Electric vehicle shock absorber defect detection method and system based on machine vision

Abstract

The invention relates to the technical field of shock absorber defect detection, and discloses an electric vehicle shock absorber defect detection method and system based on machine vision, and the method comprises the steps: collecting an initial image set under the irradiation of a multi-angle light source through high-resolution imaging collection equipment; performing denoising and filtering processing according to the initial image set to obtain clear image data; performing defect classification and spatial distribution analysis according to the clear image data to obtain surface feature vectors containing defect types and defect spatial distribution; carrying out vibration amplitude acquisition and phase angle measurement operation according to the surface feature vector, and carrying out spectral analysis to construct a performance parameter vector; performing data fusion according to the performance parameter vector and the surface feature vector to obtain a fusion feature; inputting the fusion features into a pre-constructed association prediction model to obtain defect prediction data; and performing defect influence degree analysis according to the defect prediction data to obtain a defect evaluation result. The method provides a basis for quality control and performance optimization of the shock absorber.

Description

Technical Field

[0001] The present invention relates to the technical field of shock absorber defect detection, and in particular to a method and system for detecting shock absorber defects of an electric vehicle based on machine vision. Background Art

[0002] Currently, electric vehicles, as a core form of new energy transportation, have a direct impact on driving safety and user experience. The structural integrity of the shock absorber is a key factor in determining the overall vehicle's dynamic balance. As the electric vehicle market expands and performance requirements rise, the limitations of traditional inspection methods are becoming increasingly prominent. Manual visual inspection is susceptible to subjective factors, making it difficult to standardize inspection standards. Single-parameter monitoring lacks in-depth analysis of defect root causes, making it impossible to accurately correlate defect type with dynamic balance performance fluctuations. A key challenge facing current technologies lies in the complex and diverse morphology of shock absorber surface defects (such as cracks and wear). The quantitative mapping between their visual characteristics and dynamic balance parameters remains unclear, and relying solely on images or single-dimensional data makes it difficult to assess the actual impact of defects on performance. Therefore, building a comprehensive inspection system that integrates surface image features and dynamic balance parameters, while overcoming the technical barriers of collaborative analysis of heterogeneous data and defect-performance correlation modeling, is urgently needed to achieve intelligent, high-precision defect detection for electric vehicle shock absorbers.

[0003] In one existing technology, the surface of the shock absorber is first inspected visually by hand. The inspector uses a magnifying glass or industrial endoscope to observe whether there are defects such as cracks, wear or deformation, and judges the severity of the defects based on experience. At the same time, the shock absorber is installed on a dynamic balancing test bench to collect single parameter data such as its vibration amplitude and frequency response under simulated load. During the inspection process, the location and shape of the surface defects are manually recorded and simply compared with the dynamic balancing test data. If the dynamic balancing parameters are found to exceed the preset threshold, the shock absorber is determined to have potential defects. However, this technology only relies on manual observation and static threshold judgment, and has not established an accurate correlation model between surface defect characteristics and dynamic balancing parameters. The final inspection results rely on the subjective experience of the inspectors and lack an automated data fusion and quantitative analysis mechanism.

[0004] Therefore, the existing technology cannot provide a basis for shock absorber quality control and performance optimization. Summary of the Invention

[0005] The present invention provides a method and system for detecting defects in electric vehicle shock absorbers based on machine vision, so as to provide a basis for shock absorber quality control and performance optimization.

[0006] In a first aspect, in order to solve the above technical problems, the present invention provides a method for detecting defects in electric vehicle shock absorbers based on machine vision, comprising: Collect an initial image set under multi-angle light source illumination by a high-resolution imaging acquisition device; Performing denoising filtering on the initial image set to obtain clear image data; Perform defect classification and spatial distribution analysis based on the clear image data to obtain a surface feature vector including defect type and defect spatial distribution; Performing vibration amplitude acquisition and phase angle measurement operations according to the surface feature vector, and performing spectrum analysis to construct a performance parameter vector; Performing data fusion according to the performance parameter vector and the surface feature vector to obtain a fusion feature; Inputting the fusion features into a pre-built association prediction model to obtain defect prediction data; A defect impact degree analysis is performed based on the defect prediction data to obtain a defect assessment result.

[0007] In an optional embodiment, performing denoising filtering on the initial image set to obtain clear image data includes: Based on the initial image set, pixel noise is suppressed by using a median filter. If salt and pepper noise interference is detected, a secondary smoothing denoising operation is performed by using a Gaussian filter to obtain a first preprocessed image; performing an overall brightness distribution remapping operation on the first preprocessed image by using an adaptive histogram equalization algorithm to obtain a second preprocessed image; performing a pixel gradient calculation operation using a Sobel edge detection operator on the second preprocessed image, identifying continuous high gradient sequences and marking them as suspected crack regions, thereby obtaining a third preprocessed image; According to the third preprocessed image set, a wear area contour extraction operation is performed through a morphological opening operation to highlight crack and wear features and obtain clear image data.

[0008] In an optional embodiment, the defect classification and spatial distribution analysis are performed based on the clear image data to obtain a surface feature vector including the defect type and the defect spatial distribution, including: Based on the clear image data, a pixel clustering operation is performed on the surface area of ​​the shock absorber using a watershed segmentation algorithm to obtain a pixel classification set; Extracting grayscale value variations according to the pixel classification set, and taking pixel classification sets whose grayscale value variations exceed a preset grayscale variation threshold as potential defect areas, to obtain a defect area set; Based on the defect area set, a continuous contour outlining operation is performed using the Canny edge detection algorithm to extract the crack boundary coordinates and wear texture shape, and morphological matching classification is performed in combination with a pre-stored defect template library to obtain a classification feature set including crack type and wear grade; Extract defect boundary points based on the classification feature set and perform curvature analysis, record the corresponding plane coordinates of the defect boundary points whose curvature exceeds a preset curvature threshold, and obtain the defect spatial distribution; According to the spatial distribution of the defects, the length and width of the defect area are quantified using a minimum circumscribed rectangle algorithm, and structured data is generated in combination with the classification feature set to obtain a surface feature vector including the defect type, defect location and defect size.

[0009] In an optional embodiment, the vibration amplitude acquisition and phase angle measurement operations are performed according to the surface feature vector, and spectrum analysis is performed to construct a performance parameter vector, including: Based on the surface feature vectors, the shock absorber is controlled to operate within a set speed range through a dynamic balancing test platform, an acceleration sensor is used to collect vibration amplitude, and a photoelectric encoder is used to perform phase angle measurement to obtain an initial signal set including speed, amplitude, and phase. Based on the initial signal set, a fast Fourier transform algorithm is used to perform time-frequency conversion to generate a spectrum diagram, and a power spectrum density analysis is used to extract a frequency response peak synchronized with the rotational speed. When the frequency response peak exceeds a preset frequency response threshold, an abnormal frequency band is marked, and a frequency response characteristic distribution diagram of the marked abnormal frequency band is obtained; According to the frequency response characteristic distribution diagram, the vibration waveform of the abnormal frequency band is nonlinearly time-aligned with a pre-stored fault template library using a dynamic time warping algorithm, and the waveform amplitude deviation is calculated to obtain vibration signal data including speed, frequency and waveform amplitude deviation; According to the vibration signal data, a time-frequency conversion operation is performed using a fast Fourier transform algorithm to generate a spectrum diagram, and the amplitude characteristics of the preset primary and secondary frequencies are extracted to obtain vibration signal characteristics; Based on the vibration signal characteristics, energy integration is performed on the frequency band exceeding the preset frequency threshold, and an initial performance vector is generated in combination with a preset dynamic balance weight coefficient; Based on the initial performance vector, feature dimension reduction is performed using a principal component analysis algorithm, and a performance parameter vector is extracted.

[0010] In an optional embodiment, performing data fusion according to the performance parameter vector and the surface feature vector to obtain a fusion feature includes: According to the crack boundary coordinates, the crack morphology similarity is calculated by using the Frechet distance algorithm to extract a set of geometric parameters; Extracting intrinsic fluctuation components according to the performance parameter vector and calculating the frequency band energy integral to generate dominant frequency characteristics; According to the geometric parameter set and the dominant frequency feature, weight coefficients are allocated by an entropy weight method, and weighted fusion is performed to obtain a fusion feature. In an optional embodiment, the process of constructing the association prediction model includes: Obtain historical fusion features and corresponding historical defect data; The support vector machine algorithm is used to build the initial association prediction model, the classification boundary is initialized by the RBF kernel function, and the penalty coefficient is set to determine the tolerance error threshold; Based on the historical fusion features, the kernel density estimation method is used to calculate the curvature adjustment parameter of the classification boundary, and the curvature of the classification boundary is dynamically adjusted through the RBF kernel function; Optimizing the penalty coefficient by a grid search method based on the historical defect data to determine an error tolerance threshold; Perform cross-validation on the adjusted classification boundaries and calculate the classification accuracy and error distribution set after boundary adjustment; When the error distribution set exceeds the error tolerance threshold, the curvature parameter of the classification boundary is re-optimized using the gradient descent method, and the classification boundary is iteratively updated until the error distribution set does not exceed the error tolerance threshold; The optimized classification boundary is matched with the defect data in multi-dimensional space through RBF kernel function mapping to determine the hyperplane equation corresponding to each defect data; Combined with the time series characteristics of the historical defect data, the sliding window method is used to extract the dynamic trend parameters of the defect data changes, and the dynamic trend parameters are associated with the hyperplane equation to construct the final association prediction model.

[0011] In an optional embodiment, performing defect impact analysis based on the defect prediction data to obtain a defect assessment result includes: Performing a matching operation on the defect prediction data through a pre-stored historical defect-performance degradation mapping library to obtain a baseline performance degradation reference value; Wherein, the defect prediction data includes defect type code, location coordinate set, size parameter vector and performance interference coefficient; Based on the performance interference coefficient and the reference value of the benchmark performance attenuation, a finite element stress simulation analysis is performed in combination with the pre-stored hardness of the shock absorber material to obtain the stress in the defect area and the damping performance attenuation; Based on the stress in the defect area and the damping performance attenuation, a risk level determination is performed. When the stress in the defect area exceeds the preset material yield strength threshold or the damping performance attenuation exceeds the preset dynamic balance tolerance threshold, a high-risk mark is output, and finally a defect assessment result including the defect location, size, risk level, and damping performance attenuation is generated.

[0012] In a second aspect, the present invention provides an electric vehicle shock absorber defect detection system based on machine vision, comprising: A data acquisition module is used to acquire an initial image set under multi-angle light source illumination through a high-resolution imaging acquisition device; An image filtering module, configured to perform denoising filtering on the initial image set to obtain clear image data; A surface feature module, configured to perform defect classification and spatial distribution analysis based on the clear image data to obtain a surface feature vector including defect type and defect spatial distribution; A performance parameter module is used to collect vibration amplitude and phase angle measurements based on the surface feature vector, and to perform spectrum analysis to construct a performance parameter vector; A feature fusion module, configured to perform data fusion based on the performance parameter vector and the surface feature vector to obtain a fusion feature; A defect prediction module, configured to input the fusion features into a pre-built association prediction model to obtain defect prediction data; A defect assessment module is configured to perform defect impact analysis based on the defect prediction data to obtain a defect assessment result. In a third aspect, the present invention further provides an electronic device comprising a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein when the processor executes the computer program, it implements any of the above-described methods for detecting electric vehicle shock absorbers based on machine vision.

[0013] In a fourth aspect, the present invention also provides a computer-readable storage medium, which includes a stored computer program, wherein when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute any one of the above-mentioned electric vehicle shock absorber defect detection methods based on machine vision.

[0014] Compared with the prior art, the present invention has the following beneficial effects: (1) By adopting the weighted fusion of surface feature vector and performance parameter vector (vibration spectrum / energy ratio), the weight coefficients of geometric features and frequency features are assigned by entropy weight method to solve the problem of insufficient performance degradation correlation analysis in traditional pure visual inspection, realize cross-domain correlation analysis between surface defects and dynamic performance, and improve the accuracy of defect cause determination.

[0015] (2) The classification boundary curvature is dynamically adjusted by the RBF kernel function, and the penalty coefficient is optimized by the grid search method and the parameters are iteratively updated by the gradient descent method. This breaks through the limitations of the traditional support vector machine in the classification rigidity of time-varying vibration data, realizes the autonomous adaptation of the defect prediction model to the real-time vibration signal characteristics, and reduces the misjudgment rate under complex working conditions.

[0016] (3) Based on the Frechet distance algorithm, the similarity of crack morphology is quantified, and the energy proportion of abnormal frequency bands is extracted by combined power spectrum density analysis. The fault waveform template is matched by the dynamic time warping algorithm to solve the problem that traditional methods are insufficient in mining the correlation between defect geometric characteristics and vibration response, and establish a full-dimensional defect assessment system across the visual-vibration domain.

[0017] (4) Through finite element stress simulation, the stress distribution in the defect area is quantified. Combining the historical defect-performance degradation mapping library with the dynamic balance tolerance threshold, the material yield strength is used as the high-risk judgment benchmark to solve the problem of inaccurate prediction of implicit performance degradation by traditional manual sampling, and realize a closed-loop decision chain from defect identification to service risk warning. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 1 is a flow chart of a method for detecting defects in an electric vehicle shock absorber based on machine vision provided by the first embodiment of the present invention; Figure 2 It is a structural diagram of an electric vehicle shock absorber defect detection system based on machine vision provided by the second embodiment of the present invention. DETAILED DESCRIPTION

[0019] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0020] Reference Figure 1 The first embodiment of the present invention provides a method for detecting defects in an electric vehicle shock absorber based on machine vision, comprising the following steps: S11, collecting an initial image set under multi-angle light source illumination by a high-resolution imaging acquisition device; S12, performing denoising filtering processing on the initial image set to obtain clear image data; S13, performing defect classification and spatial distribution analysis based on the clear image data to obtain a surface feature vector including defect type and defect spatial distribution; S14, performing vibration amplitude acquisition and phase angle measurement operations based on the surface feature vector, and performing spectrum analysis to construct a performance parameter vector; S15, performing data fusion based on the performance parameter vector and the surface feature vector to obtain a fusion feature; S16, inputting the fusion features into a pre-built association prediction model to obtain defect prediction data; S17, performing defect impact analysis based on the defect prediction data to obtain a defect assessment result.

[0021] In step S11, a high-resolution imaging acquisition device is used to acquire an initial set of images illuminated by multi-angle light sources.

[0022] Specifically, an industrial-grade high-resolution CMOS camera is used as the imaging device, in conjunction with a programmable annular light source array, which includes a top light source for vertical illumination, a side light source for oblique illumination, and a bottom light source for penetrating illumination. The shock absorber sample is fixed on a three-dimensional electric rotation stage, and the controller synchronizes the rotation stage step angle and the light source switching timing. The rotation stage rotates the shock absorber at a fixed step angle (for example, 30°), and triggers the light source switching every time it rotates to a preset specific angular position (for example, 0°, 120°, 240°). Each angular position activates three light source modes in turn: the top light source is used for vertical illumination to obtain surface morphological features; the side light source is used for oblique incidence to enhance texture contrast (for example, crack shadows); and the bottom light source penetrates the material to detect internal structural anomalies (for example, pore defects). The camera captures a single frame image in each light source mode, and a single rotation cycle generates multiple original images to form an initial image set. The camera exposure time is dynamically adjusted based on the reflective properties of the material (for example, the average grayscale value of the image is monitored in real time through a light feedback mechanism. When the grayscale value is above a preset upper threshold (for example, >200), the exposure time is automatically shortened (for example, to 20ms); when the grayscale value is below a preset lower threshold (for example, <50), the exposure time is automatically extended (for example, to 80ms) to maintain the image grayscale in a stable range (for example, 60-180)). The light source intensity threshold is determined through a pre-calibration process: based on a standard defect-free sample, a specific signal-to-noise ratio level (for example, ≥40dB) is used as the illumination benchmark, and the light source power is iteratively adjusted until the signal-to-noise ratio requirement is met. Multi-angle collaborative imaging addresses the detection limitations of a single light source. For example, side light can highlight surface microstructure features (such as cracks) within a specific depth range (for example, depth ≥0.1mm), and penetrating light can identify internal anomalies (such as pores) of a specific size (for example, diameter ≥0.5mm).

[0023] In step S12, denoising filtering is performed on the initial image set to obtain clear image data.

[0024] In a specific embodiment, performing denoising filtering on the initial image set to obtain clear image data includes: Based on the initial image set, pixel noise is suppressed by using a median filter. If salt and pepper noise interference is detected, a secondary smoothing denoising operation is performed by using a Gaussian filter to obtain a first preprocessed image; performing an overall brightness distribution remapping operation on the first preprocessed image by using an adaptive histogram equalization algorithm to obtain a second preprocessed image; performing a pixel gradient calculation operation using a Sobel edge detection operator on the second preprocessed image, identifying continuous high gradient sequences and marking them as suspected crack regions, thereby obtaining a third preprocessed image; According to the third preprocessed image set, a wear area contour extraction operation is performed through a morphological opening operation to highlight crack and wear features and obtain clear image data.

[0025] Specifically, a median filter is first applied to the initial image set: a 3×3 pixel window is used to traverse each image, the grayscale values ​​of the pixels within the window are sorted, and the median is used to replace the central pixel value to suppress impulse noise (typically black and white noise). If the grayscale standard deviation of a certain image region exceeds a preset threshold (determined by historical noise sample statistics), it is determined that salt and pepper noise interference is present. An additional Gaussian filter is then applied: a 5×5 Gaussian convolution kernel is used to weight the average pixel values ​​(with the weights within the kernel following a normal distribution) to achieve quadratic smoothing and denoising, and the first preprocessed image is output.

[0026] Based on the first preprocessed image, adaptive histogram equalization is performed: the image is divided into several local blocks (the block size is dynamically set according to the image resolution), the grayscale histogram is independently calculated for each block, and the cumulative distribution function mapping is performed to eliminate the effects of uneven lighting (such as overexposed areas caused by local reflections), make the overall brightness distribution uniform, and output the second preprocessed image.

[0027] The second preprocessed image is processed using the Sobel edge detection operator: the transverse gradient value Gx (horizontal convolution kernel) and the longitudinal gradient value Gy (vertical convolution kernel) are calculated. The gradient values ​​in these two directions are squared for each pixel, added together, and the square root of the sum is taken. This calculation process generates the pixel gradient magnitude for each pixel. The entire image is scanned to identify continuous high-gradient sequences (gradient magnitude exceeding an empirical threshold and the number of consecutive pixels ≥5). These sequences are marked as suspected crack areas (e.g., linear high-gradient bands), and the third preprocessed image is output.

[0028] Finally, a morphological opening operation is performed on the third preprocessed image set: a circular structuring element (with a radius determined by the minimum defect size) is first eroded to eliminate small noise points, and then dilated to restore the outline of the valid area. This operation focuses on extracting features of worn areas (such as irregular depressions) and enhancing crack continuity (such as repairing fracture edges), resulting in clear image data that highlights the defect features.

[0029] In step S13, defect classification and spatial distribution analysis are performed based on the clear image data to obtain a surface feature vector including defect type and defect spatial distribution.

[0030] In a specific embodiment, the defect classification and spatial distribution analysis are performed based on the clear image data to obtain a surface feature vector including the defect type and the defect spatial distribution, including: Based on the clear image data, a pixel clustering operation is performed on the surface area of ​​the shock absorber using a watershed segmentation algorithm to obtain a pixel classification set; Extracting grayscale value variations according to the pixel classification set, and taking pixel classification sets whose grayscale value variations exceed a preset grayscale variation threshold as potential defect areas, to obtain a defect area set; Based on the defect area set, a continuous contour outlining operation is performed using the Canny edge detection algorithm to extract the crack boundary coordinates and wear texture shape, and morphological matching classification is performed in combination with a pre-stored defect template library to obtain a classification feature set including crack type and wear grade; Extract defect boundary points based on the classification feature set and perform curvature analysis, record the corresponding plane coordinates of the defect boundary points whose curvature exceeds a preset curvature threshold, and obtain the defect spatial distribution; According to the spatial distribution of the defects, the length and width of the defect area are quantified using a minimum circumscribed rectangle algorithm, and structured data is generated in combination with the classification feature set to obtain a surface feature vector including the defect type, defect location and defect size.

[0031] Specifically, the watershed segmentation algorithm, using clear image data as input, clusters pixels on the shock absorber surface. This algorithm characterizes the intensity of brightness changes by calculating the gradient of each pixel in the image. Specifically, the Sobel operator is used to calculate the horizontal and vertical gradient components of each pixel. For the horizontal gradient, the brightness of the pixel and its surrounding neighborhood (typically a 3x3 region) is multiplied by a horizontal edge detection template (with the center column set to zero, the left columns to negative values, and the right columns to positive values) at corresponding locations, and the sum is calculated to obtain the horizontal brightness change of the pixel. Similarly, for the vertical gradient, the brightness of the pixel's neighborhood is multiplied by a specific vertical edge detection template (with the center row set to zero, the upper rows to negative values, and the lower rows to positive values) at corresponding locations, and the sum is calculated to obtain the vertical brightness change of the pixel. Next, the calculated horizontal and vertical gradient values ​​are squared, added together, and the square root of the sum is taken to obtain the overall gradient magnitude for the pixel. Due to the significant brightness step change in the crack edge area, its comprehensive gradient amplitude is usually more than twice that of the normal surface area (for example, the average gradient amplitude in the normal area is about 5, while that at the crack edge can reach more than 15).

[0032] The algorithm uses points with local minima of gradient amplitude as initial seed regions. The threshold for determining these gradient amplitude minima is determined based on a standard, defect-free shock absorber surface image: First, the average gradient amplitude and its standard deviation are calculated for all pixels in the defect-free sample image, and the threshold is set to the average gradient amplitude minus N times the standard deviation (for example, if N = 1.5, the corresponding gradient amplitude threshold is approximately 5). Region expansion is then performed based on the grayscale similarity between adjacent pixels: the tolerance for grayscale value differences is set to 1.2 times the maximum observed difference in grayscale uniformity testing of a defect-free shock absorber surface (for example, if the maximum measured difference is 6 grayscale levels, the tolerance is set to ±7 grayscale levels). Starting from the seed region, the algorithm gradually expands outward, merging adjacent pixels that meet the similarity criteria, ultimately forming independent connected regions. Upon completion, a pixel classification set containing multiple such independent regions is output.

[0033] Next, we traverse each connected region in the pixel classification set and calculate the absolute difference between the average grayscale value within the region and the average grayscale value of the surrounding background area as the grayscale value change. If this difference exceeds a preset grayscale change threshold (the threshold is determined by the statistical distribution characteristics of defect-free samples), the connected region is marked as a potential defect area and summarized to form a defect area set. Canny edge detection is then performed on the defect region set. First, the image is smoothed using a Gaussian filter to eliminate noise (for example, using a 5x5 Gaussian kernel with a standard deviation of 1.5). The gradient magnitude and direction of each pixel are then calculated (also using the Sobel operator). Non-maximum gradient suppression is then performed, retaining only pixels with local maxima in the gradient direction. Finally, a double-threshold filter is performed to extract continuous contours: a high threshold is set, with the high threshold taking 70% of the maximum gradient magnitude in the image, and a low threshold taking 30% of the maximum gradient magnitude in the image. Pixels with gradient magnitudes greater than the high threshold are considered strong edge points, while pixels with gradient magnitudes between the low and high thresholds are considered weak edge points. Weak edge points are retained only if they are directly connected to strong edge points. The output is a sequence of crack boundary coordinates (e.g., a sequence of point coordinates) and a binary texture shape of the worn area. This result is matched with a pre-existing defect template library (containing typical crack morphologies and wear patterns): by calculating the contour shape similarity and texture feature matching (determined according to the preset classification threshold), the crack type code (for example, 01 represents radial cracks) and wear level (for example, 1 represents mild wear) are output, forming a classification feature set containing defect types and levels.

[0034] Furthermore, a coordinate sequence of defect boundary points is extracted from the classification feature set. Points are selected at regular intervals along the boundary, and the local curvature value (reflecting the degree of boundary curvature) is calculated for each point. If the curvature value exceeds a preset curvature threshold (this threshold is determined based on the material's ability to resist stress concentration), the plane coordinates of that point are recorded. All high-curvature points constitute a spatially distributed dataset describing the key morphology of the defect. Finally, based on the defect spatial distribution dataset, a minimum enclosing rectangle algorithm is applied to each defect region: the algorithm finds the smallest rectangle that completely encloses the defect point and outputs its center coordinates, length, and width. These geometric parameters are then combined with the defect type code and wear level from the classification feature set according to a predefined structure (including defect type, center abscissa, center ordinate, length, width, and wear level) to generate a structured surface feature vector. This step uses watershed segmentation and curvature analysis to precisely quantify defect geometric features (e.g., points of high curvature correspond to stress concentration areas at the crack tip). This, combined with morphological matching, enables automated defect classification, significantly reducing the rate of missed detection of small cracks during manual visual inspection. The generated surface feature vector contains key parameters for defect location, size, and type, providing a spatial reference for subsequent vibration performance correlation analysis. For example, the defect center coordinates are directly used to locate the finite element stress simulation area, and the defect length parameter is used in crack growth risk assessment calculations, forming the basis for a closed-loop analysis from visual features to performance prediction.

[0035] In step S14, vibration amplitude acquisition and phase angle measurement operations are performed according to the surface feature vector, and spectrum analysis is performed to construct a performance parameter vector.

[0036] In a specific embodiment, the vibration amplitude acquisition and phase angle measurement operations are performed according to the surface feature vector, and the spectrum analysis is performed to construct the performance parameter vector, including: Based on the surface feature vectors, the shock absorber is controlled to operate within a set speed range through a dynamic balancing test platform, an acceleration sensor is used to collect vibration amplitude, and a photoelectric encoder is used to perform phase angle measurement to obtain an initial signal set including speed, amplitude, and phase. Based on the initial signal set, a fast Fourier transform algorithm is used to perform time-frequency conversion to generate a spectrum diagram, and a power spectrum density analysis is used to extract a frequency response peak synchronized with the rotational speed. When the frequency response peak exceeds a preset frequency response threshold, an abnormal frequency band is marked, and a frequency response characteristic distribution diagram of the marked abnormal frequency band is obtained; According to the frequency response characteristic distribution diagram, the vibration waveform of the abnormal frequency band is nonlinearly time-aligned with a pre-stored fault template library using a dynamic time warping algorithm, and the waveform amplitude deviation is calculated to obtain vibration signal data including speed, frequency and waveform amplitude deviation; According to the vibration signal data, a time-frequency conversion operation is performed using a fast Fourier transform algorithm to generate a spectrum diagram, and the amplitude characteristics of the preset primary and secondary frequencies are extracted to obtain vibration signal characteristics; Based on the vibration signal characteristics, energy integration is performed on the frequency band exceeding the preset frequency threshold, and an initial performance vector is generated in combination with a preset dynamic balance weight coefficient; Based on the initial performance vector, feature dimension reduction is performed using a principal component analysis algorithm, and a performance parameter vector is extracted.

[0037] Specifically, based on the defect location coordinates recorded in the surface feature vector, the dynamic balancing test platform controls the shock absorber to operate within a preset speed range (e.g., 800-3000 rpm) covering its rated operating range, with the speed increased in steps of fixed lengths (e.g., 100 rpm). An accelerometer is installed at a specified location on the shock absorber surface (based on the defect coordinates) to collect vibration amplitude data in real time. A photoelectric encoder is used to simultaneously measure the rotational phase angle, generating an initial signal set containing speed, amplitude, and phase angle values.

[0038] Perform a fast Fourier transform on the initial signal set: This converts the time-domain vibration signal into a frequency-domain signal, generating a spectrogram. Power spectral density analysis is used to extract frequency response peaks that are synchronized with the rotational speed (e.g., fundamental frequency amplitude A1, double frequency amplitude A2). If a frequency response peak exceeds a preset frequency response threshold (this threshold is based on the maximum allowable vibration amplitude of a defect-free sample at the same rotational speed), the frequency band is marked as abnormal, and a frequency response characteristic distribution diagram of the anomaly-marked band is output.

[0039] For the marked abnormal frequency band, the corresponding waveform segment is extracted from the time-domain vibration signal. Using a dynamic time warping algorithm, this waveform segment is nonlinearly aligned with a library of typical fault templates (including waveforms for typical faults such as imbalance and misalignment) generated through cluster analysis of multiple sets of vibration waveform data collected from standard faulty parts at corresponding speeds. This involves stretching or compressing the time axis to achieve a minimum path distance match between the waveform under test and the template waveform. The absolute deviation of the aligned waveform amplitude relative to the template is calculated, generating vibration signal data containing speed values, abnormal frequency values, and waveform amplitude deviation values.

[0040] A Fast Fourier Transform (FFT) is performed again on the vibration signal data to extract the amplitude characteristics of preset primary and secondary frequencies (such as the fundamental frequency, double frequency, and triple frequency). This results in a vibration signal signature consisting of the amplitudes of key frequency points. Based on this signature, the proportion of the energy in the frequency band exceeding a preset frequency threshold (for example, the high frequency band above 1kHz) to the total energy is calculated. This proportion is then weighted and combined with a preset dynamic balancing weight coefficient (assigned based on the rotor's dynamic characteristics) to generate an initial performance vector (for example, [high frequency energy proportion 0.35, fundamental frequency amplitude deviation 0.12, double frequency amplitude deviation 0.08]).

[0041] Finally, the principal component analysis algorithm is used to reduce the dimension of the initial performance vector: the variance contribution rate of each feature dimension is calculated, the principal component components with a cumulative contribution rate exceeding 85% are retained, and the performance parameter vector after dimensionality reduction is extracted.

[0042] This step uses dual spectrum analysis combined with dynamic waveform matching to accurately extract vibration features that are strongly correlated with defects. The first fast Fourier transform locates the abnormal frequency band, and the combined dynamic time warping eliminates the waveform distortion caused by speed fluctuations to ensure the comparability of fault features; the second fast Fourier transform focuses on the amplitude of the primary and secondary frequencies, supplemented by the high-frequency energy ratio to quantify the abnormal vibration energy distribution; finally, principal component analysis is used to eliminate redundant dimensions, so that the performance parameter vector can focus on characterizing the core vibration mode caused by the defect. The generated performance parameter vector is directly related to the degree of interference of the defect on the dynamic balancing performance. For example, the high-frequency energy ratio effectively reflects the nonlinear vibration phenomenon caused by the crack, providing quantitative performance indicators for the subsequent cross-domain fusion of surface features, and laying the data foundation for defect-performance correlation modeling.

[0043] In step S15, data fusion is performed based on the performance parameter vector and the surface feature vector to obtain a fusion feature.

[0044] In a specific embodiment, performing data fusion according to the performance parameter vector and the surface feature vector to obtain a fusion feature includes: According to the crack boundary coordinates, the crack morphology similarity is calculated by using the Frechet distance algorithm to extract a set of geometric parameters; Extracting intrinsic fluctuation components according to the performance parameter vector and calculating the frequency band energy integral to generate dominant frequency characteristics; According to the geometric parameter set and the dominant frequency feature, weight coefficients are allocated by an entropy weight method, and weighted fusion is performed to obtain a fusion feature.

[0045] Specifically, a crack boundary coordinate sequence (denoted as a set of coordinate points P) is extracted from the surface feature vector. The Frechet distance algorithm is then used to calculate the degree of similarity between the crack morphology and corresponding templates in a pre-stored standard crack template library. This standard crack template library is constructed based on typical crack samples (such as radial cracks and circumferential cracks) confirmed by metallographic examination. The crack contour coordinate point set is obtained through high-precision laser scanning and stored as a reference template after smoothing filtering and key point sampling. The algorithm calculates the maximum value of the Euclidean distances between all corresponding points on two curves as a morphological difference measure. The algorithm also extracts the crack length (obtained by summing the distances between adjacent boundary points) and the boundary fractal dimension. This value is calculated using a box counting method: the crack boundary contour is covered with a square grid of varying side lengths. The logarithmic slope of the minimum number of grid cells required to cover the contour is recorded as a function of the grid side length. A larger slope indicates a higher degree of contour irregularity (for example, the fractal dimension of fatigue cracks is typically greater than 1.3). This slope then forms a set of geometric parameters (including crack length and fractal dimension). Extracting intrinsic fluctuation components from the performance parameter vector: Using the empirical mode decomposition algorithm, the vibration signal is decomposed into multiple intrinsic mode functions. The component with the highest energy contribution is selected as the core fluctuation component. The energy integral of this component within a preset frequency band (e.g., 1-2 kHz) is calculated to generate the dominant frequency signature (including the core component energy value and the frequency band energy value). The geometric parameter set and the dominant frequency features are combined into an initial feature matrix, and the weight coefficients are automatically assigned using the entropy weight method: first, the information entropy value of each feature dimension is calculated (based on the discrete degree of the eigenvalue distribution), and the feature with the smaller entropy value is assigned a higher weight; the weight coefficient is calculated based on the inverse relationship of the entropy value (for example, the crack length weight coefficient is 0.3, and the frequency band energy weight coefficient is 0.4); finally, the eigenvalues ​​are weighted linearly fused according to the weights, and the fused features are output (for example, 0.3×crack length + 0.4×frequency band energy + the weighted sum of other features). This step uses the Frechet distance to precisely quantify the deviation of the crack geometry from the standard template (for example, an increase in fractal dimension indicates crack propagation risk). Combined with the frequency band energy integration of the vibration intrinsic components (for example, a surge in high-frequency energy reflects the nonlinear response triggered by the crack), this approach overcomes the limitations of traditional single-dimensional feature analysis. The entropy weighting method dynamically assigns weights based on the inherent information content of the features, avoiding subjective experience bias and ensuring that the fused features simultaneously capture the strong correlation between geometric defect characteristics and dynamic performance degradation (for example, high-weighted frequency band energy is directly related to the predicted damping performance degradation). This provides comprehensive input across the visual and vibration domains for subsequent defect prediction models.

[0046] In step S16, the fusion features are input into a pre-built association prediction model to obtain defect prediction data.

[0047] In a specific embodiment, the process of constructing the association prediction model includes: Obtain historical fusion features and corresponding historical defect data; The support vector machine algorithm is used to build the initial association prediction model, the classification boundary is initialized by the RBF kernel function, and the penalty coefficient is set to determine the tolerance error threshold; Based on the historical fusion features, the kernel density estimation method is used to calculate the curvature adjustment parameter of the classification boundary, and the curvature of the classification boundary is dynamically adjusted through the RBF kernel function; Optimizing the penalty coefficient by a grid search method based on the historical defect data to determine an error tolerance threshold; Perform cross-validation on the adjusted classification boundaries and calculate the classification accuracy and error distribution set after boundary adjustment; When the error distribution set exceeds the error tolerance threshold, the curvature parameter of the classification boundary is re-optimized using the gradient descent method, and the classification boundary is iteratively updated until the error distribution set does not exceed the error tolerance threshold; The optimized classification boundary is matched with the defect data in multi-dimensional space through RBF kernel function mapping to determine the hyperplane equation corresponding to each defect data; Combined with the time series characteristics of the historical defect data, the sliding window method is used to extract the dynamic trend parameters of the defect data changes, and the dynamic trend parameters are associated with the hyperplane equation to construct the final association prediction model.

[0048] Specifically, in the implementation process of building the association prediction model, the historical fusion feature dataset (including a set of geometric parameters such as crack length and fractal dimension generated from crack boundary coordinates, as well as dominant frequency features such as the core fluctuation component energy value and the band energy integral value extracted from the performance parameter vector) and the corresponding historical defect data (including defect type code, location coordinate set, size parameter vector, and performance interference coefficient) are first obtained. The support vector machine algorithm is used to establish the initial model framework, and the classification boundary is initialized using the radial basis function kernel function: the historical fusion feature is used as the input vector, the historical defect type code is used as the classification label, and an initial penalty coefficient is set (set in the range of 0.1-10) to control the classification error tolerance threshold (for example, the initial maximum allowable classification error rate is set to 5%).

[0049] Based on the spatial distribution characteristics of historical fusion features, a kernel density estimation operation is performed: each feature dimension (such as crack length and frequency band energy value) is traversed, and the density of data points in a unit area around each coordinate point in the feature space is calculated. When the density value in a certain area exceeds a preset density threshold (determined by the statistical quantile of the feature value distribution), the width parameter of the radial basis function kernel function in that area is increased (for example, from 0.5 to 1.0), increasing the curvature of the classification boundary to fit the high-density data clusters. Conversely, in low-density areas (density values ​​below the lower limit of the statistical quantile), the width parameter is reduced (for example, from 0.5 to 0.2) to reduce the boundary curvature and avoid overfitting.

[0050] Based on the actual classification results of historical defect data, the penalty coefficient is optimized using a grid search method: The penalty coefficient is iterated within a preset coefficient range (0.1-10) with a step size of 0.1. Each iteration performs 10-fold cross-validation, and the classification accuracy and high-risk defect missed detection rate (e.g., the proportion of cracks misclassified as non-defective) are calculated for each penalty coefficient. The penalty coefficient that minimizes the high-risk defect missed detection rate (e.g., ≤2%) and maximizes the overall accuracy (e.g., ≥95%) is selected as the optimal value. The error tolerance threshold is simultaneously updated (e.g., the high-risk defect missed detection threshold is set to 2%). Cross-validation is performed on the curvature-adjusted classification boundary. If the high-risk defect missed detection rate in the error distribution set exceeds 2%, iterative optimization is performed using gradient descent. Using the classification error rate as the loss function, the partial derivative of the loss function with respect to the width parameter of the radial basis function kernel is calculated. The width parameter is updated along the gradient descent direction (e.g., from 0.8 to 0.7). Verification is repeated until the high-risk defect missed detection rate is ≤2%.

[0051] After parameter optimization, the defect data is converted to a high-dimensional space using radial basis function kernel mapping. For each historical defect data point (e.g., a eigenvector containing a crack length of 3.2 mm and a band energy value of 0.45), the kernel function distance (e.g., Gaussian distance) to the support vector is calculated, and the optimal classification hyperplane equation is solved (e.g., the crack hyperplane equation is a linear combination of the weight vector and the eigenvector equal to the offset). Finally, taking into account the temporal continuity of the historical defect data, a sliding window method with a fixed window size (set to 30 days based on the data collection period) is employed: defect size parameter vectors within the continuous window are intercepted in chronological order, and the linear regression slope of the crack length within the window is calculated as the dynamic trend parameter (e.g., monthly expansion of 0.15 mm). This dynamic expansion rate parameter is combined with the hyperplane equation to construct a final prediction model that integrates static classification and dynamic evolution.

[0052] During the construction of the association prediction model, the radial basis function kernel dynamically adjusts the curvature of the classification boundary. Specifically, based on a historical fusion feature dataset (including geometric parameters such as crack length and fractal dimension, and dominant frequency features such as band energy), a kernel density estimation algorithm is used to calculate the density distribution values ​​for each region of the feature space. When the density value in a region exceeds a density threshold determined by the statistical quantile of the historical data, the width parameter of the radial basis function kernel in that region is automatically increased, increasing the curvature of the classification boundary to tightly surround high-density data clusters (for example, the width parameter is increased from a baseline value of 0.5 to 0.7). Conversely, the width parameter is decreased in low-density regions (for example, from 0.5 to 0.3), reducing the boundary curvature and preventing overfitting. This operation enables the classification boundary to adaptively deform according to the data distribution characteristics, addressing the classification response lag caused by the time-varying nature of the vibration signal in traditional models (for example, the fluctuation of classification accuracy is significantly improved when the speed changes suddenly).

[0053] In step S17, defect impact analysis is performed based on the defect prediction data to obtain a defect assessment result.

[0054] In a specific embodiment, performing defect impact analysis based on the defect prediction data to obtain a defect assessment result includes: Performing a matching operation on the defect prediction data through a pre-stored historical defect-performance degradation mapping library to obtain a baseline performance degradation reference value; Wherein, the defect prediction data includes defect type code, location coordinate set, size parameter vector and performance interference coefficient; Based on the performance interference coefficient and the reference value of the benchmark performance attenuation, a finite element stress simulation analysis is performed in combination with the pre-stored hardness of the shock absorber material to obtain the stress in the defect area and the damping performance attenuation; Based on the stress in the defect area and the damping performance attenuation, a risk level determination is performed. When the stress in the defect area exceeds the preset material yield strength threshold or the damping performance attenuation exceeds the preset dynamic balance tolerance threshold, a high-risk mark is output, and finally a defect assessment result including the defect location, size, risk level, and damping performance attenuation is generated.

[0055] Specifically, the system uses defect prediction data as input (including a defect type code, such as crack type code 01; a location coordinate set, such as the crack center coordinates [x, y]; a dimension parameter vector, such as a crack length of 3.2 mm and a width of 0.5 mm; and a performance interference coefficient, such as a vibration frequency band energy fraction of 0.35). The system then performs a matching operation by querying a pre-stored historical defect-performance degradation mapping library. This mapping library is constructed based on historical experimental data (for example, using the average performance degradation corresponding to defect samples with the same type code and a size within a ±10% error range as a benchmark) and outputs a baseline performance degradation reference value for the current defect (e.g., a damping performance degradation reference value of 15%).

[0056] Combined with the pre-stored hardness value of the shock absorber material (such as the spring steel hardness value HRC52), finite element stress simulation analysis is used: the simulation area grid is defined by the defect position coordinate set and the size parameter vector, and the performance interference coefficient is used as the boundary condition load (for example, the frequency band energy ratio of 0.35 is converted into an equivalent dynamic stress load). By solving the material constitutive equation (using Hooke's law and the damping dissipation model), the maximum stress value of the defect area (such as a stress peak of 380MPa) and the damping performance attenuation (such as an actual attenuation of 18%) are calculated.

[0057] During the risk level determination phase, the stress value in the defect area is compared with the material yield strength threshold (this threshold is determined based on the material hardness value from standard material mechanical properties tables, such as 355 MPa for spring steel). The damping performance degradation is also compared with the dynamic balance tolerance threshold (this threshold is statistically determined through dynamic balancing bench testing, such as a maximum allowable degradation of 20%). If the stress value in the defect area exceeds the preset material yield strength threshold or the damping performance degradation exceeds the preset dynamic balance tolerance threshold, a high risk indicator (such as the risk level code "R4") is assigned. Finally, a structural defect assessment result is generated, containing the following fields: defect location coordinates, defect size, risk level R4, and damping performance degradation.

[0058] Through dual verification of the historical mapping library and finite element simulation (such as the algorithm for converting the frequency band energy ratio into stress load), the problem of misjudgment of implicit performance degradation due to traditional manual experience is solved; the threshold dynamic judgment mechanism driven by material hardness (such as automatically matching the yield strength threshold according to HRC52) realizes closed-loop quantitative evaluation from defect characteristics to service risk level (for example, accurately warning of structural failure risks exceeding the yield strength), providing data support for proactive maintenance of shock absorbers.

[0059] Reference Figure 2 The second embodiment of the present invention provides an electric vehicle shock absorber defect detection system based on machine vision, comprising: A data acquisition module is used to acquire an initial image set under multi-angle light source illumination through a high-resolution imaging acquisition device; An image filtering module, configured to perform denoising filtering on the initial image set to obtain clear image data; A surface feature module, configured to perform defect classification and spatial distribution analysis based on the clear image data to obtain a surface feature vector including defect type and defect spatial distribution; A performance parameter module is used to collect vibration amplitude and phase angle measurements based on the surface feature vector, and to perform spectrum analysis to construct a performance parameter vector; A feature fusion module, configured to perform data fusion based on the performance parameter vector and the surface feature vector to obtain a fusion feature; A defect prediction module, configured to input the fusion features into a pre-built association prediction model to obtain defect prediction data; The defect assessment module is used to perform defect impact analysis based on the defect prediction data to obtain a defect assessment result.

[0060] It should be noted that the electric vehicle shock absorber defect detection device based on machine vision provided in an embodiment of the present invention is used to execute all the process steps of the electric vehicle shock absorber defect detection method based on machine vision in the above embodiment. The working principles and beneficial effects of the two correspond one to one, so they will not be repeated here.

[0061] An embodiment of the present invention further provides an electronic device. The electronic device includes: a processor, a memory, and a computer program stored in the memory and executable on the processor, such as a machine vision-based electric vehicle shock absorber defect detection program. When the processor executes the computer program, the steps in the above-mentioned embodiments of the electric vehicle shock absorber defect detection method based on machine vision are implemented, such as Figure 1 Alternatively, when the processor executes the computer program, the functions of the modules / units in the above-mentioned device embodiments are realized, such as a machine vision-based electric vehicle shock absorber defect detection module.

[0062] Exemplarily, the computer program may be divided into one or more modules / units, which are stored in the memory and executed by the processor to implement the present invention. The one or more modules / units may be a series of computer program instruction segments capable of implementing specific functions, and the instruction segments are used to describe the execution process of the computer program in the electronic device.

[0063] The electronic device may be a computing device such as a desktop computer, notebook, PDA, or smart tablet. The electronic device may include, but is not limited to, a processor and memory. Those skilled in the art will appreciate that the aforementioned components are merely examples of electronic devices and do not constitute a limitation of the electronic device. The electronic device may include more or fewer components than those described above, or a combination of certain components, or different components. For example, the electronic device may also include input / output devices, network access devices, buses, and the like.

[0064] The processor may be a central processing unit (CPU), other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA), other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor. The processor is the control center of the electronic device and connects various parts of the entire electronic device using various interfaces and lines.

[0065] The memory can be used to store the computer programs and / or modules. The processor implements the various functions of the electronic device by running or executing the computer programs and / or modules stored in the memory and accessing the data stored in the memory. The memory may primarily include a program storage area and a data storage area. The program storage area may store an operating system and at least one application required for a function (such as a sound playback function or an image playback function); the data storage area may store data generated based on the use of the mobile phone (such as audio data, a phone book, etc.). Furthermore, the memory may include high-speed random access memory and non-volatile memory, such as a hard disk, internal memory, a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, at least one disk storage device, a flash memory device, or other volatile solid-state storage device.

[0066] If the module / unit integrated into the electronic device is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present invention can implement all or part of the process steps in the above-mentioned method embodiments by using a computer program to instruct the relevant hardware. The computer program can be stored in a computer-readable storage medium. When executed by a processor, the computer program can implement the steps of each of the above-mentioned method embodiments. The computer program includes computer program code, which can be in source code form, object code form, executable file, or some intermediate form. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording medium, USB flash drive, mobile hard drive, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electric carrier signal, telecommunication signal, and software distribution medium. It should be noted that the content of the computer-readable medium can be appropriately increased or decreased based on the requirements of legislation and patent practice in a jurisdiction. For example, in some jurisdictions, based on legislation and patent practice, computer-readable media does not include electric carrier signals and telecommunication signals.

[0067] It should be noted that the device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed across multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the present embodiment. In addition, in the drawings of the device embodiments provided by the present invention, the connection relationship between the modules indicates that there is a communication connection between them, which may be specifically implemented as one or more communication buses or signal lines. A person of ordinary skill in the art can understand and implement the present invention without inventive effort.

[0068] The specific embodiments described above further illustrate the objectives, technical solutions, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. In particular, it should be noted that any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included within the scope of protection of the present invention for those skilled in the art.

Claims

1. A method for detecting defects in electric vehicle shock absorbers based on machine vision, characterized in that: include: Collecting an initial set of images under multi-angle light sources using a high-resolution imaging acquisition device; Performing denoising filtering on the initial image set to obtain clear image data; Perform defect classification and spatial distribution analysis based on the clear image data to obtain a surface feature vector including defect type and defect spatial distribution; Performing vibration amplitude acquisition and phase angle measurement operations according to the surface feature vector, and performing spectrum analysis to construct a performance parameter vector; Performing data fusion based on the performance parameter vector and the surface feature vector to obtain a fusion feature; Inputting the fusion features into a pre-built association prediction model to obtain defect prediction data; A defect impact degree analysis is performed based on the defect prediction data to obtain a defect assessment result.

2. The electric vehicle shock absorber defect detection method based on machine vision according to claim 1, characterized in that: The performing denoising filtering processing on the initial image set to obtain clear image data includes: Based on the initial image set, pixel noise is suppressed by using a median filter. If salt and pepper noise interference is detected, a secondary smoothing denoising operation is performed by using a Gaussian filter to obtain a first preprocessed image; performing an overall brightness distribution remapping operation on the first preprocessed image by using an adaptive histogram equalization algorithm to obtain a second preprocessed image; performing a pixel gradient calculation operation on the second preprocessed image using a Sobel edge detection operator, identifying continuous high gradient sequences and marking them as suspected crack regions, thereby obtaining a third preprocessed image; According to the third preprocessed image set, a wear area contour extraction operation is performed through a morphological opening operation to highlight crack and wear features and obtain clear image data.

3. The electric vehicle shock absorber defect detection method based on machine vision according to claim 1, characterized in that: The defect classification and spatial distribution analysis are performed based on the clear image data to obtain a surface feature vector including the defect type and the defect spatial distribution, including: Based on the clear image data, a pixel clustering operation is performed on the surface area of ​​the shock absorber using a watershed segmentation algorithm to obtain a pixel classification set; Extracting grayscale value variations according to the pixel classification set, and taking pixel classification sets whose grayscale value variations exceed a preset grayscale variation threshold as potential defect areas, to obtain a defect area set; Based on the defect area set, a continuous contour outlining operation is performed using the Canny edge detection algorithm to extract the crack boundary coordinates and wear texture shape, and morphological matching classification is performed in combination with a pre-stored defect template library to obtain a classification feature set including crack type and wear grade; Extract defect boundary points based on the classification feature set and perform curvature analysis, record the corresponding plane coordinates of the defect boundary points whose curvature exceeds a preset curvature threshold, and obtain the defect spatial distribution; According to the spatial distribution of the defects, the length and width of the defect area are quantified using a minimum circumscribed rectangle algorithm, and structured data is generated in combination with the classification feature set to obtain a surface feature vector including the defect type, defect location and defect size.

4. The electric vehicle shock absorber defect detection method based on machine vision according to claim 3 is characterized in that: The vibration amplitude acquisition and phase angle measurement operations are performed according to the surface feature vector, and spectrum analysis is performed to construct a performance parameter vector, including: Based on the surface feature vectors, the shock absorber is controlled to operate within a set speed range through a dynamic balancing test platform, an acceleration sensor is used to collect vibration amplitude, and a photoelectric encoder is used to perform phase angle measurement to obtain an initial signal set including speed, amplitude, and phase. Based on the initial signal set, a fast Fourier transform algorithm is used to perform time-frequency conversion to generate a spectrum diagram, and a power spectrum density analysis is used to extract a frequency response peak synchronized with the rotational speed. When the frequency response peak exceeds a preset frequency response threshold, an abnormal frequency band is marked, and a frequency response characteristic distribution diagram of the marked abnormal frequency band is obtained; According to the frequency response characteristic distribution diagram, the vibration waveform of the abnormal frequency band is nonlinearly time-aligned with a pre-stored fault template library using a dynamic time warping algorithm, and the waveform amplitude deviation is calculated to obtain vibration signal data including speed, frequency and waveform amplitude deviation; According to the vibration signal data, a time-frequency conversion operation is performed using a fast Fourier transform algorithm to generate a spectrum diagram, and the amplitude characteristics of the preset primary and secondary frequencies are extracted to obtain vibration signal characteristics; Based on the vibration signal characteristics, energy integration is performed on the frequency band exceeding the preset frequency threshold, and an initial performance vector is generated in combination with a preset dynamic balance weight coefficient; Based on the initial performance vector, feature dimension reduction is performed using a principal component analysis algorithm, and a performance parameter vector is extracted.

5. The electric vehicle shock absorber defect detection method based on machine vision according to claim 3, characterized in that: The performing data fusion according to the performance parameter vector and the surface feature vector to obtain a fusion feature includes: According to the crack boundary coordinates, the crack morphology similarity is calculated by using the Frechet distance algorithm to extract a set of geometric parameters; Extracting intrinsic fluctuation components according to the performance parameter vector and calculating the frequency band energy integral to generate dominant frequency characteristics; According to the geometric parameter set and the dominant frequency feature, weight coefficients are allocated by an entropy weight method, and weighted fusion is performed to obtain a fusion feature.

6. The electric vehicle shock absorber defect detection method based on machine vision according to claim 1, characterized in that: The construction process of the association prediction model includes: Obtain historical fusion features and corresponding historical defect data; The support vector machine algorithm is used to build the initial association prediction model, the classification boundary is initialized by the RBF kernel function, and the penalty coefficient is set to determine the tolerance error threshold; Based on the historical fusion features, the kernel density estimation method is used to calculate the curvature adjustment parameter of the classification boundary, and the curvature of the classification boundary is dynamically adjusted through the RBF kernel function; Optimizing the penalty coefficient by a grid search method based on the historical defect data to determine an error tolerance threshold; Perform cross-validation on the adjusted classification boundaries and calculate the classification accuracy and error distribution set after boundary adjustment; When the error distribution set exceeds the error tolerance threshold, the curvature parameter of the classification boundary is re-optimized using the gradient descent method, and the classification boundary is iteratively updated until the error distribution set does not exceed the error tolerance threshold; The optimized classification boundary is matched with the defect data in multi-dimensional space through RBF kernel function mapping to determine the hyperplane equation corresponding to each defect data; Combined with the time series characteristics of the historical defect data, the sliding window method is used to extract the dynamic trend parameters of the defect data changes, and the dynamic trend parameters are associated with the hyperplane equation to construct the final association prediction model.

7. The electric vehicle shock absorber defect detection method based on machine vision according to claim 1, characterized in that: The defect impact analysis is performed based on the defect prediction data to obtain a defect assessment result, including: Performing a matching operation on the defect prediction data through a pre-stored historical defect-performance degradation mapping library to obtain a baseline performance degradation reference value; Wherein, the defect prediction data includes defect type code, location coordinate set, size parameter vector and performance interference coefficient; According to the performance interference coefficient and the reference value of the benchmark performance attenuation, a finite element stress simulation analysis is performed in combination with the pre-stored hardness of the shock absorber material to obtain the stress of the defect area and the damping performance attenuation; Based on the stress in the defect area and the damping performance attenuation, a risk level determination is performed. When the stress in the defect area exceeds the preset material yield strength threshold or the damping performance attenuation exceeds the preset dynamic balance tolerance threshold, a high-risk mark is output, and finally a defect assessment result including the defect location, size, risk level, and damping performance attenuation is generated.

8. A machine vision-based electric vehicle shock absorber defect detection system, characterized in that: include: A data acquisition module is used to acquire an initial image set under multi-angle light source illumination through a high-resolution imaging acquisition device; An image filtering module, configured to perform denoising filtering on the initial image set to obtain clear image data; A surface feature module, configured to perform defect classification and spatial distribution analysis based on the clear image data to obtain a surface feature vector including defect type and defect spatial distribution; A performance parameter module is used to collect vibration amplitude and phase angle measurements based on the surface feature vector, and to perform spectrum analysis to construct a performance parameter vector; A feature fusion module, configured to perform data fusion based on the performance parameter vector and the surface feature vector to obtain a fusion feature; A defect prediction module, configured to input the fusion features into a pre-built association prediction model to obtain defect prediction data; The defect assessment module is used to perform defect impact analysis based on the defect prediction data to obtain a defect assessment result.

9. An electronic device, characterized in that: The method comprises a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein when the processor executes the computer program, the method for detecting defects of an electric vehicle shock absorber based on machine vision is implemented as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium includes a stored computer program, wherein when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute the electric vehicle shock absorber defect detection method based on machine vision as described in any one of claims 1 to 7.

Images