l;

[0070] The angle of incidence of the incident light ray is angle θ;

[0071] Calculation formula 1: Angle θ = arccos[(2×h) / l];

[0072] Step S252: setting the complementary angle of the incident angle to angle a;

[0073] Calculation 2: Angle a = 90 o - angle θ;

[0074] Let the angle subtended by the reflected ray be angle b;

[0075] Calculation 3: Angle b = 180 o -2×angle θ;

[0076] Let the supplement of (angle a + angle b) be angle c;

[0077] Calculation 4: Angle c = 180 o -2×(angle a+angle b);

[0078] Step S253: Set the complementary angle of angle c to angle d;

[0079] Calculation 5: Angle d = 90 o - Angle c;

[0080] Let the deviation angle of the outgoing light relative to the incident light be angle e;

[0081] From the triangle law and the definition of parallel lines, we can know that formula 6: angle d + angle e = angle θ;

[0082] Step S254: Set the gray value of the incident light to HHi; Set the gray value of the outgoing light to HHo;

[0083] Calculate the attenuation coefficient of the gray value, denoted as do; do = HHo / HHi;

[0084] Calculate the angle of angle e, angle e = arccos(do);

[0085] Substitute the angle e and the concave width into the calculation formula 6-1 in reverse order, and use the angle change relationship and trigonometric functions to obtain the concave depth.

[0086] Furthermore, the specific steps of step S28 are as follows:

[0087] Step S281: Calculate the weighted defect depth of vehicle surface A;

[0088] Definition relation 21: lim(h(A) i / ha(A))→1;

[0089] Define relation 22: lim(h(A) i / hb(A))→1;

[0090] Definition relation 23: lim(h(A) i / hc(A))→1;

[0091] Where i is a positive number, and its value range is 1~pa; h(A) i represents the depth of the i-th defect in the vehicle surface A;

[0092] Step S282: h(A) 1 ~h(A) pa ; Substitute into equations 21 to 23, and record the number of equations that satisfy equation 21 as a1, the number of equations that satisfy equation 22 as a2, and the number of equations that satisfy equation 23 as a3;

[0093] Calculate h(A) 1 ~h(A) pa The average value of is denoted as ah (A);

[0094] Step S283: Calculate the weighted defect depth of the vehicle surface A, recorded as Qh(A); the calculation formula of Qh(A) is as follows:

[0095] ;

[0096] Step S284: Repeat the same steps of calculating the weighted defect depth of vehicle surface A to calculate the weighted defect depths of vehicle surfaces B, C and D.

[0097] Furthermore, the specific steps of step S3 are as follows:

[0098] Step S31: Obtain side images, driving records and driver ages of all defective vehicles in the target area to obtain mixed information;

[0099] Driving records include: driving section and driving speed;

[0100] Step S32: Obtain the vehicle flow in the target area, recorded as ZT; count the total number of vehicles with defects, recorded as QT;

[0101] Substitute the side image in the mixed information into the vehicle defect recognition model, extract and count the total number of defects on different surfaces and the weighted defect depths on different surfaces corresponding to all vehicles in the mixed information, and obtain a defect data set;

[0102] The total number of defects corresponding to the vehicle A surface in the defect data set is recorded as qa, and the weighted defect depth corresponding to each defect is recorded as haa 1 、haa 2 ~haa qa Among them, haa 1 Indicates the weighted defect depth of the first defect on the A side of the vehicle in the defect dataset; haa 2 represents the weighted defect depth of the second defect on the A side of the vehicle in the defect data set; and so on, haa qa represents the weighted defect depth of the qath defect on the vehicle A side in the defect dataset;

[0103] The total number of defects corresponding to the vehicle B surface in the defect data set is recorded as qb, and the weighted defect depth corresponding to each defect is recorded as hbb 1 、hbb 2 ~hbb qb ; Among them, hbb 1 represents the weighted defect depth of the first defect on the B side of the vehicle in the defect dataset; hbb 2 represents the weighted defect depth of the second defect on the B side of the vehicle in the defect dataset; and so on, hbb qb represents the weighted defect depth of the qbth defect on the B side of the vehicle in the defect dataset;

[0104] The total number of defects corresponding to the C surface of the vehicle in the defect data set is recorded as qc, and the weighted defect depth corresponding to each defect is recorded as hcc 1 、hcc 2 ~hcc qc ; Among them, hcc 1 Indicates the weighted defect depth of the first defect on the C surface of the vehicle in the defect dataset; hcc 2 represents the weighted defect depth of the second defect on the C surface of the vehicle in the defect dataset; and so on, hcc qc represents the weighted defect depth of the qc-th defect on the C surface of the vehicle in the defect dataset;

[0105] The total number of defects corresponding to the vehicle D surface in the defect data set is recorded as qd, and the weighted defect depth corresponding to each defect is recorded as hdd 1 、hdd 2 ~hdd qd ; Among them, hdd 1 Indicates the weighted defect depth of the first defect on the D surface of the vehicle in the defect dataset; hdd 2 Indicates the weighted defect depth of the second defect on the D surface of the vehicle in the defect dataset; and so on, hdd qd represents the weighted defect depth of the qd-th defect on the D surface of the vehicle in the defect dataset;

[0106] Step S33: Analyze and construct relation equations 1 and 2 between defects, road sections and speeds according to the driving records and defect data set;

[0107] Step S34: taking the driver's age as the independent variable and the total number of defects on different surfaces of all vehicles and the defect depths of defects on different surfaces in the mixed information as the dependent variables, repeating step S33 to obtain relational equation 3;

[0108] Relationship equations 1 to 3 are summarized as the probability occurrence equation of vehicle defects.

[0109] Furthermore, the specific steps of step S33 are as follows:

[0110] Step S331: construct relation equation 1;

[0111] Step S332: construct relation equation 2;

[0112] Step S3321: using the pandas library and statsmodels library in Python, with the total number of defects on all vehicles on different surfaces in the mixed information as the independent variable and the defect depth of defects on all vehicles on different surfaces as the dependent variable, a regression equation is constructed, which is recorded as regression equation a;

[0113] Step S3322: taking the speed of all vehicles in the mixed information as the independent variable and the total number of defects on different surfaces of all vehicles as the dependent variable, repeat step S3321 to obtain regression equation b;

[0114] The regression equation a and the regression equation b are taken as relational equation 2.

[0115] Furthermore, the specific steps of step S331 are as follows:

[0116] Step S3311: Count the number of all road sections in the target area, denoted as rn;

[0117] The number of defective vehicles corresponding to all road sections in the target area is counted, and recorded as tt(1), tt(2)~tt(rn); among them, tt(1) represents the number of defective vehicles on the first road section; tt(2) represents the number of defective vehicles on the second road section; and so on, tt(rn) represents the number of defective vehicles on the rnth road section;

[0118] Step S3312: define calculation formula 31: pt(i) = tt(i) / QT; where i is a positive number, and the value range of i is 1 to rn; it represents the probability of pt(i) having a defective vehicle on the i-th road section;

[0119] Step S3313: Obtain the number of defective vehicles on all road sections in the target area in the past three years, and use the Markov algorithm to estimate the fitting probability of defective vehicles on all road sections in the target area, recorded as bbn;

[0120] The relationship equation 1 is as follows:

[0121] ; Wherein, i is a positive number, and the value range of i is 1 to rn; C represents the permutation and combination symbol; PTT (i) represents the estimated probability of a defective vehicle occurring on the i-th road section.

[0122] The vehicle defect comprehensive detection system based on part collaboration includes:

[0123] Information acquisition module: used to obtain the side image of the vehicle, driving records and driver age;

[0124] Image analysis module: used to obtain a test picture set and build a vehicle defect recognition model based on the test picture set; use the vehicle defect recognition model to mark and extract pictures of defective areas in the side pictures to obtain a defect picture set; calculate the weighted defect depths of different surfaces of the vehicle based on the defect picture set;

[0125] Comprehensive analysis module: used to obtain the side images, driving records and driver ages of all vehicles in the target area to obtain mixed information; using the vehicle defect recognition model, extract and count the total number of defects on different surfaces and the weighted defect depths on different surfaces corresponding to all vehicles in the mixed information to obtain a defect data set;

[0126] According to the driving records and defect data set, the relationship equations 1 and 2 between defects, road sections and speed are analyzed and constructed;

[0127] Based on the driver age and defect data set, analyze and construct the relationship equation between defect and driver age3;

[0128] Summarize the relationship equations 1 to 3 as the probability occurrence equation of vehicle defects;

[0129] Continuous detection module: used to continuously obtain side images, driving records and driver ages of all vehicles in the target area, and update the probability occurrence equation.

[0130] Compared with the prior art, the present invention has the following beneficial effects:

[0131] Comprehensiveness and accuracy: The present invention can detect different aspects of the vehicle. Through image processing technology, the present invention can capture subtle defect signals and perform accurate identification and classification.

[0132] High efficiency and automation: Traditional vehicle inspection methods are easily affected by human factors. The present invention achieves high automation and can complete the inspection task in a short time. The present invention can automatically judge the inspection results and analyze the causes of vehicle defects through automatic control technology and data analysis algorithms.

[0133] Intelligence and adaptability: The present invention can automatically adjust the detection parameters and strategies according to the changing trends of historical data and real-time data to adapt to the needs of different vehicle models and detection environments; at the same time, the system also has a data traceability function, which can record the process and results of each detection, which is helpful for subsequent quality tracking and troubleshooting. BRIEF DESCRIPTION OF THE DRAWINGS

[0134] Other features, objects and advantages of the present invention will become more apparent from the detailed description of non-limiting embodiments made with reference to the following drawings:

[0135] Figure 1 It is a schematic diagram of the method of the present invention;

[0136] Figure 2 It is a schematic diagram of the system of the present invention;

[0137] Figure 3 Schematic diagrams of different sides of the vehicle of the present invention;

[0138] Figure 4 This is a defect optical schematic diagram of the present invention. DETAILED DESCRIPTION

[0139] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0140] Embodiment 1

[0141] See also Figure 1 , the comprehensive vehicle defect detection method based on part collaboration includes:

[0142] It should be noted that the "test image set" in the present invention means: a image set consisting of images of vehicle surface depressions; and the "target area" refers to the prefecture-level city where the present invention (comprehensive vehicle defect detection method and system based on part collaboration) is applied.

[0143] Step S1: Obtain the side image of the vehicle, driving record and driver age;

[0144] Step S2: obtaining a test picture set, and constructing a vehicle defect recognition model based on the test picture set; using the vehicle defect recognition model, marking and extracting pictures of defective areas in the side pictures to obtain a defect picture set; and calculating weighted defect depths of different surfaces of the vehicle based on the defect picture set;

[0145] The specific steps of step S2 are as follows:

[0146] Step S21: obtaining a test picture set, and constructing a vehicle defect recognition model according to the test picture set;

[0147] Step S211: using the torchvision.models module in the PyTorch library, importing the AlexNet model, the SqueezeNet model, and the GoogLeNet model as the basic models of the vehicle defect recognition model;

[0148] Select "softmax function" for the activation function.

[0149] Step S212: define a BN operation (a Batch Normalization operation that removes linear changes); constrain the underlying eigenvalues ​​of the basic model to a distribution with a mean of 0 and a variance of 1;

[0150] For any input B{x 1 、x 2 ……x m The process of performing BN operation is as follows (input B represents the picture):

[0151] Calculate the mean μB of the elements in set B, ;

[0152] Calculate the variance σB of the elements in set B 2 , ; Standardize the elements in set B and get ;

[0153] Calculate x 1 ~x m get of The calculation formula is as follows:

[0154] ; Wherein, input B refers to the image in the test image set (set B); i represents 1 to m, and m represents the pixel value of the image in input B; x i represents the gray value corresponding to the i-th pixel in image B, and m is the maximum number of pixels in image B;

[0155] Step S213: creating two pooling layers, Pool 1 and Pool 2, for the vehicle defect recognition model;

[0156] Create 5 convolutional layers, starting with the convolutional layer Con1 and the subsequent convolutional layers Conv2 to Conv5;

[0157] Create a connection layer Concat; create two fully connected layers Fc6~Fc8;

[0158] Step S214: upload the image of the test picture set as a primary input to the Con1 layer and the Pool1 layer, perform bottom-level feature extraction, and obtain extracted feature 1;

[0159] Take the extracted feature 1 as the secondary input and upload it to the Conv2 layer and the Pool2 layer to obtain the extracted feature 2;

[0160] Backup extracted feature 2 as the underlying feature;

[0161] Step S125: repeat step S24, upload the extracted feature 2 to the Conv3 layer for processing, and obtain the extracted feature 3; upload the extracted feature 3 to the Conv4 layer for processing, and obtain the extracted feature 4; upload the extracted feature 4 to the Conv5 layer for processing, and obtain the extracted feature 5 as the high-level feature;

[0162] Step S216: perform BN operation on the bottom-level features and high-level features respectively to obtain fused features; use the fused features as the input of the fully connected layer fc6;

[0163] The cross entropy loss function is used as the loss function of the vehicle defect recognition model;

[0164] Step S217: defining the output of the vehicle defect recognition model;

[0165] If the side image is associated with the test image set (i.e., there is an intersection), the output is defective; if the side image is not associated with the test image set (i.e., there is no intersection), no output is performed;

[0166] Step S22: Please refer to Figure 3 , substitute the side image of the vehicle into the vehicle defect recognition model, mark and extract the area images where defects occur on the A side, B side, C side and D side of the vehicle in turn, and obtain a defect image set (i.e., a collection of area images);

[0167] Count the number of area images on the vehicle's A side, denoted as pa;

[0168] Count the number of area images on the vehicle's B side, denoted as pb;

[0169] Count the number of area images on the C side of the vehicle, denoted as pc;

[0170] Count the number of area images of vehicle surface D, recorded as pd;

[0171] pn=pa+pb+pc+pd; where pn represents the number of images in the defect image set;

[0172] Step S23: binarize the images in the defect image set to obtain the grayscale value of each defect image;

[0173] The first region image is recorded as image p, and image p is binarized;

[0174] Step S231: construct a histogram of the image p with the pixel brightness value (usually between 0 and 255, corresponding to the pixel value range of an 8-bit grayscale image) as the abscissa and the frequency of occurrence of the pixel brightness value in the image (i.e., the number of pixels) as the ordinate;

[0175] Step S232: import the OpenCV library; use the cv2.equalizeHist histogram equalization function in the OpenCV library to perform image enhancement processing on the image p to obtain the image p1;

[0176] Step S233: Use the cv2.adaptiveThreshold() function in the OpenCV library to process the image A2 and obtain an adaptive threshold;

[0177] Use the cv2.threshold() function in the OpenCV library to binarize the image p2 according to the adaptive threshold to obtain the image 1 to be analyzed;

[0178] Step S24: repeat the same steps of binarizing the first regional image (steps S231 to S233), and binarize the second to pn-th images to obtain the image 2 to the image pn to be analyzed;

[0179] Step S25: Please refer to Figure 4 , define calculation formulas 1 to 6 for calculating the "defect depth" based on the "grayscale of the defect image and the defect width";

[0180] Step S251: set the depth of the depression to h and the width of the depression to l;

[0181] The angle of incidence of the incident light (i.e. the angle between the incident light and the concave surface) is angle θ (i.e. angle ABC in triangle ABC);

[0182] Calculation formula 1: Angle θ = arccos[(2×h) / l];

[0183] Step S252: let the complementary angle of the incident angle be angle a (i.e. angle ACB in triangle ABC);

[0184] Calculation 2: Angle a = 90 o - angle θ;

[0185] Let the angle subtended by the reflected ray be angle b (i.e. angle BCD in triangle BCD);

[0186] Calculation 3: Angle b = 180 o -2×angle θ;

[0187] Let the supplement of (angle a + angle b) be angle c (that is, angle DCE in triangle CDE);

[0188] Calculation 4: Angle c = 180 o -2×(angle a+angle b);

[0189] Step S253: let the complementary angle of angle c be angle d (i.e. angle CDE in triangle CDE);

[0190] Calculation 5: Angle d = 90 o - Angle c;

[0191] Let the deviation angle of the outgoing ray relative to the incident ray be angle e (i.e. angle FGH in triangle FGH);

[0192] From the triangle laws and the definition of parallel lines, we can know that formula 6: angle d + angle e = angle θ (that is, the angle EDF in triangle DEF is equal to the angle e);

[0193] Step S254: set the gray value of the incident light to HHi (i.e., the gray value of the "non-defective" part in the defect image); set the gray value of the outgoing light to HHo (i.e., the gray value of the "defective" part in the defect image);

[0194] Calculate the attenuation coefficient of the gray value, recorded as do; do = HHo / HHi (i.e. cos(e) = do);

[0195] Calculate the angle of angle e, angle e = arccos(do);

[0196] Substitute the angle e and the concave width into the calculation formula 6-1 in reverse order, and use the angle change relationship and trigonometric functions to obtain the concave depth;

[0197] Step S26: Obtain the grayscale value of each image to be analyzed and the width of each defect, and calculate the depth of each defect on each surface of the vehicle according to the calculation formulas 1 to 6 defined in step S25 to obtain:

[0198] The depth of each defect on the vehicle surface A is denoted as h(A) 1 、h(A) 2 ~h(A) pa ;

[0199] Where h(A) 1 Indicates the depth of the first defect on the A side of the vehicle; h (A) 2 Indicates the depth of the second defect on the A side of the vehicle; and so on, h(A) pa Indicates the depth of the pa-th defect on the vehicle surface A;

[0200] The depth of each defect on the B side of the vehicle, denoted as h(B) 1 、h(B) 2 ~h(B) pb ;

[0201] Among them, h(B) 1 Indicates the depth of the first defect on the B side of the vehicle; h(B) 2 Indicates the depth of the second defect on the B side of the vehicle; and so on, h(B) pb Indicates the depth of the pbth defect on the B side of the vehicle;

[0202] The depth of each defect on the C surface of the vehicle, denoted as h(C) 1 、h(C) 2 ~h(C) pc ;

[0203] Among them, h (C) 1 Indicates the depth of the first defect on the C surface of the vehicle; h(C) 2 Indicates the depth of the second defect on the C surface of the vehicle; and so on, h(C) pc Indicates the depth of the pcth defect on the C surface of the vehicle;

[0204] The depth of each defect on the vehicle surface D is denoted as h(D) 1 、h(D) 2 ~h(D) pd ;

[0205] Among them, h (D) 1 Indicates the depth of the first defect on the A side of the vehicle; h (D) 2 Indicates the depth of the second defect on the A side of the vehicle; and so on, h(D) pd Indicates the depth of the pdth defect on the D surface of the vehicle;

[0206] Step S27: Extract the maximum value, minimum value and mode of the defect depth of each surface of the vehicle to obtain:

[0207] For the A side of the vehicle, the maximum defect depth is denoted as ha(A), the minimum defect depth is denoted as hb(A), and the mode defect depth is denoted as hc(A);

[0208] For vehicle side B, the maximum defect depth is denoted as ha(B), the minimum defect depth is denoted as hb(B), and the mode defect depth is denoted as hc(B);

[0209] For the C side of the vehicle, the maximum defect depth is recorded as ha(C), the minimum defect depth is recorded as hb(C), and the mode defect depth is recorded as hc(C);

[0210] For the vehicle D surface, the maximum defect depth is denoted as ha(D), the minimum defect depth is denoted as hb(D), and the mode defect depth is denoted as hc(D);

[0211] Step S28: Calculate the weighted defect depth of each surface of the vehicle;

[0212] Step S281: Calculate the weighted defect depth of vehicle surface A;

[0213] Definition relation 21: lim(h(A) i / ha(A))→1;

[0214] Define relation 22: lim(h(A) i / hb(A))→1;

[0215] Definition relation 23: lim(h(A) i / hc(A))→1;

[0216] Where i is a positive number, and its value range is 1~pa; h(A) i represents the depth of the i-th defect in the vehicle surface A;

[0217] Step S282: h(A) 1 ~h(A) pa ; Substitute into equations 21 to 23, and record the number of equations that satisfy equation 21 as a1, the number of equations that satisfy equation 22 as a2, and the number of equations that satisfy equation 23 as a3;

[0218] Calculate h(A) 1 ~h(A) pa The average value of is denoted as ah (A);

[0219] Step S283: Calculate the weighted defect depth of the vehicle surface A, recorded as Qh(A); the calculation formula of Qh(A) is as follows:

[0220] ;

[0221] Step S284: Repeat the same steps of calculating the weighted defect depth of vehicle surface A to calculate the weighted defect depths of vehicle surfaces B, C and D.

[0222] Step S3: Obtain the side images, driving records and driver ages corresponding to all (defective) vehicles in the target area to obtain mixed information; use the vehicle defect recognition model to extract and count the total number of defects on different surfaces (i.e., the total number of defect images) and weighted defect depths on different surfaces corresponding to all vehicles in the mixed information to obtain a defect data set;

[0223] According to the driving records and defect data set, the relationship equations 1 and 2 between defects, road sections and speed are analyzed and constructed;

[0224] Based on the driver age and defect data set, analyze and construct the relationship equation between defect and driver age3;

[0225] Summarize the relationship equations 1 to 3 as the probability occurrence equation of vehicle defects;

[0226] The specific steps of step S3 are as follows:

[0227] Step S31: Obtain side images, driving records and driver ages of all defective vehicles in the target area to obtain mixed information;

[0228] Driving records include: the driving route and driving speed (when the vehicle defect occurred);

[0229] Step S32: Obtain the vehicle flow in the target area, recorded as ZT; count the total number of vehicles with defects, recorded as QT;

[0230] Substituting the side image in the mixed information into the vehicle defect recognition model (perform step S2), extracting and counting the total number of defects on different surfaces (i.e., the total number of defect images) and weighted defect depths on different surfaces corresponding to all vehicles in the mixed information, and obtaining a defect data set;

[0231] The total number of defects corresponding to the vehicle A surface in the defect data set is recorded as qa, and the weighted defect depth corresponding to each defect is recorded as haa 1 、haa 2 ~haa qa Among them, haa 1 Indicates the weighted defect depth of the first defect on the A side of the vehicle in the defect dataset; haa 2 represents the weighted defect depth of the second defect on the A side of the vehicle in the defect data set; and so on, haa qa represents the weighted defect depth of the qath defect on the vehicle A side in the defect dataset;

[0232] The total number of defects corresponding to the vehicle B surface in the defect data set is recorded as qb, and the weighted defect depth corresponding to each defect is recorded as hbb 1 、hbb 2 ~hbb qb ; Among them, hbb 1 represents the weighted defect depth of the first defect on the B side of the vehicle in the defect dataset; hbb 2 represents the weighted defect depth of the second defect on the B side of the vehicle in the defect dataset; and so on, hbb qb represents the weighted defect depth of the qbth defect on the B side of the vehicle in the defect dataset;

[0233] The total number of defects corresponding to the C surface of the vehicle in the defect data set is recorded as qc, and the weighted defect depth corresponding to each defect is recorded as hcc 1 、hcc 2 ~hcc qc ; Among them, hcc 1 Indicates the weighted defect depth of the first defect on the C surface of the vehicle in the defect dataset; hcc 2 represents the weighted defect depth of the second defect on the C surface of the vehicle in the defect dataset; and so on, hcc qc represents the weighted defect depth of the qc-th defect on the C surface of the vehicle in the defect dataset;

[0234] The total number of defects corresponding to the vehicle D surface in the defect data set is recorded as qd, and the weighted defect depth corresponding to each defect is recorded as hdd 1 、hdd 2 ~hdd qd ; Among them, hdd 1 Indicates the weighted defect depth of the first defect on the D surface of the vehicle in the defect dataset; hdd 2 Indicates the weighted defect depth of the second defect on the D surface of the vehicle in the defect dataset; and so on, hdd qd represents the weighted defect depth of the qd-th defect on the D surface of the vehicle in the defect dataset;

[0235] Step S33: Analyze and construct relation equations 1 and 2 between defects, road sections and speeds according to the driving records and defect data set;

[0236] Step S331: construct relation equation 1;

[0237] Step S3311: Count the number of all road sections in the target area, denoted as rn;

[0238] The number of defective vehicles corresponding to all road sections in the target area is counted, and recorded as tt(1), tt(2)~tt(rn); among them, tt(1) represents the number of defective vehicles on the first road section; tt(2) represents the number of defective vehicles on the second road section; and so on, tt(rn) represents the number of defective vehicles on the rnth road section;

[0239] Step S3312: define calculation formula 31: pt(i) = tt(i) / QT; where i is a positive number, and the value range of i is 1 to rn; it represents the probability of pt(i) having a defective vehicle on the i-th road section;

[0240] Step S3313: Obtain the number of defective vehicles on all road sections in the target area in the past three years, and use the Markov algorithm to estimate the fitting probability of defective vehicles on all road sections in the target area, recorded as bbn;

[0241] The relationship equation 1 is as follows:

[0242] ; Where i is a positive number, and the value range of i is 1 to rn; C represents the permutation and combination symbol; PTT (i) represents the estimated probability of a defective vehicle occurring on the i-th road section;

[0243] Step S332: construct relation equation 2;

[0244] Step S3321: using the pandas library and statsmodels library in Python, with the total number of defects on different surfaces of all vehicles in the mixed information as the independent variable and the defect depth of defects on different surfaces of all vehicles as the dependent variable, a regression equation is constructed, which is recorded as regression equation a;

[0245] Step S3322: taking the speed of all vehicles in the mixed information as the independent variable and the total number of defects on different surfaces of all vehicles as the dependent variable, repeat step S3321 to obtain regression equation b;

[0246] Regression equation a and regression equation b are taken as relationship equation 2;

[0247] Step S333: taking the age of the driver as the independent variable and the total number of defects on different surfaces of all vehicles and the defect depths of defects on different surfaces in the mixed information as the dependent variables, repeating step S332 to obtain relational equation 3;

[0248] Relationship equations 1 to 3 are summarized as the probability occurrence equation of vehicle defects.

[0249] Step S4: Continuously obtain the side images, driving records and driver ages of all vehicles in the target area, and update the probability occurrence equation.

[0250] Embodiment 2

[0251] See also Figure 2 The vehicle defect comprehensive detection system based on part collaboration includes: an information acquisition module, an image analysis module, a comprehensive analysis module, a continuous detection module, a database and a server; wherein the information acquisition module, the image analysis module, the comprehensive analysis module and the continuous detection module are connected to the database and the server respectively.

[0252] Information acquisition module: used to obtain the side image of the vehicle, driving records and driver age;

[0253] Image analysis module: used to obtain a test picture set and build a vehicle defect recognition model based on the test picture set; use the vehicle defect recognition model to mark and extract pictures of defective areas in the side pictures to obtain a defect picture set; calculate the weighted defect depths of different surfaces of the vehicle based on the defect picture set;

[0254] Comprehensive analysis module: used to obtain the side images, driving records and driver ages of all vehicles in the target area to obtain mixed information; using the vehicle defect recognition model, extract and count the total number of defects on different surfaces (i.e. the total number of defect images) and weighted defect depths on different surfaces corresponding to all vehicles in the mixed information to obtain a defect data set;

[0255] According to the driving records and defect data set, the relationship equations 1 and 2 between defects, road sections and speed are analyzed and constructed;

[0256] Based on the driver age and defect data set, analyze and construct the relationship equation between defect and driver age3;

[0257] Summarize the relationship equations 1 to 3 as the probability occurrence equation of vehicle defects;

[0258] Continuous detection module: used to continuously obtain side images, driving records and driver ages of all vehicles in the target area, and update the probability occurrence equation.

[0259] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters in the formula are set by technical personnel in this field according to actual conditions. For example, if there are weight coefficients and proportional coefficients, their set sizes are to quantify each parameter to obtain a specific value, which is convenient for subsequent comparison. Regarding the size of the weight coefficient and the proportional coefficient, it is sufficient as long as it does not affect the proportional relationship between the parameter and the quantized value.

[0260] Finally, it should be noted that the above-described embodiments are only specific implementations of the present invention, which are used to illustrate the technical solutions of the present invention, rather than to limit them. The protection scope of the present invention is not limited thereto. Although the present invention is described in detail with reference to the above-described embodiments, ordinary technicians in the field should understand that any technician familiar with the technical field can still modify the technical solutions recorded in the above-described embodiments within the technical scope disclosed by the present invention, or can easily think of changes, or make equivalent replacements for some of the technical features therein; and these modifications, changes or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention shall be based on the protection scope of the claims.

Claims

1. A comprehensive vehicle defect detection method based on part collaboration, characterized in that: The method comprises: Step S1: Obtain the side image of the vehicle, driving record and driver age; Step S2: obtaining a test picture set, and constructing a vehicle defect recognition model based on the test picture set; using the vehicle defect recognition model, marking and extracting pictures of defective areas in the side pictures to obtain a defect picture set; and calculating weighted defect depths of different surfaces of the vehicle based on the defect picture set; Step S3: Obtain side images, driving records and driver ages corresponding to all vehicles in the target area to obtain mixed information; use the vehicle defect recognition model to extract and count the total number of defects on different surfaces and weighted defect depths on different surfaces corresponding to all vehicles in the mixed information to obtain a defect data set; According to the driving records and defect data set, the relationship equations 1 and 2 between defects, road sections and speed are analyzed and constructed; Based on the driver age and defect data set, analyze and construct the relationship equation between defect and driver age3; Summarize the relationship equations 1 to 3 as the probability occurrence equation of vehicle defects; Step S4: Continuously obtain side images, driving records and driver ages of all vehicles in the target area, and update the probability equation; The specific steps of step S2 are as follows: Step S21: obtaining a test picture set, and constructing a vehicle defect recognition model according to the test picture set; Step S22: Substituting the side image of the vehicle into the vehicle defect recognition model, marking and extracting the images of the areas where defects occur on the A side, B side, C side and D side of the vehicle in sequence, and obtaining a defect image set; Count the number of area images on the vehicle's A side, denoted as pa; Count the number of area images on the vehicle's B side, denoted as pb; Count the number of area images on the C side of the vehicle, denoted as pc; Count the number of area images of vehicle surface D, recorded as pd; pn=pa+pb+pc+pd; where pn represents the number of images in the defect image set; Step S23: binarize the images in the defect image set to obtain the grayscale value of each defect image; Step S24: repeat the same steps of binarizing the first regional image, and binarize the second to pn-th images to obtain the to-be-analyzed images 2 to pn; Step S25: defining calculation formulas 1 to 6 for calculating the defect depth according to the grayscale of the defect image and the defect width; Step S26: Obtain the grayscale value of each image to be analyzed and the width of each defect, and calculate the depth of each defect on each surface of the vehicle according to the calculation formulas 1 to 6 defined in step S25, and obtain: the depth of each defect on surface A of the vehicle, recorded as h(A)1, h(A)2~h(A) pa ; Where h(A)1 represents the depth of the first defect on the A side of the vehicle; h(A)2 represents the depth of the second defect on the A side of the vehicle; and so on. pa Indicates the depth of the pa-th defect on the vehicle surface A; The depth of each defect on the B side of the vehicle is denoted as h(B)1, h(B)2~h(B) pb ; Among them, h(B)1 represents the depth of the first defect on the B side of the vehicle; h(B)2 represents the depth of the second defect on the B side of the vehicle; and so on. pb Indicates the depth of the pbth defect on the B side of the vehicle; The depth of each defect on the C surface of the vehicle is recorded as h(C)1, h(C)2~h(C) pc ; Among them, h(C)1 represents the depth of the first defect on the C surface of the vehicle; h(C)2 represents the depth of the second defect on the C surface of the vehicle; and so on. pc Indicates the depth of the pcth defect on the C surface of the vehicle; The depth of each defect on the vehicle surface D is recorded as h(D)1, h(D)2~h(D) pd ; Among them, h(D)1 represents the depth of the first defect on the A side of the vehicle; h(D)2 represents the depth of the second defect on the A side of the vehicle; and so on. pd Indicates the depth of the pdth defect on the D surface of the vehicle; Step S27: Extract the maximum value, minimum value and mode of the defect depth of each surface of the vehicle to obtain: For the A side of the vehicle, the maximum defect depth is denoted as ha(A), the minimum defect depth is denoted as hb(A), and the mode defect depth is denoted as hc(A); For vehicle side B, the maximum defect depth is denoted as ha(B), the minimum defect depth is denoted as hb(B), and the mode defect depth is denoted as hc(B); For the C side of the vehicle, the maximum defect depth is recorded as ha(C), the minimum defect depth is recorded as hb(C), and the mode defect depth is recorded as hc(C); For the vehicle D surface, the maximum defect depth is denoted as ha(D), the minimum defect depth is denoted as hb(D), and the mode defect depth is denoted as hc(D); Step S28: Calculate the weighted defect depth of each surface of the vehicle and proceed to step S3; The specific steps of step S25 are as follows: Step S251: set the depth of the depression to h and the width of the depression to l; The angle of incidence of the incident light ray is angle θ; Calculation formula 1: Angle θ = arccos[(2×h) / l]; Step S252: setting the complementary angle of the incident angle to angle a; Calculation 2: Angle a = 90 o - angle θ; Let the angle subtended by the reflected ray be angle b; Calculation 3: Angle b = 180 o -2×angle θ; Let the supplement of (angle a + angle b) be angle c; Calculation 4: Angle c = 180 o -2×(angle a+angle b); Step S253: Set the complementary angle of angle c to angle d; Calculation 5: Angle d = 90 o - Angle c; Let the deviation angle of the outgoing light relative to the incident light be angle e; From the triangle law and the definition of parallel lines, we can know that formula 6: angle d + angle e = angle θ; Step S254: Set the gray value of the incident light to HHi; Set the gray value of the outgoing light to HHo; Calculate the attenuation coefficient of the gray value, denoted as do; do = HHo / HHi; Calculate the angle e, angle e = arccos(do); Substitute the angle e and the concave width into the calculation formula 6-1 in reverse order, and use the angle change relationship and trigonometric functions to obtain the concave depth; The specific steps of step S28 are as follows: Step S281: Calculate the weighted defect depth of vehicle surface A; Definition relation 21: lim(h(A) i / ha(A))→1; Define relation 22: lim(h(A) i / hb(A))→1; Definition relation 23: lim(h(A) i / hc(A))→1; Where i is a positive number, and its value range is 1~pa; h(A) i represents the depth of the i-th defect in the vehicle surface A; Step S282: h(A)1~h(A) pa Substituting into equations 21 to 23, the number of items satisfying equation 21 is recorded as a1, the number of items satisfying equation 22 is recorded as a2, and the number of items satisfying equation 23 is recorded as a3; Calculate h(A)1~h(A) pa The average value of is denoted as ah (A); Step S283: Calculate the weighted defect depth of the vehicle surface A, recorded as Qh(A); the calculation formula of Qh(A) is as follows: ； Step S284: Repeat the same steps of calculating the weighted defect depth of vehicle surface A to calculate the weighted defect depths of vehicle surfaces B, C and D.

2. The vehicle defect comprehensive detection method based on part coordination according to claim 1 is characterized in that: The specific steps of step S21 are as follows: Step S211: using the torchvision.models module in the PyTorch library, importing the AlexNet model, the SqueezeNet model, and the GoogLeNet model as the basic models of the vehicle defect recognition model; The activation function selects the softmax function; Step S212: define a BN operation; constrain the underlying eigenvalues ​​of the basic model to a distribution with a mean equal to 0 and a variance equal to 1; For any input B{x1, x2, ... m The process of performing BN operation is as follows: Calculate the mean μB of the elements in set B, ; Calculate the variance σB of the elements in set B 2 , ; Standardize the elements in set B and get ; Calculate x1~x m get of The calculation formula is as follows: ； Among them, input B refers to the picture in the test picture set; i represents 1 to m, x i represents the gray value corresponding to the i-th pixel in image B, and m is the maximum number of pixels in image B; Step S213: creating two pooling layers, Pool 1 and Pool 2, for the vehicle defect recognition model; Create 5 convolutional layers, starting with Conv1 and subsequent convolutional layers Conv2 to Conv5; Create a connection layer Concat; create two fully connected layers Fc6~Fc8; Step S214: upload the image of the test picture set as a primary input to the Con1 layer and the Pool1 layer, perform bottom-level feature extraction, and obtain extracted feature 1; Take the extracted feature 1 as the secondary input and upload it to the Conv2 layer and the Pool2 layer to obtain the extracted feature 2; Backup extracted feature 2 as the underlying feature; Step S125: repeat step S24, upload the extracted feature 2 to the Conv3 layer for processing, and obtain the extracted feature 3; upload the extracted feature 3 to the Conv4 layer for processing, and obtain the extracted feature 4; upload the extracted feature 4 to the Conv5 layer for processing, and obtain the extracted feature 5 as the high-level feature; Step S216: perform BN operation on the bottom-level features and high-level features respectively to obtain fused features; use the fused features as the input of the fully connected layer Fc6; The cross entropy loss function is used as the loss function of the vehicle defect recognition model; Step S217: defining the output of the vehicle defect recognition model; If the side image is associated with the test image set, the output is defective; if the side image is not associated with the test image set, no output is performed.

3. The vehicle defect comprehensive detection method based on part coordination according to claim 2 is characterized in that: The specific steps of step S23 are as follows: Step S231: record the first region image as image p, and perform binarization processing on image p; Construct a histogram of the image p with the pixel brightness value as the horizontal coordinate and the frequency of the pixel brightness value in the image as the vertical coordinate; Step S232: import OpenCV library; Use the cv2.equalizeHist histogram equalization function in the OpenCV library to perform image enhancement processing on the image p to obtain the image p1; Step S233: Use the cv2.adaptiveThreshold() function in the OpenCV library to process the image p1 and obtain an adaptive threshold; Use the cv2.threshold() function in the OpenCV library to binarize the image p1 according to the adaptive threshold to obtain the image 1 to be analyzed; Step S24: Repeat the same steps of binarizing the first regional image, and binarize the second to pn-th images to obtain the image 2 to the image pn to be analyzed.

4. The vehicle defect comprehensive detection method based on part coordination according to claim 3 is characterized in that: The specific steps of step S3 are as follows: Step S31: Obtain side images, driving records and driver ages of all defective vehicles in the target area to obtain mixed information; Driving records include: driving section and driving speed; Step S32: Obtain the vehicle flow in the target area, recorded as ZT; count the total number of vehicles with defects, recorded as QT; Substitute the side image in the mixed information into the vehicle defect recognition model, extract and count the total number of defects on different surfaces and the weighted defect depths on different surfaces corresponding to all vehicles in the mixed information, and obtain a defect data set; The total number of defects corresponding to the vehicle A surface in the defect data set is recorded as qa, and the weighted defect depth corresponding to each defect is recorded as haa1, haa2~haa qa ; where haa1 represents the weighted defect depth of the first defect on the A side of the vehicle in the defect data set; haa2 represents the weighted defect depth of the second defect on the A side of the vehicle in the defect data set; and so on, haa qa represents the weighted defect depth of the qath defect on the vehicle A side in the defect dataset; The total number of defects corresponding to the vehicle B surface in the defect data set is recorded as qb, and the weighted defect depth corresponding to each defect is recorded as hbb1, hbb2~hbb qb ; where hbb1 represents the weighted defect depth of the first defect on the B side of the vehicle in the defect data set; hbb2 represents the weighted defect depth of the second defect on the B side of the vehicle in the defect data set; and so on, hbb qb represents the weighted defect depth of the qbth defect on the B side of the vehicle in the defect dataset; The total number of defects corresponding to the C surface of the vehicle in the defect data set is recorded as qc, and the weighted defect depth corresponding to each defect is recorded as hcc1, hcc2~hcc qc ; where hcc1 represents the weighted defect depth of the first defect on the C surface of the vehicle in the defect data set; hcc2 represents the weighted defect depth of the second defect on the C surface of the vehicle in the defect data set; and so on, hcc qc represents the weighted defect depth of the qc-th defect on the C surface of the vehicle in the defect dataset; The total number of defects corresponding to the vehicle D surface in the defect data set is recorded as qd, and the weighted defect depth corresponding to each defect is recorded as hdd1, hdd2~hdd qd ; where hdd1 represents the weighted defect depth of the first defect on the D surface of the vehicle in the defect data set; hdd2 represents the weighted defect depth of the second defect on the D surface of the vehicle in the defect data set; and so on, hdd qd represents the weighted defect depth of the qd-th defect on the D surface of the vehicle in the defect dataset; Step S33: Analyze and construct relation equations 1 and 2 between defects, road sections and speeds according to the driving records and defect data set; Step S34: taking the driver's age as the independent variable and the total number of defects on different surfaces of all vehicles and the defect depths of defects on different surfaces in the mixed information as the dependent variables, repeating step S33 to obtain relational equation 3; Relationship equations 1 to 3 are summarized as the probability occurrence equation of vehicle defects.

5. The vehicle defect comprehensive detection method based on part coordination according to claim 4 is characterized in that: The specific steps of step S33 are as follows: Step S331: construct relation equation 1; Step S332: construct relation equation 2; Step S3321: using the pandas library and statsmodels library in Python, with the total number of defects on different surfaces of all vehicles in the mixed information as the independent variable and the defect depth of defects on different surfaces of all vehicles as the dependent variable, a regression equation is constructed, which is recorded as regression equation aa; Step S3322: taking the speed of all vehicles in the mixed information as the independent variable and the total number of defects on different surfaces of all vehicles as the dependent variable, repeat step S3321 to obtain regression equation bb; The regression equation aa and the regression equation bb are taken as relational equation 2.

6. The vehicle defect comprehensive detection method based on part coordination according to claim 5 is characterized in that: The specific steps of step S331 are as follows: Step S3311: Count the number of all road sections in the target area, denoted as rn; The number of defective vehicles corresponding to all road sections in the target area is counted, and recorded as tt(1), tt(2)~tt(rn); among them, tt(1) represents the number of defective vehicles on the first road section; tt(2) represents the number of defective vehicles on the second road section; and so on, tt(rn) represents the number of defective vehicles on the rnth road section; Step S3312: define calculation formula 31: pt(i) = tt(i) / QT; where i is a positive number, and the value range of i is 1 to rn; it represents the probability of pt(i) having a defective vehicle on the i-th road section; Step S3313: Obtain the number of defective vehicles on all road sections in the target area in the past three years, and use the Markov algorithm to estimate the fitting probability of defective vehicles on all road sections in the target area, recorded as bbn; The relationship equation 1 is as follows: ； Among them, i is a positive number, and the value range of i is 1 to rn; C represents the permutation and combination symbol; PTT (i) represents the estimated probability of a defective vehicle occurring on the i-th road section.

7. A vehicle defect comprehensive detection system based on part coordination, applicable to the vehicle defect comprehensive detection method based on part coordination according to any one of claims 1 to 6, characterized in that: The detection system comprises: Information acquisition module: used to obtain the side image of the vehicle, driving records and driver age; Image analysis module: used to obtain a test picture set and build a vehicle defect recognition model based on the test picture set; use the vehicle defect recognition model to mark and extract pictures of defective areas in the side pictures to obtain a defect picture set; calculate the weighted defect depths of different surfaces of the vehicle based on the defect picture set; Comprehensive analysis module: used to obtain the side images, driving records and driver ages of all vehicles in the target area to obtain mixed information; using the vehicle defect recognition model, extract and count the total number of defects on different surfaces and the weighted defect depths on different surfaces corresponding to all vehicles in the mixed information to obtain a defect data set; According to the driving records and defect data set, the relationship equations 1 and 2 between defects, road sections and speed are analyzed and constructed; Based on the driver age and defect data set, analyze and construct the relationship equation between defect and driver age3; Summarize the relationship equations 1 to 3 as the probability occurrence equation of vehicle defects; Continuous detection module: used to continuously obtain side images, driving records and driver ages of all vehicles in the target area, and update the probability occurrence equation.

Images