A wheel hub defect detection method and system based on visual images

Abstract

The application relates to the field of hub defect detection, in particular to a hub defect detection method and system based on visual images; the method comprises the following steps: obtaining hub manufacturing image data and manufacturing data, obtaining hub target data according to the hub manufacturing image data, and obtaining a hub coordinate group according to the hub target data; obtaining hub state data and hub real-time image data, obtaining hub real-time data according to the hub real-time image data; detecting the hub real-time data to obtain a hub image group and a defect detection value; obtaining a defect detection value group according to the hub image group, constructing a hub prediction model according to the hub image group, and obtaining a defect change value group according to the hub prediction model and the defect detection value group; analyzing each hub prediction image through the defect change value group to obtain a comprehensive defect trend value, and evaluating the hub through the comprehensive defect trend value. The application can improve the accuracy of hub defect detection.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of hub defect detection, in particular to a hub defect detection method and system based on visual images. BACKGROUND

[0002] In recent years, with the rapid development of computer vision, image processing and deep learning technology, defect detection methods based on visual images have been widely used in industrial detection. This kind of method realizes the automatic detection and classification of defects by combining image preprocessing, feature extraction and intelligent recognition algorithm through high-resolution industrial cameras to collect hub surface images. As an important component that carries, connects tires and transmits power during vehicle driving, the structural integrity and surface quality of the hub are directly related to the running safety and comfort of the whole vehicle.

[0003] The Chinese patent with publication number CN115830033A discloses a method for detecting surface defects of automobile hub based on machine vision, which includes: obtaining a gray image of the automobile hub area to obtain the original scale parameter of each pixel point in the gray image; obtaining the attention points in the gray image and the approximate degree of defects of each attention point; performing three-dimensional reconstruction on the gray image to obtain surface images under all viewing angles, and obtaining the improvement factor of the corresponding attention point according to the difference in the approximate degree of defects of each attention point in the gray image and the surface image thereof; obtaining the adaptive scale parameter of the corresponding attention point by using the improvement factor of each attention point; based on the adaptive scale parameter and the original scale parameter, using Retinex algorithm to enhance the gray image, and detecting the surface defects of the automobile hub based on the enhanced gray image.

[0004] There are various defects in the hub, and the degree of defects is different. In the prior art, the hub detection method is complex, and the hub defects are prone to be missed, resulting in errors in hub defect detection. The defect detection standard is not clear, which increases the error of hub defect detection. The state of the hub also affects the generation of hub defects. How to improve the accuracy of hub defect detection has become a problem to be solved. SUMMARY

[0005] The present application aims to solve the problems in the background art and provides a hub defect detection method based on visual images.

[0006] The technical scheme of the present application is a hub defect detection method based on visual images, comprising the following steps:

[0007] S1, obtaining hub manufacturing image data and manufacturing data, analyzing the hub manufacturing image data to obtain hub target data, and analyzing the hub target data to obtain hub coordinate groups;

[0008] S2, obtaining wheel hub real-time data according to wheel hub real-time image data through real-time acquisition of wheel hub state data and wheel hub real-time image data, and detecting the wheel hub real-time data through the wheel hub target data and the wheel hub coordinate group to obtain a wheel hub image group and a defect detection value;

[0009] S3, analyzing the wheel hub image group to obtain a defect detection value group, constructing a wheel hub prediction model according to the wheel hub image group, and analyzing the wheel hub prediction model and the defect detection value group to obtain a defect change value group;

[0010] S4, analyzing each wheel hub prediction image through the defect change value group to obtain a comprehensive defect trend value, and evaluating the wheel hub through the comprehensive defect trend value to obtain an evaluation result.

[0011] Preferably, the process of obtaining wheel hub manufacturing image data and manufacturing data, analyzing the wheel hub manufacturing image data to obtain wheel hub target data, and analyzing the wheel hub target data to obtain a wheel hub coordinate group comprises:

[0012] The wheel hub manufacturing image data comprises a wheel hub manufacturing image and a wheel hub position; the manufacturing data comprises manufacturing time, vehicle manufacturing number and wheel hub number;

[0013] Analyzing the wheel hub manufacturing image to obtain a wheel hub hole binary image, and performing geometric analysis and air core feature extraction on the wheel hub hole binary image to obtain a wheel hub air core hole image;

[0014] Performing reference positioning processing on the wheel hub manufacturing image, the reference positioning processing process being: marking the hole outline in the wheel hub manufacturing image through the wheel hub hole binary image, dividing the wheel hub manufacturing image into a plurality of images according to the size and shape of the wheel hub air core hole image as the image division standard to obtain a wheel hub target image, recording the position of the wheel hub air core hole image as a target position base point of the wheel hub manufacturing image, and obtaining a wheel hub target position according to the target position base point of the wheel hub manufacturing image; recording the wheel hub target image, the target position base point and the wheel hub target position as the wheel hub target data;

[0015] Analyzing the wheel hub target image, the target position base point and the wheel hub target position of the wheel hub target data, and constructing a wheel hub coordinate group according to the wheel hub position of the wheel hub manufacturing image and the wheel hub target position.

[0016] Preferably, the process of obtaining wheel hub real-time data according to wheel hub real-time image data through real-time acquisition of wheel hub state data and wheel hub real-time image data comprises:

[0017] The wheel hub state data and the wheel hub real-time image data of each wheel hub of the vehicle are acquired in real time; the wheel hub state data includes static wheel hub data and dynamic wheel hub data; the static wheel hub data includes static time and static state; the dynamic wheel hub data includes dynamic time, dynamic state and rotating speed; the wheel hub real-time image data includes wheel hub real-time image, real-time wheel hub position and real-time time;

[0018] The wheel hub real-time image is analyzed to obtain the wheel hub real-time image, real-time position base point and real-time wheel hub position, and the wheel hub real-time image, real-time position base point and real-time wheel hub position are recorded as wheel hub real-time data.

[0019] Preferably, the process of detecting the wheel hub real-time data by the wheel hub target data and the wheel hub coordinate group to obtain the wheel hub image group and the defect detection value is as follows:

[0020] The wheel hub real-time data is detected by the wheel hub target data and the wheel hub coordinate group, and when the real-time wheel hub position and the real-time wheel hub position are consistent with the corresponding positions in the wheel hub coordinate group, the wheel hub target image corresponding to the wheel hub coordinate group is matched with the wheel hub real-time image to construct the wheel hub image group.

[0021] The wheel hub real-time image and the wheel hub target image in the wheel hub image group are subjected to LBP processing to obtain target feature values and real-time feature values; when the target feature values and the real-time feature values are not equal, the defect detection value is obtained according to the target feature values and the real-time feature values.

[0022] Preferably, the process of analyzing the wheel hub image group to obtain the defect detection value group includes:

[0023] The wheel hub image group is continuously analyzed, and the continuous analysis refers to continuously acquired wheel hub images, and the acquired wheel hub images are recorded through the wheel hub image group, so that the wheel hub image group is recorded as [wheel hub target image, wheel hub image t1, wheel hub image t2, …, wheel hub real-time image tv]; t1, t2, …, tv respectively refer to the acquisition time of the wheel hub image.

[0024] Each wheel hub image in the wheel hub image group is subjected to LBP processing to respectively obtain feature value 1, feature value 2, …, feature value v; the target feature value, feature value 1, feature value 2, …, and feature value v are analyzed, and when the adjacent feature values are not equal, the defect detection v value is obtained according to the latter adjacent feature value and the former feature value, and the defect detection value group is obtained by analyzing the defect detection v value.

[0025] Preferably, the wheel hub prediction model is constructed according to the wheel hub image group, and the process of analyzing the wheel hub prediction model and the defect detection value group to obtain the defect change value group includes:

[0026] The wheel hub images in each wheel hub image group and the acquisition time are taken as a training set and a test set, and the training set and the test set are input into the wheel hub prediction model to train the wheel hub prediction model, and a trained wheel hub prediction model is obtained, and corresponding wheel hub prediction images and prediction time are output;

[0027] The wheel hub prediction images are subjected to LBP processing to obtain prediction characteristic values, and prediction defect detection values are obtained according to the prediction characteristic values and the characteristic values v; defect change v values and prediction defect change values are obtained according to the prediction defect detection values, the prediction time, the defect detection value groups and the acquisition time; and defect change value groups are obtained according to the defect change v values and the prediction defect change values.

[0028] Preferably, the wheel hub prediction images are analyzed by the defect change value groups to obtain comprehensive defect trend values, and the wheel hubs are evaluated by the comprehensive defect trend values to obtain evaluation results, and the process includes:

[0029] The wheel hub prediction images of each wheel hub of the vehicle are analyzed by the wheel hub manufacturing images and the wheel hub positions of the wheel hub manufacturing image data, the average change rates of the defect change v values and the prediction defect change values in the defect change value groups corresponding to each wheel hub prediction image of each wheel hub are calculated, the average change rates of each wheel hub prediction image are obtained, the average values of the average change rates of each wheel hub prediction image in each wheel hub are calculated, the maximum average value is taken as the comprehensive defect trend value;

[0030] A defect trend threshold value is set, when the comprehensive defect trend value is greater than or equal to the defect trend threshold value, the vehicle is immediately stopped, the wheel hub defects of the vehicle are too many, and dangerous events are prone to occur, and an evaluation result is generated;

[0031] When the comprehensive defect trend value is less than the defect trend threshold value, the wheel hub images of the vehicle are continuously monitored.

[0032] The application also discloses a wheel hub defect detection system based on visual images, which comprises a management center, and the management center is communicatively connected with a data acquisition module, a data detection module, a data processing module and a data evaluation module.

[0033] The data acquisition module is used for acquiring wheel hub manufacturing image data and manufacturing data, analyzing the wheel hub manufacturing image data to obtain wheel hub target data, and analyzing the wheel hub target data to obtain a wheel hub coordinate group.

[0034] The data detection module is used for acquiring wheel hub state data and wheel hub real-time image data in real time through the manufacturing data, obtaining wheel hub real-time data according to the wheel hub real-time image data, detecting the wheel hub real-time data through the wheel hub target data and the wheel hub coordinate group, and obtaining a wheel hub image group and a defect detection value.

[0035] The data processing module is used for analyzing the hub image group to obtain a defect detection value group, constructing a hub prediction model according to the hub image group, analyzing the hub prediction model and the defect detection value group to obtain a defect change value group;

[0036] The data evaluation module is used for analyzing each hub prediction image through the defect change value group to obtain a comprehensive defect trend value, evaluating the hub through the comprehensive defect trend value to obtain an evaluation result.

[0037] Compared with the prior art, the above technical scheme of the present application has the following beneficial technical effects: the hub manufacturing image data and manufacturing data are obtained, the hub manufacturing image data is analyzed to obtain hub target data, and the hub target data is analyzed to obtain a hub coordinate group; the hub state data and hub real-time image data are obtained in real time through the manufacturing data, the hub real-time data is obtained according to the hub real-time image data; the hub image group and the defect detection value are obtained by detecting the hub real-time data through the hub target data and the hub coordinate group, the accuracy and real-time performance of hub defect detection are improved by analyzing the hub target data and the hub real-time data through the hub coordinate group; the hub image group is analyzed to obtain a defect detection value group, a hub prediction model is constructed according to the hub image group, the hub prediction model and the defect detection value group are analyzed to obtain a defect change value group; the change process of hub defects is embodied through the defect change value group, the accuracy of hub defect detection is improved, and the error of hub defect detection is reduced; the trend of hub defects is predicted through the hub image group, the defect detection value and the hub prediction model, all the hubs of the vehicle are evaluated through the defect change value group, the hub prediction image and the comprehensive defect trend value, and the hub defect detection is intelligentized and systematized in combination with the hub running state. BRIEF DESCRIPTION OF DRAWINGS

[0038] Figure 1 A flowchart of an embodiment of the present application is shown. DETAILED DESCRIPTION

[0039] Embodiment one, as shown in the figure, the present application puts forward a kind of based on visual image's hub defect detection method, including the following steps: Figure 1

[0040] S1, obtain hub manufacturing image data and manufacturing data, analyze the hub manufacturing image data to obtain hub target data, and analyze the hub target data to obtain a hub coordinate group;

[0041] ​S2, obtain the wheel hub real-time data according to the wheel hub real-time image data through the wheel hub real-time image data and the wheel hub real-time manufacturing data; and obtain the wheel hub image group and the defect detection value through the wheel hub target data and the wheel hub coordinate group and the wheel hub real-time data detection;

[0042] S3, analyze the wheel hub image group to obtain the defect detection value group, construct the wheel hub prediction model according to the wheel hub image group, and analyze the wheel hub prediction model and the defect detection value group to obtain the defect change value group;

[0043] S4, analyze each wheel hub prediction image through the defect change value group to obtain the comprehensive defect trend value, and evaluate the wheel hub through the comprehensive defect trend value to obtain the evaluation result.

[0044] It should be further explained that, in the specific implementation process, the wheel hub manufacturing image data and the manufacturing data are obtained, the wheel hub target data is obtained by analyzing the wheel hub manufacturing image data, and the wheel hub coordinate group is obtained by analyzing the wheel hub target data.

[0045] The wheel hub manufacturing image data includes wheel hub manufacturing images and wheel hub positions; the wheel hub manufacturing image refers to the overall image of each wheel hub of the vehicle when the vehicle manufacturing is completed; the wheel hub position refers to the wheel hub position corresponding to the wheel hub manufacturing image, and the determination standard of the wheel hub position is unified;

[0046] It should be further explained that, in the specific implementation process, the wheel hub position is exemplified, assuming that the vehicle has four wheel hubs, then the positions of the wheel hubs are left front, left rear, right front and right rear respectively based on the position opposite to the vehicle head, and the determination standard of the wheel hub position of all vehicles is consistent with the above example content;

[0047] The manufacturing data includes manufacturing time, vehicle manufacturing number and wheel hub number; the manufacturing time refers to the time when the vehicle manufacturing is completed; the vehicle manufacturing number refers to the number of the completed vehicle manufacturing, and the vehicle manufacturing number has uniqueness;

[0048] The wheel hub manufacturing image data is analyzed, and the small hole feature extraction is performed on the wheel hub manufacturing image; the small hole feature extraction process refers to gray processing of the wheel hub manufacturing image, converting the wheel hub manufacturing image into a gray image, denoising the gray image, and performing Canny edge detection on the denoised image to obtain a wheel hub hole binary image, and performing geometric analysis and air core feature extraction on the wheel hub hole binary image to obtain a wheel hub air core image;

[0049] The hub manufacturing image is subjected to a reference positioning process, which is a process of marking the hole profile in the hub manufacturing image by a hub hole binary image, dividing the hub manufacturing image into a plurality of images by taking the size and shape of the hub core hole image as the image division standard, obtaining a hub target image, obtaining the position of the hub target image, taking the position of the hub core hole image as the target position base point of the hub manufacturing image, updating the position of the hub target image based on the target position base point of the hub manufacturing image, and obtaining a hub target position; the hub target image, the target position base point and the hub target position are taken as hub target data;

[0050] The hub target data, i.e., the hub target image, the target position base point and the hub target position, are analyzed, and a hub coordinate group [x i ,y i ] is constructed based on the hub position of the hub manufacturing image where the hub target image is located and the hub target position corresponding to the hub target image; x i refers to the hub position of the hub manufacturing image where the hub target image is located; y i refers to the hub target position; and i refers to the number of the hub target image.

[0051] It should be further explained that in the specific implementation process, the hub state data and the hub real-time image data are obtained in real time based on the manufacturing data, the hub real-time data is obtained based on the hub real-time image data, and the hub image group and the defect detection value are obtained by detecting the hub real-time data based on the hub target data and the hub coordinate group.

[0052] The hub state data and the hub real-time image data of each hub of the vehicle are obtained in real time based on the vehicle manufacturing number and the hub number of the manufacturing data; the hub state data refers to the state of the hub on the vehicle, including static hub data and dynamic hub data; the static hub data includes static time and static state; the dynamic hub data includes dynamic time, dynamic state and rotational speed.

[0053] The hub real-time image data refers to the real-time overall image of each hub of the vehicle after the vehicle is manufactured, including the hub real-time image, the real-time hub position and the real-time time.

[0054] The hub real-time image data is analyzed, the small hole feature of the hub real-time image is extracted, the hub real-time image, the real-time position base point and the hub real-time position are obtained, and the hub real-time image, the real-time position base point and the hub real-time position are taken as hub real-time data.

[0055] It needs to be further explained that in the specific implementation process, the specific process of extracting small hole features from the wheel hub real-time image is consistent with the specific process of extracting small hole features from the wheel hub manufacturing image; the wheel hub real-time image is obtained through a camera, and if the wheel hub state is a dynamic state, the shooting speed and mode of the camera can be adjusted according to the rotating speed in the dynamic state, for example, high-speed shooting, etc.

[0056] The wheel hub real-time data is detected through the wheel hub target data and the wheel hub coordinate group, and the wheel hub real-time position and the real-time wheel hub position of the wheel hub real-time data are found through the wheel hub coordinate group [x i ,y i ]; when the wheel hub real-time position and the real-time wheel hub position are consistent with the corresponding positions in the wheel hub coordinate group [x i ,y i ], the wheel hub target image corresponding to the wheel hub coordinate group [x i ,y i ] is matched with the wheel hub real-time image, and the wheel hub image group [wheel hub target image, wheel hub real-time image] is constructed; the wheel hub image group will be continuously expanded according to the acquisition time sequence of the wheel hub real-time image.

[0057] The images in the wheel hub image group are analyzed, the wheel hub real-time image and the wheel hub target image in the wheel hub image group are subjected to LBP processing, and target feature values and real-time feature values are obtained; when the target feature values and the real-time feature values are equal, the wheel hub real-time image in the wheel hub image group has no change compared with the wheel hub target image; when the target feature values and the real-time feature values are not equal, the wheel hub real-time image in the wheel hub image group has a change compared with the wheel hub target image, and the absolute value of the difference between the target feature values and the real-time feature values is calculated to obtain a defect detection value.

[0058] It needs to be further explained that in the specific implementation process, the wheel hub image group is analyzed to obtain a defect detection value group, a wheel hub prediction model is constructed according to the wheel hub image group, and the process of analyzing the wheel hub prediction model and the defect detection value group to obtain a defect change value group is as follows:

[0059] The wheel hub image group is continuously analyzed, and the continuous analysis refers to continuously acquired wheel hub images, and the acquired wheel hub images are recorded through the wheel hub image group, so the wheel hub image group is recorded as [wheel hub target image, wheel hub image t1, wheel hub image t2, …, wheel hub real-time image tv];

[0060] It needs to be further explained that in the specific implementation process, [hub target image, hub image t1, hub image t2, …, hub real-time image tv] is the composition of the hub image group, t1, t2, …, tv respectively refer to the acquisition time of the hub image, and tv is the current real-time time; the continuous acquisition in the continuous analysis refers to continuously repeating S2: acquiring the hub real-time image of the hub real-time image data and the hub state data in real time through the manufacturing data; the hub images t1, t2, …, tV in the hub image group [hub target image, hub image t1, hub image t2, …, hub real-time image tv] are sequentially subjected to LBP processing, and feature values 1, 2, …, v are respectively obtained;

[0061] The target feature value, feature value 1, feature value 2, …, and feature value v are analyzed, when the adjacent feature values are not equal, then the hub image changes, and the difference between the next adjacent feature value and the previous feature value is calculated to obtain a defect detection v value. The defect detection v value is summarized in order of v from small to large to obtain a defect detection value group.

[0062] It needs to be further explained that in the specific implementation process, for example, if the defect detection v value is obtained by the difference between the feature value 3 and the feature value 2, then v=3; if the defect detection v value is obtained by the difference between the feature value 6 and the feature value 5, then v=6; and so on. If the defect detection v value is obtained by the difference between the feature value u and the feature value u-1, and u

[0063] A plurality of hub image groups and acquisition times in each hub image group are used as a training set and a test set, and the training set and the test set are input into the hub prediction model to train the hub prediction model, obtain the trained hub prediction model, and output the corresponding hub prediction image and prediction time;

[0064] The hub prediction image is subjected to LBP processing to obtain a prediction feature value, the prediction feature value is compared with the feature value v, when the prediction feature value is not equal to the feature value v, the difference between the prediction feature value and the feature value v is calculated to obtain a prediction defect detection value; the change rate between the defect detection values is calculated by the prediction defect detection value, the prediction time, the defect detection value group, and the acquisition time, to obtain a defect change v value and a prediction defect change value, and the defect change v value is summarized in order of v from small to large, and the prediction defect change value is inserted into the last to obtain a defect change value group.

[0065] It needs to be further explained that, in the specific implementation process, the wheel hub prediction image of each wheel hub is analyzed by the defect change value group, the comprehensive defect trend value is obtained, the wheel hub is evaluated by the comprehensive defect trend value, and the evaluation result is obtained.

[0066] The wheel hub prediction image of each wheel hub of the vehicle is analyzed by the wheel hub manufacturing image data of the wheel hub manufacturing image and the wheel hub position, the average change rate of the defect change value group corresponding to each wheel hub prediction image of each wheel hub is calculated, the average change rate of each wheel hub prediction image is obtained, and the average value of the average change rate of each wheel hub prediction image in each wheel hub is calculated, and the maximum average value is taken as the comprehensive defect trend value.

[0067] A defect trend threshold is set, when the comprehensive defect trend value is greater than or equal to the defect trend threshold, the vehicle stops immediately, the wheel hub defect of the vehicle is too much, and dangerous events are prone to occur, and an evaluation result is generated;

[0068] When the comprehensive defect trend value is less than the defect trend threshold, the wheel hub image of the vehicle is continuously monitored.

[0069] In embodiment two, a wheel hub defect detection system based on visual image is provided, which is applied to the wheel hub defect detection method based on visual image in embodiment one, and specifically includes a management center, which is in communication connection with a data acquisition module, a data detection module, a data processing module and a data evaluation module.

[0070] The data acquisition module is used to acquire wheel hub manufacturing image data and manufacturing data, analyze the wheel hub manufacturing image data to obtain wheel hub target data, and analyze the wheel hub target data to obtain a wheel hub coordinate group.

[0071] The data detection module is used to acquire wheel hub state data and wheel hub real-time image data in real time through the manufacturing data, obtain wheel hub real-time data according to the wheel hub real-time image data, detect the wheel hub real-time data through the wheel hub target data and the wheel hub coordinate group, and obtain a wheel hub image group and a defect detection value.

[0072] The data processing module is used to analyze the wheel hub image group to obtain a defect detection value group, construct a wheel hub prediction model according to the wheel hub image group, analyze the wheel hub prediction model and the defect detection value group, and obtain a defect change value group.

[0073] The data evaluation module is used to analyze each wheel hub prediction image by the defect change value group, obtain a comprehensive defect trend value, evaluate the wheel hub by the comprehensive defect trend value, and obtain an evaluation result.

[0074] The embodiments of the present application are described in detail above with reference to the accompanying drawings, but the present application is not limited to the embodiments, and various changes can be made by those skilled in the art within the scope of knowledge acquired from the present disclosure, without departing from the spirit of the present application.

Claims

1. A wheel hub defect detection method based on visual images, characterized by, The method comprises the following steps: S1, obtaining hub manufacturing image data and manufacturing data, analyzing the hub manufacturing image data to obtain hub target data, and analyzing the hub target data to obtain a hub coordinate group, comprising: The hub manufacturing image data comprises a hub manufacturing image and a hub position; the manufacturing data comprises manufacturing time, vehicle manufacturing number, and hub number; The hub manufacturing image is analyzed to obtain a hub hole binary image, the hub hole binary image is geometrically analyzed and air core hole features are extracted to obtain a hub air core hole image; The hub manufacturing image is subjected to reference positioning processing, the reference positioning processing process being: the hole profile in the hub manufacturing image is marked by the hub hole binary image, the hub manufacturing image is divided into a plurality of images according to the size and shape of the hub air core hole image as the image division standard to obtain hub target images, the position of the hub air core hole image is recorded as the target position base point of the hub manufacturing image, and the hub target position is obtained according to the target position base point of the hub manufacturing image; the hub target images, the target position base point, and the hub target position are recorded as the hub target data; The hub target images, the target position base point, and the hub target position of the hub target data are analyzed, and the hub coordinate group is constructed according to the hub position of the hub manufacturing image and the hub target position; S2, obtaining hub state data and hub real-time image data in real time through the manufacturing data, obtaining hub real-time data according to the hub real-time image data; detecting the hub real-time data through the hub target data and the hub coordinate group to obtain a hub image group and a defect detection value; S3, analyzing the hub image group to obtain a defect detection value group, constructing a hub prediction model according to the hub image group, analyzing the hub prediction model and the defect detection value group to obtain a defect change value group; S4, analyzing each hub prediction image through the defect change value group to obtain a comprehensive defect trend value, evaluating the hub through the comprehensive defect trend value to obtain an evaluation result.

2. The wheel hub defect detection method based on visual images according to claim 1, characterized in that, The process of obtaining hub state data and hub real-time image data in real time through manufacturing data, and obtaining hub real-time data according to hub real-time image data comprises: Real-time acquisition of hub state data and hub real-time image data for each hub of a vehicle; the hub state data comprises static hub data and dynamic hub data; the static hub data comprises static time and static state; the dynamic hub data comprises dynamic time, dynamic state, and rotational speed; the hub real-time image data comprises a hub real-time image, a real-time hub position, and a real-time time; The hub real-time image is analyzed to obtain a hub real-time image, a real-time position base point, and a hub real-time position, and the hub real-time image, the real-time position base point, and the hub real-time position are recorded as the hub real-time data.

3. The wheel hub defect detection method based on visual images according to claim 2, characterized in that, The process of detecting the hub real-time data through the hub target data and the hub coordinate group to obtain a hub image group and a defect detection value comprises: The wheel hub real-time data is detected through the wheel hub target data and the wheel hub coordinate group, when the wheel hub real-time position and the wheel hub real-time position are consistent with the corresponding positions in the wheel hub coordinate group, the wheel hub target image corresponding to the wheel hub coordinate group is matched with the wheel hub real-time image, and the wheel hub image group is constructed; The wheel hub real-time image and the wheel hub target image in the wheel hub image group are subjected to LBP processing to obtain a target feature value and a real-time feature value; When the target feature value and the real-time feature value are not equal, a defect detection value is obtained according to the target feature value and the real-time feature value.

4. The wheel hub defect detection method based on visual images according to claim 3, characterized in that, The process of obtaining the defect detection value group by analyzing the wheel hub image group includes: The wheel hub image group is continuously analyzed, which means that the wheel hub images are continuously acquired and the acquired wheel hub images are recorded by the wheel hub image group, and the wheel hub image group is denoted as ; t1, t2, …, tv respectively denote the acquisition time of the wheel hub image; Each wheel hub image in the wheel hub image group is subjected to LBP processing respectively to obtain feature value 1, feature value 2,..., and feature value v; the target feature value, feature value 1, feature value 2,..., and feature value v are analyzed, when adjacent feature values are not equal, a defect detection value is obtained according to the latter adjacent feature value and the former feature value, and the defect detection value is analyzed to obtain the defect detection value group.

5. The method of claim 4, wherein the method further comprises: The process of obtaining the defect change value group by constructing the wheel hub prediction model according to the wheel hub image group, analyzing the wheel hub prediction model and the defect detection value group includes: A plurality of wheel hub images and acquisition times in each wheel hub image group are taken as a training set and a test set, the training set and the test set are input into the wheel hub prediction model, the wheel hub prediction model is trained, the wheel hub prediction model after training is completed is obtained, and corresponding wheel hub prediction images and prediction times are output; The wheel hub prediction images are subjected to LBP processing to obtain prediction feature values, prediction defect detection values are obtained according to the prediction feature values and the feature value v; defect change values and prediction defect change values are obtained according to the prediction defect detection values, the prediction times, the defect detection value group and the acquisition times, and the defect change value group is obtained according to the defect change values and the prediction defect change values.

6. The wheel hub defect detection method based on visual images according to claim 5, characterized in that, The process of obtaining the comprehensive defect trend value by analyzing each wheel hub prediction image through the defect change value group and evaluating the wheel hub through the comprehensive defect trend value to obtain an evaluation result includes: The wheel hub prediction images of each wheel hub of the vehicle are analyzed through the wheel hub manufacturing images and the wheel hub positions of the wheel hub manufacturing image data, the average change rates of the defect change values and the prediction defect change values in the defect change value groups corresponding to each wheel hub prediction image of each wheel hub are calculated, the average change rates of each wheel hub prediction image are obtained, the average values of the average change rates of each wheel hub prediction image in each wheel hub are calculated, the maximum average value is taken as the comprehensive defect trend value; A defect trend threshold is set, when the comprehensive defect trend value is greater than or equal to the defect trend threshold, the vehicle stops immediately, the wheel hub defects of the vehicle are too many, dangerous events are prone to occur, and an evaluation result is generated; When the comprehensive defect trend value is less than the defect trend threshold, the wheel hub images of the vehicle are continuously monitored.

7. A visual image based wheel hub defect detection system, particularly applied to the visual image based wheel hub defect detection method of any one of claims 1 to 6, comprising a management center, characterized in that, The management center is communicatively connected with a data acquisition module, a data detection module, a data processing module and a data evaluation module: The data acquisition module is configured to acquire hub manufacturing image data and manufacturing data, analyze the hub manufacturing image data, obtain hub target data, and analyze the hub target data to obtain a hub coordinate group; The data detection module is configured to acquire hub state data and hub real-time image data in real time through the manufacturing data, obtain hub real-time data according to the hub real-time image data, detect the hub real-time data through the hub target data and the hub coordinate group, and obtain a hub image group and a defect detection value; The data processing module is configured to analyze the hub image group to obtain a defect detection value group, construct a hub prediction model according to the hub image group, analyze the hub prediction model and the defect detection value group, and obtain a defect change value group; The data evaluation module is configured to analyze each hub prediction image through the defect change value group to obtain a comprehensive defect trend value, evaluate the hub through the comprehensive defect trend value, and obtain an evaluation result.

Images