Automobile steering wheel appearance defect detection method based on machine vision

Abstract

The present application provides a method for detecting appearance defects of automobile steering wheels based on machine vision, which relates to the field of defect detection technology, including: activating an ultrasonic probe to perform high-frequency acoustic wave detection on a target steering wheel; performing curve processing on the timing of a first echo signal of a first detection point; determining whether the first curve deviation meets a curve deviation threshold; if not, activating a laser transmitter to perform laser scanning detection on the first detection point; performing local three-dimensional modeling on the first detection point; comparing the first local point cloud model with a preset point cloud model of the target; analyzing the first point cloud deviation set to obtain the target appearance defect. This application can solve the technical problem in the prior art of inaccurate detection of non-obvious defects due to the fact that some defects only appear under specific conditions. By combining ultrasonic detection, laser scanning and three-dimensional modeling, the appearance defects of the steering wheel can be detected from multiple levels, thereby improving the accuracy of defect detection.

Description

Technical Field

[0001] The present application relates to the technical field of defect detection, and in particular to a method for detecting appearance defects of automobile steering wheels based on machine vision. Background Art

[0002] A car's steering wheel is a crucial component of vehicle control. It must not only offer excellent handling and comfort, but also meet stringent quality standards for appearance. Common cosmetic defects include, but are not limited to, scratches, cracks, bubbles, color variations, indentations, and irregular shapes. The steering wheel's material (leather, plastic, metal) and the lighting conditions during production can affect machine vision recognition. Visual inspection typically relies on visual images, capturing visible defects on an object's surface through a camera. This method can typically only detect surface defects, making it extremely difficult to identify less noticeable defects, those located within the material, or those that are difficult to detect in bright light (such as microcracks, bubbles, or deformation caused by internal stress).

[0003] In summary, the prior art has a technical problem in that certain defects only appear under specific conditions, resulting in inaccurate detection of non-visible defects. Summary of the Invention

[0004] The purpose of this application is to provide a method for detecting appearance defects of automobile steering wheels based on machine vision, so as to solve the technical problem in the prior art that some defects are only visible under specific conditions, resulting in inaccurate detection of non-visible defects.

[0005] In order to achieve the above-mentioned purpose, the present application provides a method for detecting the appearance defects of a car steering wheel based on machine vision, wherein the method for detecting the appearance defects of a car steering wheel based on machine vision includes: activating an ultrasonic probe to perform high-frequency acoustic wave detection on a target steering wheel and obtaining an echo signal record; performing curve processing on the first echo signal time series of a first detection point in the echo signal record to obtain a first echo spline curve; judging whether the first curve deviation obtained by comparing the first echo spline curve with a preset spline curve meets a curve deviation threshold; if the first curve deviation does not meet the curve deviation threshold, activating a laser emitter to perform laser scanning detection on the first detection point to obtain a first laser record; performing local three-dimensional modeling on the first detection point of the target steering wheel based on the first laser record to obtain a first local point cloud model; comparing the first local point cloud model with the target preset point cloud model of the target steering wheel to obtain a first point cloud deviation set; analyzing the first point cloud deviation set to obtain a target appearance defect of the target steering wheel.

[0006] Optionally, a preset influence structure diagram is obtained, wherein the preset influence structure diagram includes a preset ultrasonic influence structure diagram; an ultrasonic influence factor group is formed according to the preset ultrasonic influence structure diagram; dynamic monitoring of high-frequency sound wave detection is performed based on the ultrasonic influence factor group to obtain an ultrasonic influence factor parameter set; the ultrasonic influence factor parameter set is rendered to the preset ultrasonic influence structure diagram to obtain a dynamic ultrasonic influence structure diagram; and the first echo signal timing is calibrated and adjusted according to the dynamic ultrasonic influence structure diagram.

[0007] Optionally, obtain a historical ultrasonic detection database; extract a first historical detection data group from the historical ultrasonic detection database, the first historical detection data group including first historical detection condition information and first historical detection accuracy; perform a correlation analysis on the first historical detection condition information and the first historical detection accuracy to obtain a first correlation result; and construct the preset ultrasonic influence structure diagram based on the first correlation result.

[0008] Optionally, the first historical detection condition information includes first historical steering wheel condition information, first historical probe condition information and first historical environmental condition information.

[0009] Optionally, obtain any condition indicator; match any correlation coefficient corresponding to the any condition indicator in the first correlation result; when the any correlation coefficient reaches a correlation threshold, call a predetermined rendering strategy to render the any condition indicator to the preset ultrasonic impact structure diagram.

[0010] Optionally, according to the predetermined rendering strategy, the arbitrary condition indicator is used as the influence endpoint and the arbitrary correlation coefficient is used as the influence length; the arbitrary influence structure of the arbitrary condition indicator is obtained according to the influence endpoint and the influence length; the arbitrary influence structure is added to the initial influence structure diagram to obtain the preset ultrasonic influence structure diagram.

[0011] Optionally, matrix vectorization processing is performed on the dynamic ultrasonic influence structure diagram to obtain a dynamic ultrasonic influence vector; and the dynamic vector eigenvalue of the dynamic ultrasonic influence vector is used as a weight to calibrate and adjust the timing of the first echo signal.

[0012] Optionally, a first echo scatter plot of the first echo signal time series is obtained; polynomial regression fitting is performed on the first echo scatter plot to obtain a first echo fitting formula; and a first echo fitting curve of the first echo fitting formula is used as the first echo spline curve.

[0013] Optionally, a preset laser influence structure diagram is extracted from the preset influence structure diagram; a laser influence factor group is formed according to the preset laser influence structure diagram; dynamic monitoring of laser scanning detection is performed based on the laser influence factor group to obtain a laser influence factor parameter set; the laser influence factor parameter set is rendered to the preset laser influence structure diagram to obtain a dynamic laser influence structure diagram; and the first laser record is calibrated and adjusted according to the dynamic laser influence structure diagram.

[0014] The technical solution provided in this application has at least the following technical effects or advantages:

[0015] The target steering wheel is subjected to high-frequency acoustic wave detection by activating an ultrasonic probe, and an echo signal record is obtained; the first echo signal time series of the first detection point in the echo signal record is curve-processed to obtain a first echo spline curve; it is determined whether the first curve deviation obtained by comparing the first echo spline curve with the preset spline curve meets the curve deviation threshold; if the first curve deviation does not meet the curve deviation threshold, the laser transmitter is activated to perform laser scanning detection on the first detection point to obtain a first laser record; based on the first laser record, local three-dimensional modeling is performed on the first detection point of the target steering wheel to obtain a first local point cloud model; the first local point cloud model is compared with the target preset point cloud model of the target steering wheel to obtain a first point cloud deviation set; the first point cloud deviation set is analyzed to obtain the target appearance defect of the target steering wheel. In other words, ultrasonic detection is used to perform high-frequency sound wave detection on the steering wheel, and the echo signal is processed into a curve and compared with the preset spline curve to determine whether there is a deviation. If there is a deviation, the laser emitter is activated for laser scanning, and local three-dimensional modeling is performed. It is compared with the target preset point cloud model, the point cloud deviation is analyzed, and the appearance defects are identified, thereby improving the accuracy of the appearance defect detection of the automobile steering wheel.

[0016] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, which can be implemented in accordance with the contents of the description, and to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are specifically listed below. It should be understood that the content described in this section is not intended to identify the key or important features of the embodiments of the present application, nor is it intended to limit the scope of the present application. Other features of the present application will become easy to understand through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions in this application or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are merely exemplary. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without any creative work.

[0018] Figure 1 This is a flow chart of a method for detecting appearance defects of an automobile steering wheel based on machine vision in this application;

[0019] Figure 2 This is a flow chart of calibrating and adjusting the timing of the first echo signal in the machine vision-based automobile steering wheel appearance defect detection method of this application. DETAILED DESCRIPTION

[0020] This application provides a machine vision-based method for detecting exterior defects in automotive steering wheels, resolving the existing technical problem of inaccurate detection of non-visible defects, which occurs when certain defects only appear under specific conditions. Ultrasonic detection performs high-frequency acoustic detection on the steering wheel, curves the echo signal, and compares it with a preset spline curve to determine whether there is a deviation. If so, a laser transmitter is activated for laser scanning, and local 3D modeling is performed. This is then compared with a preset point cloud model of the target, analyzing point cloud deviations and identifying exterior defects, thereby improving the accuracy of exterior defect detection for automotive steering wheels.

[0021] Below, the technical solutions in this application will be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of this application, rather than all the embodiments of this application. It should be understood that this application is not limited to the example embodiments described herein. Based on the embodiments of this application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application. It should also be noted that, for the convenience of description, only the parts related to this application, rather than all of them, are shown in the accompanying drawings.

[0022] For examples, please see the attached Figure 1 The present application provides a method for detecting appearance defects of an automobile steering wheel based on machine vision, wherein the method for detecting appearance defects of an automobile steering wheel based on machine vision specifically comprises the following steps:

[0023] S100: Activate the ultrasonic probe to perform high-frequency sound wave detection on the target steering wheel and obtain and record the echo signal.

[0024] Specifically, an appropriate ultrasonic probe is selected and activated, emitting high-frequency sound waves into the target steering wheel. Commonly used ultrasonic probe tools, such as inspection-type ultrasonic flaw detectors like the Panametrics 5800, transmit and receive ultrasonic signals within the object being inspected. The appropriate frequency and probe type are set based on the material and size of the object being measured to ensure test accuracy and effectiveness. An ultrasonic probe is an instrument that utilizes ultrasonic technology for testing. It emits high-frequency sound waves (typically ranging from tens of kilohertz to several megahertz) into an object and receives echo signals reflected from the object's interior or surface. It is typically used to inspect the internal structure of materials and can identify defects such as cracks, bubbles, and holes.

[0025] The high-frequency sound wave signal emitted by the ultrasonic probe penetrates the surface of the steering wheel and produces different reflected echoes depending on the internal structure or defects (such as cracks, bubbles, and deformation). For example, if the target steering wheel is made of plastic and there is a tiny crack in it, the ultrasonic wave will reflect at the crack and return to the probe. If the crack is small enough, the echo signal will be weaker than the echo signal from a normal area, and the reflection time will also vary. By analyzing the changes in the echo signal, the location and size of the crack can be determined.

[0026] After the ultrasonic signal is transmitted and reflected, the echo signal is automatically recorded. Ultrasonic detection can detect defects inside or on the surface of the steering wheel, particularly hidden defects such as tiny cracks, bubbles, and material unevenness. By analyzing the timing and waveform changes of the echo signal, internal defects can be effectively distinguished from normal areas, providing important data support for subsequent inspections (such as laser scanning and 3D modeling), improving the accuracy and efficiency of overall defect detection.

[0027] S200: Curve processing is performed on the time sequence of the first echo signal of the first detection point in the echo signal record to obtain a first echo spline curve.

[0028] Specifically, the echo signal record is the acoustic signal reflected from the surface of the target object (such as the steering wheel) back to the probe after high-frequency acoustic wave testing of the target steering wheel. This includes the acoustic wave emitted by the probe and the echo data reflected from the target object. It is usually stored in a time series, recording the change in echo intensity over time. The first detection point refers to a specific location that the ultrasonic probe focuses on during the inspection process. Different detection points can represent different locations on the target object's surface, potentially involving different surface characteristics and defects. The first echo signal time series associated with the first detection point is obtained from the ultrasonic inspection.

[0029] The first echo signal timing is processed to form a curve, and the data in the echo signal timing is processed to form a smooth curve. Before this processing, the first echo signal timing must be calibrated. A preset ultrasonic influence structure diagram is constructed using historical test conditions and their corresponding test accuracy in the historical ultrasonic test database. An ultrasonic influence factor group is then constructed based on this preset ultrasonic influence structure diagram, encompassing all factors that influence ultrasonic test results (such as temperature, humidity, probe frequency, steering wheel material, etc.). The high-frequency acoustic wave test process is monitored using this ultrasonic influence factor group, and the parameter values ​​of all influencing factors are collected. These real-time monitoring parameters are mapped to the preset ultrasonic influence structure diagram, reflecting the impact of real-time changes in each influencing factor on the ultrasonic test results during the actual test process. Based on the preset ultrasonic influence structure diagram, which maps the real-time monitoring parameters, the first echo signal timing is calibrated to ensure signal accuracy, reduce interference or errors, and thus improve test accuracy.

[0030] By selecting an appropriate mathematical model (such as polynomial regression or spline curve) to fit discrete data points, discrete echo signal data points are connected into a smooth curve. Fitting methods are typically used to smooth the data. A scatter plot is drawn based on the time series of the first echo signal, with each data point represented as a scatter point on the graph. Polynomial regression is performed on the time series data of the first echo signal, using a polynomial model to fit the data points in the scatter plot. Based on the regression order, an appropriate polynomial is selected to fit the signal. The regression order (such as quadratic or cubic polynomial) is selected. A low order may not accurately capture signal changes, while a high order may lead to overfitting. Using the regression formula, the echo signal values ​​at different time points are calculated and the echo fitting curve is plotted, resulting in a smooth echo signal trend line.

[0031] The echo fitting curve is further processed using a spline curve. An appropriate spline type (such as a cubic spline) is selected, and a piecewise polynomial is used to fit the data, ensuring continuity and smoothness at each node. Spline curves effectively avoid the overfitting problem that occurs with polynomial regression, especially when there is significant fluctuation between data points. Through polynomial regression and spline curve fitting, the echo signal is smoothed, removing noise and irregular fluctuations, making the signal trend more clear. Regression and spline curve fitting effectively reduce the interference of noise on signal analysis.

[0032] S300: Determine whether a first curve deviation obtained by comparing the first echo spline curve with a preset spline curve meets a curve deviation threshold.

[0033] Specifically, a pre-defined spline curve is pre-constructed based on known normal conditions or reference data. It represents the expected trends or characteristics of an echo signal under normal circumstances and is generated based on theoretical models, historical data, or ideal test results. The deviation between the first echo spline curve and the pre-defined spline curve can be calculated by comparing the echo signal strength values ​​at each time point to calculate the difference between the two curves, or by calculating the average deviation, maximum deviation, or root mean square error (RMSE) over a period of time to assess the overall deviation. The calculated first curve deviation is then compared with a curve deviation threshold to determine whether the deviation is within an acceptable range. The curve deviation threshold is a preset tolerance used to determine whether the deviation between the two curves is within an acceptable range. If the deviation meets the threshold, the echo signal performance meets expectations, and further testing and analysis can proceed. If the deviation does not meet the threshold, it indicates an anomaly in the echo signal, possibly indicating a defect (such as a surface crack or pore), requiring further testing or remedial measures, such as using more precise methods such as laser scanning and 3D modeling to locate the defect. By comparing curve deviations, potential defects can be identified at the moment when the signal changes significantly, reducing the subjectivity of manual judgment, making the detection process more automated and less susceptible to the limitations of operator experience.

[0034] S400: If the first curve deviation does not meet the curve deviation threshold, activate the laser transmitter to perform laser scanning detection on the first detection point to obtain a first laser record.

[0035] Specifically, if the first curve deviation does not meet the curve deviation threshold, it indicates an abnormality in the echo signal. The laser transmitter is automatically activated to perform laser scanning at the first detection point, accurately measuring subtle changes and defects on the target steering wheel surface (or other object surface). A laser transmitter is a device that emits a laser beam and is typically used for non-contact surface measurement and scanning. In defect detection, lasers can be used to accurately measure the shape, displacement, or defects of a target surface. The laser transmitter emits a laser beam toward the target steering wheel surface. The laser beam hits the target surface and reflects back. A laser receiver captures the reflected light signal and calculates the three-dimensional coordinates of each surface point. The laser scanning results are recorded to generate a first laser record, which includes the three-dimensional coordinates of the surface points, surface morphology, depth information, and more. Laser scanning provides highly accurate three-dimensional surface data, capable of detecting subtle defects (such as microcracks, dents, or irregular surface deformations) that ultrasonic echo signals cannot detect. Ultrasonic waves are suitable for detecting deep or large defects, while laser scanners are used to accurately detect surface defects. Combining the two effectively improves the comprehensiveness and accuracy of defect detection.

[0036] S500: Performing local three-dimensional modeling on the first detection point of the target steering wheel based on the first laser recording to obtain a first local point cloud model.

[0037] Specifically, after the laser scan, the first laser record is calibrated and adjusted to eliminate measurement errors caused by environmental influences, surface reflectivity differences, laser emission angle, and other factors. This calibration results in a more accurate first laser record, containing corrected data about the target surface obtained by the laser scan. Based on the first laser record, a local 3D model is generated for the first detection point of the target steering wheel. The first detection point is a specific location during the laser scan, corresponding to data for a specific area on the steering wheel surface. This can be a center point, an edge point, or a uniquely shaped area.

[0038] Extract point cloud data around the first detection point from the first laser recording, and use 3D modeling technology to process these points to generate a 3D model. By connecting adjacent points in the point cloud, a triangular mesh is generated to construct a surface model of the target area. Use algorithms (such as the least squares method) to fit the point cloud data to generate smooth surfaces or geometric shapes. Convert the point cloud data into a uniform voxel grid model to represent objects in three-dimensional space. Local 3D modeling refers to the construction of a 3D geometric model of a local area of ​​the target object based on the point cloud data obtained by laser scanning. Usually, a 3D geometric model is a point cloud composed of several points. These points are connected or fitted through a certain algorithm to form a 3D surface.

[0039] After local 3D modeling, the first local point cloud model is obtained. This model is composed of a large amount of point cloud data, with each point having a 3D coordinate representing its position in space. A point cloud model, composed of a series of discrete 3D points, typically represents the surface morphology of the target object. The 3D points are then used to further analyze the structure, surface flatness, and possible defects of the target object. Local 3D modeling allows for precise reconstruction of the surface morphology of the target steering wheel, faithfully reflecting the geometric features and surface structure of the target object, allowing for rapid identification of abnormal areas on the steering wheel surface and accurate positioning.

[0040] S600: Compare the first local point cloud model with the target preset point cloud model of the target steering wheel to obtain a first point cloud deviation set.

[0041] Specifically, the first local point cloud model is compared with a target preset point cloud model of the target steering wheel to identify and quantify differences between the two. The target preset point cloud model of the target steering wheel is a pre-defined 3D point cloud model of the target steering wheel, based on design or quality standards. It represents the ideal surface geometry and expected dimensions, shape, and other features of the steering wheel. It is typically a standard model generated using computer-aided design (CAD) tools. The preset point cloud model can be generated through measurement and calculation, is generally unaffected by errors in the actual manufacturing process, and represents a perfect steering wheel shape, representing the ideal geometry and surface features of the target steering wheel.

[0042] The corresponding points in the two models are compared through the point cloud processing algorithm, and the distance between each point in the first local point cloud model and the nearest point in the target preset point cloud model is calculated. The iterative closest point algorithm is used to align the two point clouds, repeatedly looking for the nearest point of each point in one point cloud in the other point cloud, and then continuously optimizing the transformation parameters between the point clouds until the minimum error is achieved. After the point cloud comparison, the first point cloud deviation set is generated, which records the spatial position of all points with differences and their deviation values. The first point cloud deviation set contains the deviation values ​​between all points, indicating the deviations in the local area due to manufacturing errors, material properties, surface unevenness, etc. Each deviation point usually represents the difference in spatial position between the actual surface and the ideal model. By obtaining the point cloud deviation set, the defects on the steering wheel surface are quantitatively analyzed to help determine the size, shape and location of the defects.

[0043] S700: Analyze the first point cloud deviation set to obtain target appearance defects of the target steering wheel.

[0044] Specifically, the first point cloud deviation set is analyzed to identify cosmetic defects on the steering wheel surface, including poor roundness, warping, surface unevenness, cracks, or breakage. Poor roundness refers to a deviation in the circular portion of the steering wheel, which may appear as one part protruding or recessed relative to the rest. By comparing the deviations of each point in the deviation set with the ideal circular model, the roundness error can be calculated. By comparing the point cloud data with the ideal circular model, a circle fitting algorithm (such as the least squares method) is used to fit the point cloud to a circle, and the fitting error is calculated. If the deviation exceeds a certain threshold, a poor roundness defect is considered to exist.

[0045] Warping refers to an uneven steering wheel surface with curves or bulges. Data analysis of point cloud deviations reveals large deviations in certain areas of the surface, manifesting as consistent shape changes (such as curvature or bulges). Using surface fitting algorithms, such as least squares plane fitting or B-spline surface fitting, the deviation between the actual point cloud and the fitted plane is calculated. If the deviation values ​​show a regular pattern (such as a gradual increase from the center outward), warping may be present.

[0046] An uneven surface refers to the presence of local depressions or ridges on the surface of the steering wheel. In certain areas where point cloud deviations are concentrated, the deviation values ​​may be significantly higher than in other areas, indicating surface irregularities. By analyzing the local curvature or discreteness of the point cloud, check whether there are obvious depressions or ridges in the local area. The local curvature calculation method can be used to determine whether the surface is flat. Cracks or damage usually appear as obvious discontinuities on the steering wheel surface. Some points in the point cloud data may show obvious missing or morphological mutations. Analyze sparse or discontinuous areas in the point cloud data to determine whether there are cracks or surface missing. By calculating the sharp changes in the deviation value, possible fractures or damaged areas can be identified.

[0047] Based on the analysis of the first point cloud deviation set, a defect report is generated, documenting the defect type, location, severity, and repair recommendations. This comprehensive analysis of the point cloud deviation set enables the detection of defects across multiple dimensions, such as shape, surface quality, and size. This significantly reduces the workload of manual inspections, reduces human error, and improves inspection efficiency. Point cloud data analysis identifies subtle defects imperceptible to traditional inspection methods, providing higher inspection accuracy and helping to detect subtle cosmetic issues such as warpage, poor roundness, or surface inequality.

[0048] Further, as attached Figure 2 As shown, this application S200 includes:

[0049] S210: Obtain a preset influence structure diagram, which includes a preset ultrasonic influence structure diagram; S220: Establish an ultrasonic influence factor group based on the preset ultrasonic influence structure diagram; S230: Perform dynamic monitoring of high-frequency sound wave detection based on the ultrasonic influence factor group to obtain an ultrasonic influence factor parameter set; S240: Render the ultrasonic influence factor parameter set to the preset ultrasonic influence structure diagram to obtain a dynamic ultrasonic influence structure diagram; S250: Calibrate and adjust the first echo signal timing according to the dynamic ultrasonic influence structure diagram.

[0050] Specifically, a preset influence structure diagram is obtained. This is a pre-defined graphical structure that contains various conditions and factors that affect ultrasonic testing accuracy. A preset influence structure diagram contains the relationships between these factors, including a preset ultrasonic influence structure diagram, which is a specific type of preset influence structure diagram specifically used to describe the impact of various factors (such as the steering wheel, probe, and environment) on testing accuracy during the ultrasonic testing process.

[0051] Based on the historical ultrasonic testing database, multiple testing conditions and their corresponding multiple testing accuracies are determined. A data set is randomly selected, and the corresponding first historical testing condition information and first historical testing accuracy are extracted. A correlation analysis is performed on these first historical testing condition information and the first historical testing accuracy to determine the degree of influence of each condition indicator in the first historical testing condition information on the testing accuracy, i.e., the correlation coefficient. An initial influence structure diagram is constructed for each of the first historical testing condition information and the first historical testing accuracy. Similarly, corresponding influence structure diagrams are constructed for each of the multiple testing conditions and their corresponding testing accuracy, ultimately completing the construction of the preset ultrasonic influence structure diagram, which displays the various factors influencing ultrasonic testing and the relationships between them.

[0052] Based on a pre-set ultrasonic influence structure diagram, an ultrasonic influence factor group is constructed, comprising all conditions that influence ultrasonic signal detection results, such as physical conditions (e.g., temperature, humidity), equipment settings (e.g., probe frequency, probe angle), and target object characteristics (e.g., steering wheel material, thickness). Dynamic monitoring of high-frequency ultrasonic testing is performed based on this ultrasonic influence factor group, with real-time monitoring and recording of various factors (e.g., probe frequency, steering wheel material, etc.) during the ultrasonic testing process. During the high-frequency ultrasonic testing process, changes in various influencing factors, such as environmental conditions and equipment settings, are monitored in real time, and their parameter sets are recorded to create the ultrasonic influence factor parameter set, including temperature, humidity, probe frequency, and acoustic reflectivity.

[0053] Render the recorded ultrasonic impact factor parameter set onto a preset ultrasonic impact structure diagram to create a dynamic ultrasonic impact structure diagram. Based on the impact factor parameters collected in real time, the rendered ultrasonic impact factor parameter set is mapped to the corresponding position in the preset impact structure diagram. For example, if the probe frequency changes, the corresponding frequency portion of the impact structure diagram will dynamically update to display the new impact length or direction. Through rendering, the dynamic ultrasonic impact structure diagram can reflect the current test conditions and their impact on the ultrasonic signal in real time.

[0054] The timing of the first echo signal is calibrated based on the dynamic ultrasonic influence structure diagram. The timing of the echo signal is affected by many factors, such as ambient temperature, humidity changes, or probe settings. If these conditions change, the timing of the echo signal may be offset, thereby affecting the detection accuracy. First, the dynamic ultrasonic influence structure diagram is converted into a matrix form, and the data therein is converted into a vector form through vectorization processing. Based on each dynamic ultrasonic influence vector, the dynamic vector eigenvalue is extracted, reflecting the weighted effect and importance of the influencing factor, which is used as the weight. The timing of the first echo signal is calibrated and adjusted according to the weight, and finally a more accurate echo signal timing is obtained. By rendering the dynamic ultrasonic influence structure diagram, the real-time impact of each influencing factor on the detection process is determined. The real-time calibration and adjustment of the first echo signal timing reduces the signal offset caused by environmental and equipment changes, ensures the accuracy and consistency of the echo signal, and thus improves the detection accuracy.

[0055] Furthermore, the present application further comprises the following steps:

[0056] S211: Acquire a historical ultrasonic detection database; S212: Extract a first historical detection data group from the historical ultrasonic detection database, wherein the first historical detection data group includes first historical detection condition information and first historical detection accuracy; S213: Perform a correlation analysis on the first historical detection condition information and the first historical detection accuracy to obtain a first correlation result; S214: Construct the preset ultrasonic influence structure diagram based on the first correlation result.

[0057] S2131: The first historical detection condition information includes first historical steering wheel condition information, first historical probe condition information, and first historical environmental condition information.

[0058] Specifically, a historical ultrasonic testing database is obtained from ultrasonic testing equipment, including test records from different batches of steering wheels during the production process, or records provided by historical testing and testing instruments. These records typically include the conditions used for each test (e.g., probe type, test environment, specific state of the steering wheel) and the corresponding test accuracy. A data set is randomly selected from the historical ultrasonic testing database as a first historical test data set, including first historical test condition information and first historical test accuracy.

[0059] The first historical detection condition information includes various conditions that affect the accuracy and results of ultrasonic testing, such as first historical steering wheel condition information, first historical probe condition information, and first historical environmental condition information. The first historical steering wheel condition information refers to information related to the physical properties of the steering wheel, such as its material, shape, and surface condition. The first historical probe condition information refers to information related to the ultrasonic probe itself, such as its frequency, model, and sensitivity. The first historical environmental condition information refers to external environmental conditions during the testing process, such as temperature, humidity, noise, and lighting.

[0060] Correlation analysis, such as the Pearson correlation coefficient and regression analysis, is performed between the first historical detection condition information and the first historical detection accuracy to determine the impact of different detection conditions (such as the environment, probe, and steering wheel characteristics) on detection accuracy. Taking the Pearson correlation coefficient analysis as an example, the correlation coefficient between each pair of variables is calculated, that is, the correlation between each historical detection condition information and detection accuracy. The result ranges from -1 to +1, with +1 indicating a perfect positive correlation, -1 indicating a perfect negative correlation, and 0 indicating no correlation. The first correlation result is derived based on the specific relationship between each detection condition (steering wheel material, probe frequency, ambient temperature, etc.) and detection accuracy, as well as the degree of influence of each variable on detection accuracy.

[0061] Based on each condition indicator and its corresponding correlation coefficient in the first historical detection condition information, a predetermined rendering strategy is introduced. Using the condition indicator as the impact endpoint and the correlation coefficient as the impact length, the impact structures corresponding to each condition indicator are obtained and added to the initial impact structure diagram. This ultimately completes the construction of the preset ultrasonic impact structure diagram, which includes the degree of influence of each condition indicator in the first historical detection condition information on detection accuracy. By deeply analyzing the impact of different detection conditions on detection accuracy, the most critical factors for ultrasonic detection are clarified, allowing targeted adjustments and optimizations to be made during actual testing, precisely adjusting the detection process to ensure maximum detection accuracy.

[0062] Furthermore, the present application further comprises the following steps:

[0063] S2141: Obtain any condition indicator; S2142: Match any correlation coefficient corresponding to the any condition indicator in the first correlation result; S2143: When the any correlation coefficient reaches a correlation threshold, call a predetermined rendering strategy to render the any condition indicator to the preset ultrasonic impact structure diagram.

[0064] Specifically, any conditional indicator to be analyzed is determined, which is a specific condition or factor that may affect the test results during the ultrasonic detection process, such as the temperature and humidity of the detection environment, the material of the steering wheel, the frequency of the probe, etc. Different conditional indicators have different degrees of influence on the accuracy of ultrasonic detection. According to any conditional indicator, the corresponding correlation coefficient is matched in the first correlation result. When any correlation coefficient reaches the correlation threshold, the predetermined rendering strategy is called. The correlation threshold refers to the threshold set in the correlation analysis. When the correlation coefficient of a conditional indicator reaches or exceeds this threshold value, it is considered that the condition has a significant impact on the detection accuracy, and it is decided to treat it as an important factor for further analysis or visualization.

[0065] Taking any conditional indicator as the impact endpoint and the corresponding arbitrary correlation coefficient as the impact length, an arbitrary impact structure corresponding to the arbitrary conditional indicator is constructed according to a predetermined rendering strategy. The predetermined rendering strategy refers to the process of visualizing the results of the correlation analysis in the form of graphs, charts, etc. By dynamically displaying which conditions have a greater impact on detection accuracy based on the correlation coefficients of different conditions, and showing the relative importance of each factor. Rendering methods can take the form of color coding (such as red for high correlation and green for low correlation), size changes, graphics or annotations (adding annotations to the figure to clearly indicate the importance of certain conditions). The color, size, position and other parameters in the figure are dynamically adjusted according to the correlation coefficient of the conditional indicator to highlight the conditions that have a greater impact on detection accuracy.

[0066] Each condition indicator in the first historical test condition information is rendered accordingly, forming a corresponding influence structure diagram. Ultimately, a preset ultrasonic influence structure diagram is generated, which illustrates the relationship between the various influencing factors in ultrasonic testing and the correlation between each condition indicator and test accuracy. Through a preset rendering strategy, important influencing factors are automatically identified and displayed, reducing manual intervention and subjective judgment, and achieving a more intelligent testing process.

[0067] Furthermore, the present application further comprises the following steps:

[0068] S21431: According to the predetermined rendering strategy, the arbitrary condition indicator is used as the influence endpoint and the arbitrary correlation coefficient is used as the influence length; S21432: The arbitrary influence structure of the arbitrary condition indicator is obtained according to the influence endpoint and the influence length; S21433: The arbitrary influence structure is added to the initial influence structure diagram to obtain the preset ultrasonic influence structure diagram.

[0069] Specifically, according to a predetermined rendering strategy, any condition metric is used as an impact endpoint, and any correlation coefficient is used as an impact length. A predetermined rendering strategy refers to the rules and methods used to convert a detection condition and its correlation coefficient into a visual representation. It specifies how to use the impact endpoint (condition metric) and impact length (correlation coefficient) to create a graphical representation. The impact endpoint represents the starting point of a specific condition, and the impact length is determined by the correlation coefficient obtained from the first correlation result.

[0070] Based on the obtained influence endpoints and influence lengths, an arbitrary influence structure corresponding to any conditional indicator is constructed. Its length is the absolute value of the correlation coefficient, and its direction indicates the direction of influence. The generated arbitrary influence structure is added to the initial influence result map to form the preset ultrasonic influence structure map. In other words, the above calculation is performed for all conditional indicators in the first historical detection condition information, and these are used as influence endpoints and their corresponding correlation coefficients as influence lengths to form the influence structure.

[0071] The preset ultrasonic influence structure diagram is the final graphic, continuously updated, rendered, and added to. It includes the impact of all conditions and correlation coefficients, providing an intuitive understanding of the impact of each factor on ultrasonic testing accuracy. The preset ultrasonic influence structure diagram includes each condition and its corresponding impact length, and displays each condition's positive or negative impact. Different color or shape markers are used to distinguish the impact strength of different conditions. The influence structure diagram directly identifies the conditions with the greatest impact on test accuracy. The influence structure diagram is continuously updated to maintain the latest test optimization strategies and automatically adjust to different test environments.

[0072] Furthermore, the present application further comprises the following steps:

[0073] S251: performing matrix vectorization processing on the dynamic ultrasonic influence structure diagram to obtain a dynamic ultrasonic influence vector; S252: using the dynamic vector eigenvalue of the dynamic ultrasonic influence vector as a weight, calibrating and adjusting the timing of the first echo signal.

[0074] Specifically, the dynamic ultrasonic influence structure diagram is matrix-vectorized, and the multiple influencing factors and their interrelationships in the diagram are expressed in numerical form. The dynamic ultrasonic influence structure diagram is converted into a matrix, and the data therein is converted into a vector form through vectorization. Specifically, each influencing factor in the dynamic ultrasonic influence structure diagram (such as probe frequency, temperature, humidity, etc.) and the relationship between the influencing factors (such as positive or negative correlation) are converted into a matrix. For example, assuming that the influencing factors include temperature, humidity, and probe frequency, these factors are encoded into a matrix form according to their influence, where each element in the matrix represents the contribution of a factor to the ultrasonic detection results.

[0075] The dynamic ultrasonic influence vector is a vector obtained by matrix-vectorizing the influencing factor data in the dynamic ultrasonic influence structure diagram. It contains the degree, direction (positive or negative correlation), and intensity of the impact of each factor (such as temperature, humidity, and probe frequency) on the ultrasonic signal. The dimension of the dynamic ultrasonic influence vector generally depends on the number of influencing factors.

[0076] The dynamic vector eigenvalue is extracted from the dynamic ultrasonic influence vector, which reflects the properties of the vector and is usually used to represent certain important statistics or the intensity of the influence on the signal. Calculating the eigenvalue of a vector usually involves the concepts of eigenvalue and eigenvector in linear algebra. Usually, a characteristic equation is constructed based on the matrix converted from the dynamic ultrasonic influence structure diagram, and a polynomial equation is solved to obtain the eigenvalue. For multiple eigenvalues ​​found, the eigenvector is calculated through a system of linear equations. Factors with larger eigenvalues ​​usually have a greater impact on the signal, while factors with smaller eigenvalues ​​have a smaller impact on the signal. The dynamic vector eigenvalue of the dynamic ultrasonic influence vector is used as a weight to assign to each influencing factor. Generally, factors with larger eigenvalues ​​correspond to higher weights, which means that the factor has a greater impact on the calibration of the echo signal timing. The larger the eigenvalue, the greater the weight, indicating that the factor has a stronger impact on the adjustment of the signal timing.

[0077] Based on the weights, the first echo signal timing is calibrated and adjusted, adjusting the signal according to the weights of each influencing factor. Because environmental conditions or detection settings during the detection process can affect signal propagation and reflection, calibration adjustments are necessary to reduce deviations and errors. The echo signal timing is adjusted to ensure greater accuracy and meet actual detection requirements. By performing matrix vectorization and eigenvalue analysis on the dynamic ultrasonic influence structure diagram, the impact of each influencing factor on the echo signal timing is quantified, allowing for efficient calibration adjustments to reduce errors and improve echo signal accuracy.

[0078] Furthermore, the present application further comprises the following steps:

[0079] S260: Acquire a first echo scatter plot of the first echo signal time series; S270: Perform polynomial regression fitting processing on the first echo scatter plot to obtain a first echo fitting formula; S280: Use a first echo fitting curve of the first echo fitting formula as the first echo spline curve.

[0080] Specifically, the first echo signal time series refers to the time series data of the echo signal associated with the first detection point during ultrasonic testing. It records important characteristics of the sound wave, such as propagation time and reflection intensity. The data in the echo signal time series is plotted as scattered points on a scatter plot, with time (e.g., milliseconds) on the horizontal axis and echo intensity (e.g., decibels) on the vertical axis. Each data point represents the intensity of the echo signal at a specific moment. A first echo scatter plot is created using the time and echo intensity data in the dataset. Scatter plots can help quickly identify whether there are clear patterns or trends in the signal. A scatter plot is a data visualization method used to show the relationship between two variables. A first echo scatter plot plots each signal data point in the first echo signal time series as a scattered point, with the horizontal axis representing time (or other relevant parameters) and the vertical axis representing the intensity or amplitude of the echo signal.

[0081] A polynomial regression fit is performed on the first echo scatter plot, using polynomial regression analysis to fit these scattered points. Polynomial regression is a statistical method that fits data using a polynomial function. The goal is to find the best-fit curve with the smallest fitting error. When fitting the data using a polynomial function, an appropriate polynomial order (such as a quadratic or cubic polynomial) is selected to minimize the error between the data points and the fitted curve. The goal of the curve fitting is to find an optimal trend line that represents the temporal variation of the echo signal. The selection of the polynomial regression order requires consideration of fitting accuracy and model complexity. If the order is too low, it may fail to capture signal variations; if it is too high, it may overfit, making the curve sensitive to noise. Therefore, cross-validation or other methods are often used to select the optimal order. The regression fitting process involves selecting an appropriate polynomial order, calculating the regression coefficients using an optimization algorithm such as the least squares method, and obtaining the fitted curve equation.

[0082] After polynomial regression fitting, the first echo fitting formula is obtained, which represents the variation pattern of the echo signal. Based on the first echo fitting formula, a first echo fitting curve is generated, representing the temporal variation trend of the echo signal. This first echo fitting curve is used as the first echo spline curve. A spline curve is a piecewise smooth polynomial function, typically used to fit a smooth curve through discrete data points. Using a spline curve, a more refined signal trend can be obtained. Especially when the signal is noisy or has a small number of data points, a spline curve can smooth the data, reduce volatility, and obtain a more stable trend prediction. Using polynomial regression and spline curve fitting, echo signal data can be effectively smoothed, noise can be eliminated, and a clearer signal variation trend can be obtained. The spline curve can better reflect the actual signal changes, rather than being interfered with by noise.

[0083] Furthermore, the present application S400 also includes:

[0084] S410: Extracting a preset laser influence structure diagram from the preset influence structure diagram; S420: Establishing a laser influence factor group according to the preset laser influence structure diagram; S430: Performing dynamic monitoring of laser scanning detection based on the laser influence factor group to obtain a laser influence factor parameter set; S440: Rendering the laser influence factor parameter set to the preset laser influence structure diagram to obtain a dynamic laser influence structure diagram; S450: Calibrating and adjusting the first laser record according to the dynamic laser influence structure diagram.

[0085] Specifically, the preset laser influence structure diagram in the preset influence structure diagram is extracted, and the preset influence structure diagram includes not only the preset ultrasonic influence structure diagram, but also the preset laser influence structure diagram. The preset laser influence structure diagram is a structure diagram specifically used to describe and analyze the factors affecting laser scanning detection, which includes factors that may affect the laser scanning detection results, such as laser emission angle, distance, surface reflectivity, etc., as well as the relationship between these factors. Furthermore, the laser scanning detection process is dynamically monitored with the laser influence factor group as a guide to obtain dynamic data of factors affecting the laser detection quality, that is, the laser influence factor parameter set is obtained. Then, the factor parameters monitored in real time are rendered and updated to the preset laser influence structure diagram, and the first laser record is calibrated and adjusted by rendering the updated dynamic laser influence structure diagram. Correct potential errors caused by the environment, equipment or external conditions during the laser scanning process.

[0086] Based on a preset laser impact structure diagram, factors that may affect the laser scanning results, such as the laser emission angle and the reflectivity of the target surface, are identified. Based on the identified influencing factors, the first laser recording is corrected. This process is logically similar to the aforementioned process of calibrating and adjusting the timing of the first echo signal using a preset ultrasonic impact structure diagram. The difference lies in ultrasonic detection and laser scanning, which will not be discussed in detail here. By considering the environmental factors affecting laser scanning, measurement errors can be corrected more accurately. Various interference factors during the laser scanning process (such as surface reflectivity and distance changes) may cause errors in the scanning results. Calibration reduces the impact of these factors, ensuring that the measurement data is closer to the actual situation.

[0087] In summary, the method for detecting appearance defects of automobile steering wheels based on machine vision provided by this application has the following technical effects:

[0088] The target steering wheel is subjected to high-frequency acoustic wave detection by activating an ultrasonic probe, and an echo signal record is obtained; the first echo signal time series of the first detection point in the echo signal record is curve-processed to obtain a first echo spline curve; it is determined whether the first curve deviation obtained by comparing the first echo spline curve with the preset spline curve meets the curve deviation threshold; if the first curve deviation does not meet the curve deviation threshold, the laser transmitter is activated to perform laser scanning detection on the first detection point to obtain a first laser record; based on the first laser record, local three-dimensional modeling is performed on the first detection point of the target steering wheel to obtain a first local point cloud model; the first local point cloud model is compared with the target preset point cloud model of the target steering wheel to obtain a first point cloud deviation set; the first point cloud deviation set is analyzed to obtain the target appearance defect of the target steering wheel. In other words, ultrasonic detection is used to perform high-frequency sound wave detection on the steering wheel, and the echo signal is processed into a curve and compared with the preset spline curve to determine whether there is a deviation. If there is a deviation, the laser emitter is activated for laser scanning, and local three-dimensional modeling is performed. It is compared with the target preset point cloud model, the point cloud deviation is analyzed, and the appearance defects are identified, thereby improving the accuracy of the appearance defect detection of the automobile steering wheel.

[0089] The above description of the disclosed embodiments is intended to enable one skilled in the art to implement or use the present application. Various modifications to these embodiments will be readily apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application is not limited to the embodiments shown herein, but is intended to conform to the widest scope consistent with the principles and novel features disclosed herein.

[0090] Obviously, those skilled in the art may make various changes and modifications to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the present application and its equivalents, the present application is intended to include these modifications and variations.

Claims

1. A method for detecting appearance defects of automobile steering wheels based on machine vision, characterized in that: include: Activate the ultrasonic probe to perform high-frequency sound wave detection on the target steering wheel and obtain and record the echo signal; Curve processing is performed on the time sequence of the first echo signal of the first detection point in the echo signal record to obtain a first echo spline curve; determining whether a first curve deviation obtained by comparing the first echo spline curve with a preset spline curve meets a curve deviation threshold; If the first curve deviation does not meet the curve deviation threshold, activating the laser transmitter to perform laser scanning detection on the first detection point to obtain a first laser record; Performing local three-dimensional modeling on the first detection point of the target steering wheel based on the first laser recording to obtain a first local point cloud model; Comparing the first local point cloud model with a target preset point cloud model of the target steering wheel to obtain a first point cloud deviation set; The first point cloud deviation set is analyzed to obtain target appearance defects of the target steering wheel.

2. The method for detecting appearance defects of automobile steering wheels based on machine vision according to claim 1, characterized in that: Before performing curve processing on the first echo signal time sequence of the first detection point in the echo signal record to obtain the first echo spline curve, the method further includes: Acquire a preset impact structure diagram, wherein the preset impact structure diagram includes a preset ultrasonic impact structure diagram; Establishing an ultrasonic impact factor group according to the preset ultrasonic impact structure diagram; Dynamic monitoring of high-frequency acoustic wave detection is performed based on the ultrasonic impact factor group to obtain an ultrasonic impact factor parameter set; Rendering the ultrasonic impact factor parameter set to the preset ultrasonic impact structure diagram to obtain a dynamic ultrasonic impact structure diagram; The first echo signal timing is calibrated and adjusted according to the dynamic ultrasonic influence structure diagram.

3. The method for detecting appearance defects of automobile steering wheels based on machine vision according to claim 2, characterized in that: Obtaining a preset impact structure diagram, wherein the preset impact structure diagram includes a preset ultrasonic impact structure diagram, including: Access to historical ultrasonic testing database; Extracting a first historical detection data group from the historical ultrasonic detection database, wherein the first historical detection data group includes first historical detection condition information and first historical detection accuracy; Performing a correlation analysis on the first historical detection condition information and the first historical detection accuracy to obtain a first correlation result; The preset ultrasonic influence structure diagram is constructed according to the first correlation result.

4. The method for detecting appearance defects of automobile steering wheels based on machine vision according to claim 3, characterized in that: The first historical detection condition information includes first historical steering wheel condition information, first historical probe condition information, and first historical environmental condition information.

5. The method for detecting appearance defects of automobile steering wheels based on machine vision according to claim 3, characterized in that: Constructing the preset ultrasonic influence structure diagram according to the first correlation result includes: Get any conditional indicator; Matching any correlation coefficient corresponding to the arbitrary condition indicator in the first correlation result; When the arbitrary correlation coefficient reaches a correlation threshold, a predetermined rendering strategy is called to render the arbitrary condition indicator into the preset ultrasonic impact structure diagram.

6. The method for detecting appearance defects of automobile steering wheels based on machine vision according to claim 5, characterized in that: Retrieving a predetermined rendering strategy to render the arbitrary condition indicator to the preset ultrasonic impact structure diagram includes: According to the predetermined rendering strategy, the arbitrary condition indicator is used as an impact endpoint and the arbitrary correlation coefficient is used as an impact length; Obtaining an arbitrary influence structure of the arbitrary condition indicator according to the influence endpoint and the influence length; The arbitrary influencing structure is added to the initial influencing structure diagram to obtain the preset ultrasonic influencing structure diagram.

7. The method for detecting appearance defects of automobile steering wheels based on machine vision according to claim 2, characterized in that: Calibrating and adjusting the timing of the first echo signal according to the dynamic ultrasonic influence structure diagram includes: Performing matrix vectorization processing on the dynamic ultrasonic influence structure diagram to obtain a dynamic ultrasonic influence vector; The first echo signal timing is calibrated and adjusted using the dynamic vector eigenvalue of the dynamic ultrasonic influence vector as a weight.

8. The method for detecting appearance defects of automobile steering wheels based on machine vision according to claim 1, characterized in that: include: Acquire a first echo scatter plot of the first echo signal time series; Performing polynomial regression fitting processing on the first echo scatter plot to obtain a first echo fitting formula; The first echo fitting curve of the first echo fitting formula is used as the first echo spline curve.

9. The method for detecting appearance defects of automobile steering wheels based on machine vision according to claim 2, characterized in that: After obtaining the preset impact structure diagram, it also includes: Extracting a preset laser influence structure diagram from the preset influence structure diagram; Establishing a laser impact factor group according to the preset laser impact structure diagram; Dynamic monitoring of laser scanning detection is performed based on the laser impact factor group to obtain a laser impact factor parameter set; Rendering the laser impact factor parameter set to the preset laser impact structure diagram to obtain a dynamic laser impact structure diagram; The first laser recording is calibrated and adjusted according to the dynamic laser influence structure diagram.

Images