A steel wheel surface scratch detection method and system

Abstract

The present application relates to the technical field of steel wheel quality detection, and particularly relates to a steel wheel surface scratch detection method and system, which comprises the following steps: constructing a laser speckle interference detection system to collect a steel wheel laser speckle interference image; preprocessing the collected steel wheel laser speckle interference image, including image denoising and image enhancement; constructing a steel wheel surface scratch feature extraction model based on deep learning and training, inputting the enhanced steel wheel laser speckle interference image into the trained steel wheel surface scratch feature extraction model, and outputting a scratch feature vector of the image; determining a scratch detection threshold value by statistically analyzing the steel wheel surface scratch features, comparing the scratch feature vector of the model output image with the set threshold value, detecting the scratch area, and identifying the scratch type. The present application amplifies the scratch defects on the surface of the steel wheel through laser speckle interference, realizes accurate identification of the scratches on the surface of the steel wheel, and improves the detection efficiency of the scratches on the surface of the steel wheel.

Description

Technical Field

[0001] This invention relates to the field of steel wheel quality inspection technology, specifically to a method and system for detecting scratches on the surface of steel wheels. Background Technology

[0002] Steel wheels are widely used in transportation, machinery manufacturing, and other fields, and their surface quality directly affects the performance, safety, and service life of the equipment. Surface scratches are one of the most common defects in steel wheels. Even tiny scratches can become crack sources over long-term use, leading to wheel damage. Traditional methods for detecting surface scratches on steel wheels mainly rely on manual visual inspection. This method is not only inefficient and subjective, but also difficult to accurately detect tiny scratches and scratches in hidden locations. Although some existing automated inspection technologies, such as machine vision inspection, have improved inspection efficiency to some extent, they still suffer from low accuracy when detecting scratches against complex backgrounds. Therefore, developing an efficient and accurate method for detecting surface scratches on steel wheels is of significant practical importance. Summary of the Invention

[0003] To address the aforementioned technical shortcomings, this invention provides a method and system for detecting scratches on the surface of steel wheels, thereby achieving accurate detection of scratches on the surface of steel wheels.

[0004] This invention is achieved through the following technical solution:

[0005] A method for detecting scratches on the surface of a steel wheel is provided, the method comprising the following steps:

[0006] Step S10: Construct a laser speckle interferometry detection system to acquire laser speckle interferometry images of the steel wheel. The system includes a laser, a beam expander, a reflector, an imaging lens, and a high-speed camera.

[0007] Step S20: The acquired steel wheel laser speckle interferometric image is denoised using a wavelet threshold denoising algorithm, and the denoised image is enhanced using a contrast-limited adaptive histogram equalization algorithm.

[0008] Step S30: Construct and train a deep learning-based steel wheel surface scratch feature extraction model. Input the enhanced steel wheel laser speckle interferometry image into the trained steel wheel surface scratch feature extraction model. The model outputs the scratch feature vector of the steel wheel laser speckle interferometry image.

[0009] Step S40: By statistically analyzing the scratch features on the steel wheel surface, the threshold for scratch detection is determined. The scratch feature vector of the model output image is compared with the set threshold. When the feature vector is greater than the threshold, it is determined that there is a scratch in the area corresponding to the image. When the feature vector is less than or equal to the threshold, it is determined that there is no scratch in the area corresponding to the image. For areas with scratches, the type of scratch is identified and classified according to the scratch feature vector.

[0010] Preferably, step S10, which involves constructing a laser speckle interferometry detection system to acquire laser speckle interferometry images of a steel wheel, includes:

[0011] Construct a laser speckle interferometry detection system: including a laser, a beam expander, a reflector, an imaging lens, and a high-speed camera. The laser beam emitted by the laser is expanded by the beam expander and then reflected by the reflector to illuminate the surface of a steel wheel. The scattered light from the surface of the steel wheel is imaged onto the photosensitive surface of the high-speed camera by the imaging lens.

[0012] Laser speckle interferometry image acquisition of steel wheel: By applying a small vibration excitation to the surface of the steel wheel to cause deformation, a high-speed camera is used to acquire laser speckle interferometry images of the steel wheel surface before and after vibration. The frequency and amplitude of the vibration excitation are adjusted according to the material and size of the steel wheel, generally with a frequency range of 10-100Hz and an amplitude of 0.01-0.1mm.

[0013] Preferably, the step S20, which uses a wavelet thresholding denoising algorithm to denoise the acquired steel wheel laser speckle interferometric image, includes:

[0014] Selected parameters: Based on the texture and noise characteristics of the steel wheel laser speckle interferometry image, the Symlets wavelet basis, such as sym4, is selected. It has a good balance between smoothness and tight support, and is suitable for processing steel wheel laser speckle interferometry images. Based on the image size and noise of the steel wheel laser speckle interferometry image, the number of decomposition layers is set to 3.

[0015] Wavelet decomposition: A two-dimensional discrete wavelet transform is performed on the laser speckle interferometry image of the steel wheel, which decomposes the image into a low-frequency approximate subband and high-frequency detail subbands in the horizontal, vertical and diagonal directions. After three layers of decomposition, three sets of high-frequency subbands and one low-frequency subband are obtained.

[0016] Threshold determination: The SureShrink adaptive thresholding method is adopted to determine the threshold for each high-frequency sub-band. This method is based on the Stein unbiased likelihood estimation criterion and can automatically adjust the threshold according to the characteristics of the sub-band, so as to better adapt to the noise changes in different regions of the steel wheel image.

[0017] Thresholding: The wavelet coefficients of each high-frequency sub-band are processed using a soft thresholding method. When the absolute value of a wavelet coefficient is greater than the threshold, it is shrunk; when it is less than or equal to the threshold, it is set to 0.

[0018] Image reconstruction: Two-dimensional inverse discrete wavelet transform is performed on the processed wavelet coefficients to reconstruct the low-frequency subband and the processed high-frequency subband, resulting in a denoised steel wheel laser speckle interferometry image that retains key image features and reduces noise interference.

[0019] Preferably, the step S20, which involves enhancing the denoised image using a contrast-limited adaptive histogram equalization algorithm, includes:

[0020] Image segmentation: The segmentation size is determined based on the size of the steel wheel laser speckle interferometry image, and the image is divided into a certain number of non-overlapping rectangular blocks, each of which is a local region of the image;

[0021] Contrast limiting: For each small block, the grayscale histogram is statistically analyzed and a clipping limit value is set to control the maximum height of the histogram of each small block. The number of pixels exceeding the clipping limit value is clipped, and the clipped pixels are evenly redistributed to each grayscale level to obtain the grayscale histogram after contrast limiting. The selection of the clipping limit value needs to be adjusted according to the characteristics of the steel wheel laser speckle interferometry image. A suitable value can be determined through experiments.

[0022] Small patch histogram equalization: For each small patch histogram that has been contrast-limited, calculate its cumulative distribution function, and find the corresponding new gray value through the cumulative distribution function;

[0023] Bilinear interpolation stitching: The small blocks after histogram equalization are stitched together. For pixels in the steel wheel laser speckle interferometry image that are not on the boundaries of the small blocks, the gray value is determined by bilinear interpolation. Based on the position of the pixel in the four adjacent small blocks, the weighting coefficient in each small block is calculated. The weighting coefficient is related to the distance from the pixel to the center of each small block. The closer the distance, the larger the weighting coefficient. Based on the weighting coefficient, the gray values ​​of the four adjacent small blocks are weighted and averaged to obtain the final gray value of the pixel.

[0024] Output Enhanced Image: After the above steps, an enhanced laser speckle interferometry image of the steel wheel is obtained. The contrast of this image is enhanced, and the details of the scratch area are more prominent, which is beneficial for subsequent scratch feature extraction and detection.

[0025] Preferably, the step S30 of constructing and training a deep learning-based steel wheel surface scratch feature extraction model includes:

[0026] Construction of a feature extraction model for scratches on steel wheel surfaces: A convolutional neural network (CNN) model is constructed as the scratch feature extractor. This model contains multiple convolutional layers, pooling layers, and fully connected layers. The convolutional layers are used to extract local features of the scratch images on the steel wheel surface, the pooling layers are used to reduce the dimensionality of the feature maps and reduce the amount of computation, and the fully connected layers are used to classify the extracted features.

[0027] Training of the steel wheel surface scratch feature extraction model: A large number of laser speckle interferometric images of steel wheels with scratch annotations were collected as training sets to train the constructed steel wheel surface scratch feature extraction model. During the training process, the cross-entropy loss function was used as the optimization objective, and the stochastic gradient descent algorithm was used to update the model parameters and adjust the model weights so that the model can accurately extract scratch features.

[0028] Feature extraction: The laser speckle interferometric image of the steel wheel processed in step S20 is input into the trained steel wheel surface scratch feature extraction model. The model outputs the scratch feature vector of the image. The output scratch feature vector includes information such as the shape, size and location of the scratch.

[0029] Preferably, in step S40, the threshold for scratch detection is determined by statistical analysis of the scratch features on the steel wheel surface, including:

[0030] Determine scratch judgment indicators: Select scratch length, width and gray scale change, which are closely related to scratches and can effectively reflect scratch characteristics, as scratch judgment indicators among the scratch features of steel wheel surface. Extract scratch feature vectors from all steel wheel laser speckle interferometry images in the training set, classify them according to key indicators, and form a separate dataset.

[0031] Mean and standard deviation calculation: The mean μ and standard deviation σ of each dataset are calculated using statistical methods. The mean represents the average level of the key indicator in the training set, reflecting the central tendency of the scratch features. The standard deviation represents the dispersion of the data. The larger the standard deviation, the greater the dispersion of the key indicator data; conversely, the smaller the standard deviation, the smaller the dispersion.

[0032] Threshold determination: Based on the actual detection situation, a multiple k of the standard deviation is set to adjust the leniency of the threshold. The larger the multiple, the higher the threshold, the stricter the detection conditions, and the lower the probability of false detection. The smaller the multiple, the lower the threshold, the more lenient the detection conditions, and the more minor scratches can be detected. The calculated mean is added to the standard deviation of the selected multiple to obtain the scratch detection threshold T, T = μ + kσ.

[0033] Furthermore, to achieve the above objectives, the present invention also proposes a steel wheel surface scratch detection system, the steel wheel surface scratch detection system comprising:

[0034] Steel wheel laser speckle interferometric image acquisition module: used to construct a laser speckle interferometric detection system to acquire steel wheel laser speckle interferometric images. The system includes a laser, beam expander, reflector, imaging lens and high-speed camera;

[0035] Steel wheel laser speckle interferometric image preprocessing module: The wavelet threshold denoising algorithm is used to denoise the acquired steel wheel laser speckle interferometric image, and the contrast-limited adaptive histogram equalization algorithm is used to enhance the denoised image.

[0036] Steel wheel surface scratch feature extraction model construction and training module: Construct and train a deep learning-based steel wheel surface scratch feature extraction model. Input the enhanced steel wheel laser speckle interferometry image into the trained steel wheel surface scratch feature extraction model, and the model outputs the scratch feature vector of the steel wheel laser speckle interferometry image.

[0037] Steel wheel surface scratch detection module: By statistically analyzing the scratch features of the steel wheel surface, a scratch detection threshold is determined. The scratch feature vector of the model output image is compared with the set threshold. When the feature vector is greater than the threshold, it is determined that there is a scratch in the corresponding area of ​​the image; when the feature vector is less than or equal to the threshold, it is determined that there is no scratch in the corresponding area of ​​the image. For areas with scratches, the type of scratch is identified and classified according to the scratch feature vector.

[0038] Furthermore, to achieve the above objectives, the present invention also proposes a steel wheel surface scratch detection device, the device comprising: a memory, a processor, and a steel wheel surface scratch detection algorithm and other programs stored in the memory and executable on the processor, the steel wheel surface scratch detection algorithm and other programs comprising the steps of implementing the steel wheel surface scratch detection method described above.

[0039] In addition, to achieve the above objectives, the present invention also provides a computer program product, which includes programs such as a steel wheel surface scratch detection algorithm. When the steel wheel surface scratch detection algorithm and other programs are executed by a processor, they implement a steel wheel surface scratch detection method as described above.

[0040] The advantages and effects of this invention are:

[0041] This invention proposes a method and system for detecting scratches on the surface of steel wheels. It utilizes laser speckle interferometry to amplify and acquire microscopic deformation information of the steel wheel surface, and combines this with the feature extraction and classification capabilities of deep learning technology to achieve high-precision detection of scratches on the steel wheel surface, improving the accuracy, efficiency, and precision of scratch detection. Furthermore, by incorporating deep learning technology, it can adapt to scratch detection on steel wheels of different materials, sizes, and surface conditions, exhibiting good versatility and adaptability. In addition, the entire detection process is automated, reducing manual intervention and improving detection efficiency, making it suitable for large-scale steel wheel production inspection. Attached Figure Description

[0042] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0043] Figure 1 This is a flowchart of a method for detecting scratches on the surface of a steel wheel according to the present invention.

[0044] Figure 2 This is a schematic diagram of the structure of a steel wheel surface scratch detection system according to the present invention.

[0045] Figure 3 This is a schematic block diagram of an electronic device for detecting scratches on the surface of a steel wheel according to the present invention. Detailed Implementation

[0046] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0047] like Figure 1 As shown, in one embodiment of the present invention, a method for detecting scratches on the surface of a steel wheel includes the following steps:

[0048] Step S10: Construct a laser speckle interferometry detection system to acquire laser speckle interferometry images of the steel wheel. The system includes a laser, a beam expander, a reflector, an imaging lens, and a high-speed camera.

[0049] Specifically, step S10 involves constructing a laser speckle interferometry detection system to acquire laser speckle interferometry images of the steel wheel, including:

[0050] Construct a laser speckle interferometry detection system: including a laser, a beam expander, a reflector, an imaging lens, and a high-speed camera. The laser beam emitted by the laser is expanded by the beam expander and then reflected by the reflector to illuminate the surface of the steel wheel. The scattered light from the surface of the steel wheel is imaged onto the photosensitive surface of the high-speed camera by the imaging lens. For example, a solid-state laser with a wavelength of 532nm is selected, and the laser beam is expanded to a diameter of about 50mm by the beam expander. After being reflected by the reflector, it is perpendicularly illuminated onto the surface of the railway steel wheel. An imaging lens with a focal length of 100mm is used to image the scattered light from the surface of the steel wheel onto a high-speed camera with a resolution of 1280×1024.

[0051] Laser speckle interferometry image acquisition on steel wheel: By applying a small vibration excitation to the surface of the steel wheel to cause deformation, a high-speed camera is used to acquire laser speckle interferometry images of the steel wheel surface before and after vibration. The frequency and amplitude of the vibration excitation are adjusted according to the material and size of the steel wheel, generally in the range of 10-100Hz and the amplitude of 0.01-0.1mm. For example, a vibration excitation of 50Hz frequency and 0.05mm amplitude is applied to the steel wheel, and a high-speed camera is used to acquire laser speckle interferometry images of the steel wheel surface before and after vibration at a rate of 100 frames per second, for a total of 1000 frames.

[0052] Step S20: The acquired steel wheel laser speckle interferometric image is denoised using a wavelet threshold denoising algorithm, and the denoised image is enhanced using a contrast-limited adaptive histogram equalization algorithm.

[0053] Specifically, step S20, which uses a wavelet threshold denoising algorithm to denoise the acquired steel wheel laser speckle interferometric image, includes:

[0054] Selected parameters: Based on the texture and noise characteristics of the steel wheel laser speckle interferometry image, Symlets wavelet basis, such as sym4, is selected. It has a good balance between smoothness and compact support, and is suitable for processing steel wheel laser speckle interferometry images. Based on the image size and noise of the steel wheel laser speckle interferometry image, the number of decomposition layers is set to 3. This can effectively separate different frequency components without increasing the amount of computation.

[0055] Wavelet decomposition: A two-dimensional discrete wavelet transform is performed on the laser speckle interferometry image of the steel wheel, which decomposes the image into a low-frequency approximate subband and high-frequency detail subbands in the horizontal, vertical and diagonal directions. After three layers of decomposition, three sets of high-frequency subbands and one low-frequency subband are obtained.

[0056] Threshold determination: The SureShrink adaptive thresholding method is adopted to determine the threshold for each high-frequency sub-band. This method is based on the Stein unbiased likelihood estimation criterion and can automatically adjust the threshold according to the characteristics of the sub-band, so as to better adapt to the noise changes in different regions of the steel wheel image.

[0057] Thresholding: The wavelet coefficients of each high-frequency sub-band are processed using a soft thresholding method. When the absolute value of the wavelet coefficient is greater than the threshold, it is shrunk; when it is less than or equal to the threshold, it is set to 0. This can remove noise and avoid artifacts that may be produced by hard thresholding.

[0058] Image reconstruction: Two-dimensional inverse discrete wavelet transform is performed on the processed wavelet coefficients to reconstruct the low-frequency subband and the processed high-frequency subband, resulting in a denoised steel wheel laser speckle interferometry image that retains key image features and reduces noise interference.

[0059] Specifically, step S20, which involves enhancing the denoised image using a contrast-limited adaptive histogram equalization algorithm, includes:

[0060] Image segmentation: The segmentation size is determined based on the size of the steel wheel laser speckle interferometry image. The image is divided into a certain number of non-overlapping rectangular blocks. Each block is a local region of the image. For example, blocks of 8×8, 16×16, or 32×32 pixels can be selected. Smaller segmentation sizes can better capture local contrast changes, but increase the amount of computation. Larger segmentation sizes have higher computational efficiency, but will ignore some subtle local features.

[0061] Contrast limiting: For each small block, the grayscale histogram is statistically analyzed and a clipping limit value is set to control the maximum height of the histogram of each small block. The number of pixels exceeding the clipping limit value is clipped, and the clipped pixels are evenly redistributed to each grayscale level to obtain the grayscale histogram after contrast limiting. The selection of the clipping limit value needs to be adjusted according to the characteristics of the steel wheel laser speckle interferometry image. A suitable value can be determined through experiments, for example, a value between 0.01 and 0.1 can be determined.

[0062] Small patch histogram equalization: For each small patch histogram that has been contrast-limited, its cumulative distribution function is calculated, and the corresponding new gray value is found through the cumulative distribution function. Specifically, the cumulative distribution function represents the proportion of the sum of the number of pixels of all gray levels before each gray level to the total number of pixels. According to the cumulative distribution function, the pixel gray values ​​in each small patch are mapped to make the distribution of gray values ​​after mapping more uniform, thereby achieving histogram equalization of the small patch.

[0063] Bilinear interpolation stitching: The small blocks after histogram equalization are stitched together. For pixels in the steel wheel laser speckle interferometry image that are not on the boundaries of the small blocks, the gray value is determined by bilinear interpolation. Based on the position of the pixel in the four adjacent small blocks, the weighting coefficient in each small block is calculated. The weighting coefficient is related to the distance from the pixel to the center of each small block. The closer the distance, the larger the weighting coefficient. Based on the weighting coefficient, the gray values ​​of the four adjacent small blocks are weighted and averaged to obtain the final gray value of the pixel.

[0064] Output Enhanced Image: After the above steps, an enhanced laser speckle interferometry image of the steel wheel is obtained. The contrast of this image is enhanced, and the details of the scratch area are more prominent, which is beneficial for subsequent scratch feature extraction and detection.

[0065] Step S30: Construct and train a deep learning-based model for extracting scratch features on the surface of a steel wheel. Input the enhanced laser speckle interferometry image of the steel wheel into the trained model for extracting scratch features on the surface of a steel wheel. The model outputs the scratch feature vector of the laser speckle interferometry image of the steel wheel.

[0066] Specifically, step S30, which involves constructing and training a deep learning-based model for extracting surface scratches on steel wheels, includes the following steps:

[0067] Construction of a feature extraction model for scratches on steel wheel surfaces: A convolutional neural network (CNN) model is constructed as the scratch feature extractor. This model contains multiple convolutional layers, pooling layers, and fully connected layers. The convolutional layers are used to extract local features from the scratch images on the steel wheel surface. The pooling layers are used to reduce the dimensionality of the feature maps and reduce the computational cost. The fully connected layers are used to classify the extracted features. For example, a CNN model containing 5 convolutional layers, 3 pooling layers, and 2 fully connected layers is constructed. The convolutional kernel sizes of the convolutional layers are 3×3, 5×5, 3×3, 5×5, and 3×3, respectively. The pooling layers use 2×2 max pooling, and the number of neurons in the fully connected layers are 128 and 64, respectively.

[0068] Training of the steel wheel surface scratch feature extraction model: A large number of laser speckle interferometric images of steel wheels with scratch annotations were collected as the training set to train the constructed steel wheel surface scratch feature extraction model. During the training process, the cross-entropy loss function was used as the optimization objective, and the stochastic gradient descent algorithm was used to update the model parameters and adjust the model weights so that the model can accurately extract scratch features. For example, 5000 laser speckle interferometric images of railway steel wheels with scratch annotations were collected as the training set, and the images were randomly divided into training set and validation set in a ratio of 8:2. The cross-entropy loss function was used, the learning rate was set to 0.001, and the model was trained for 50 epochs.

[0069] Feature extraction: The laser speckle interferometric image of the steel wheel processed in step S20 is input into the trained steel wheel surface scratch feature extraction model. The model outputs the scratch feature vector of the image. The output scratch feature vector includes information such as the shape, size and location of the scratch.

[0070] Step S40: By statistically analyzing the scratch features on the steel wheel surface, a scratch detection threshold is determined. The scratch feature vector of the model output image is compared with the set threshold. When the feature vector is greater than the threshold, it is determined that there is a scratch in the corresponding area of ​​the image; when the feature vector is less than or equal to the threshold, it is determined that there is no scratch in the corresponding area of ​​the image. For areas with scratches, the type of scratch is identified and classified according to the scratch feature vector. For example, it is determined whether the detected steel wheel scratch is a shallow scratch or a deep scratch.

[0071] Specifically, in step S40, the threshold for scratch detection is determined by statistically analyzing the scratch characteristics on the steel wheel surface, including:

[0072] Determine scratch judgment indicators: Select scratch length, width and gray scale change, which are closely related to scratches and can effectively reflect scratch characteristics, as scratch judgment indicators among the scratch features of steel wheel surface. Extract scratch feature vectors from all steel wheel laser speckle interferometry images in the training set, classify them according to key indicators, and form a separate dataset.

[0073] Mean and standard deviation calculation: The mean μ and standard deviation σ of each dataset are calculated using statistical methods. The mean represents the average level of the key indicator in the training set, reflecting the central tendency of the scratch features. The standard deviation represents the dispersion of the data. The larger the standard deviation, the greater the dispersion of the key indicator data; conversely, the smaller the standard deviation, the smaller the dispersion.

[0074] Threshold determination: Based on the actual detection situation, a multiple k of the standard deviation is set to adjust the leniency of the threshold. The larger the multiple, the higher the threshold, the stricter the detection conditions, and the lower the probability of false detection; the smaller the multiple, the lower the threshold, the more lenient the detection conditions, and the more minor scratches can be detected. The calculated mean plus the standard deviation of the selected multiple is the scratch detection threshold T, T = μ + kσ. For example, if the scratch feature vector of the training set image is statistically analyzed, and the mean of a certain key indicator in the feature vector is calculated to be 0.5 and the standard deviation is 0.1, then the threshold T = 0.5 + 2 × 0.1, T = 0.7, which is the mean plus 2 times the standard deviation.

[0075] In addition, such as Figure 2 As shown, in one embodiment of the present invention, a steel wheel surface scratch detection system is provided, the system comprising:

[0076] Steel wheel laser speckle interferometric image acquisition module: used to construct a laser speckle interferometric detection system to acquire steel wheel laser speckle interferometric images. The system includes a laser, beam expander, reflector, imaging lens and high-speed camera;

[0077] Steel wheel laser speckle interferometric image preprocessing module: The wavelet threshold denoising algorithm is used to denoise the acquired steel wheel laser speckle interferometric image, and the contrast-limited adaptive histogram equalization algorithm is used to enhance the denoised image.

[0078] Steel wheel surface scratch feature extraction model construction and training module: Construct and train a deep learning-based steel wheel surface scratch feature extraction model. Input the enhanced steel wheel laser speckle interferometry image into the trained steel wheel surface scratch feature extraction model, and the model outputs the scratch feature vector of the steel wheel laser speckle interferometry image.

[0079] Steel wheel surface scratch detection module: By statistically analyzing the scratch features of the steel wheel surface, a scratch detection threshold is determined. The scratch feature vector of the model output image is compared with the set threshold. When the feature vector is greater than the threshold, it is determined that there is a scratch in the corresponding area of ​​the image; when the feature vector is less than or equal to the threshold, it is determined that there is no scratch in the corresponding area of ​​the image. For areas with scratches, the type of scratch is identified and classified according to the scratch feature vector. For example, it is determined whether the detected steel wheel scratch is a shallow scratch or a deep scratch.

[0080] This application provides a steel wheel surface scratch detection system, employing a steel wheel surface scratch detection method described in the above embodiments, which solves the technical problems of low detection accuracy and low detection efficiency in traditional steel wheel surface scratch detection methods. Compared with the prior art, the beneficial effects of the steel wheel surface scratch detection system provided in this application are the same as those of the steel wheel surface scratch detection method provided in the above embodiments, and other technical features of the steel wheel surface scratch detection system are the same as those disclosed in the methods of the above embodiments, and will not be repeated here.

[0081] This application provides a steel wheel surface scratch detection device, which includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform a steel wheel surface scratch detection method as described in Embodiment 1 above.

[0082] like Figure 3As shown in the illustration, in one embodiment of the present invention, a structural schematic diagram of a steel wheel surface scratch detection device suitable for implementing the embodiments of this application is presented. The steel wheel surface scratch detection device in the embodiments of this application may include, but is not limited to, mobile terminals such as mobile phones, laptops, digital radio receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Description), PMPs (Portable Media Players), etc., and fixed terminals such as digital TVs, desktop computers, etc. Figure 3 The steel wheel surface scratch detection device shown is merely an example and should not impose any limitations on the functionality and scope of use of the embodiments of this application.

[0083] Figure 3 The illustrated steel wheel surface scratch detection device may include a processing system 1001 (e.g., a central processing unit, a graphics processing unit, etc.), which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 1002 or a program loaded from a storage system 1003 into a random access memory (RAM) 1004. The RAM 1004 also stores various programs and data required for the operation of the steel wheel surface scratch detection device. The processing system 1001, ROM 1002, and RAM 1004 are interconnected via a bus 1005. An input / output (I / O) interface 1006 is also connected to the bus. Typically, the following systems can be connected to I / O interface 1006: input systems 1007 including, for example, touchscreens, touchpads, keyboards, mice, image sensors, microphones, accelerometers, gyroscopes, etc.; output systems 1008 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; storage systems 1003 including, for example, magnetic tapes, hard disks, etc.; and communication systems 1009. Communication system 1009 allows a steel wheel surface scratch detection device to communicate wirelessly or wiredly with other devices to exchange data. Although a steel wheel surface scratch detection device with various systems is shown in the figure, it should be understood that it is not required to implement or possess all the systems shown. More or fewer systems may be implemented alternatively.

[0084] Specifically, according to the embodiments disclosed in this application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments disclosed in this application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication system, or installed from storage system 1003, or installed from ROM 1002. When the computer program is executed by processing system 1001, it performs the functions defined in the methods of the embodiments disclosed in this application.

[0085] This application provides a steel wheel surface scratch detection device, which employs a steel wheel surface scratch detection method described in the above embodiments, and can solve the technical problems of low detection accuracy and low detection efficiency of traditional steel wheel surface scratch detection methods. Compared with the prior art, the beneficial effects of the steel wheel surface scratch detection device provided in this application are the same as those of the steel wheel surface scratch detection method provided in the above embodiments, and other technical features of this steel wheel surface scratch detection device are the same as those disclosed in the previous embodiment method, and will not be repeated here.

[0086] The various parts disclosed in this application can be implemented using hardware, software, firmware, or a combination thereof. In the description of the above embodiments, specific features, structures, materials, or characteristics can be combined in any suitable manner in one or more embodiments or examples.

[0087] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the steel wheel surface scratch detection method described above.

[0088] The computer program product provided in this application can solve the technical problems of low detection accuracy and low detection efficiency of traditional steel wheel surface scratch detection methods. Compared with the prior art, the beneficial effects of the computer program product provided in this application are the same as those of the steel wheel surface scratch detection method provided in the above embodiments, and will not be repeated here.

[0089] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A method for detecting scratches on the surface of a steel wheel, characterized in that, The method includes the following steps: Step S10: Construct a laser speckle interferometry detection system to acquire laser speckle interferometry images of the steel wheel. The system includes a laser, a beam expander, a reflector, an imaging lens, and a high-speed camera. Step S20: The acquired steel wheel laser speckle interferometric image is denoised using a wavelet threshold denoising algorithm, and the denoised image is enhanced using a contrast-limited adaptive histogram equalization algorithm. Step S30: Construct and train a deep learning-based steel wheel surface scratch feature extraction model. Input the enhanced steel wheel laser speckle interferometry image into the trained steel wheel surface scratch feature extraction model. The model outputs the scratch feature vector of the steel wheel laser speckle interferometry image. Step S40: By statistically analyzing the scratch features on the steel wheel surface, the threshold for scratch detection is determined. The scratch feature vector of the model output image is compared with the set threshold. When the feature vector is greater than the threshold, it is determined that there is a scratch in the area corresponding to the image. When the feature vector is less than or equal to the threshold, it is determined that there is no scratch in the area corresponding to the image. For areas with scratches, the type of scratch is identified and classified according to the scratch feature vector. Step S10 involves constructing a laser speckle interferometry detection system to acquire laser speckle interferometry images of the steel wheel, including: Construct a laser speckle interferometry detection system: including a laser, a beam expander, a reflector, an imaging lens, and a high-speed camera. The laser beam emitted by the laser is expanded by the beam expander and then reflected by the reflector to illuminate the surface of a steel wheel. The scattered light from the surface of the steel wheel is imaged onto the photosensitive surface of the high-speed camera by the imaging lens. Laser speckle interferometry image acquisition of steel wheel: By applying vibration excitation to the surface of the steel wheel to cause deformation, a high-speed camera is used to acquire laser speckle interferometry images of the steel wheel surface before and after vibration. The frequency and amplitude of vibration excitation are adjusted according to the material and size of the steel wheel. In step S40, the threshold for scratch detection is determined by statistically analyzing the scratch characteristics on the steel wheel surface, including: Determine the scratch judgment index: Select the scratch length, width and gray scale change degree among the scratch features of the steel wheel surface as scratch judgment index. Extract the scratch feature vector of all steel wheel laser speckle interferometric images in the training set, classify them according to key indexes, and form a separate data set. Mean and standard deviation calculation: The mean μ and standard deviation σ of each data set are calculated using statistical methods. Threshold determination: Based on the actual detection situation, a multiple of standard deviation k is set to adjust the leniency of the threshold. The calculated mean is added to the standard deviation of the selected multiple to obtain the scratch detection threshold T, T=μ+kσ.

2. The method for detecting scratches on the surface of a steel wheel according to claim 1, characterized in that, The step S20, which uses a wavelet threshold denoising algorithm to denoise the acquired steel wheel laser speckle interferometric image, includes: Selected parameters: Based on the texture and noise characteristics of the steel wheel laser speckle interferometry image, the Symlets wavelet basis is selected. Based on the image size and noise of the steel wheel laser speckle interferometry image, the number of decomposition layers is set to 3. Wavelet decomposition: A two-dimensional discrete wavelet transform is performed on the laser speckle interferometry image of the steel wheel, which decomposes the image into a low-frequency approximate subband and high-frequency detail subbands in the horizontal, vertical and diagonal directions. After three layers of decomposition, three sets of high-frequency subbands and one low-frequency subband are obtained. Threshold determination: The SureShrink adaptive thresholding method is adopted to determine the threshold for each high-frequency sub-band. This method is based on the Stein unbiased likelihood estimation criterion and automatically adjusts the threshold according to the characteristics of the sub-band to adapt to the noise changes in different regions of the steel wheel image. Thresholding: The wavelet coefficients of each high-frequency sub-band are processed using a soft thresholding method. When the absolute value of a wavelet coefficient is greater than the threshold, it is shrunk; when it is less than or equal to the threshold, it is set to 0. Image reconstruction: Two-dimensional inverse discrete wavelet transform is performed on the processed wavelet coefficients to reconstruct the low-frequency subband and the processed high-frequency subband, resulting in a denoised steel wheel laser speckle interferometry image.

3. The method for detecting scratches on the surface of a steel wheel according to claim 1, characterized in that, The step S20, which involves enhancing the denoised image using a contrast-limited adaptive histogram equalization algorithm, includes: Image segmentation: The segmentation size is determined based on the size of the steel wheel laser speckle interferometry image, and the image is divided into a certain number of non-overlapping rectangular blocks, each of which is a local region of the image; Contrast limiting: For each small block, the grayscale histogram is calculated and a cropping limit value is set to control the maximum height of the histogram of each small block. The number of pixels exceeding the cropping limit value is cropped, and the cropped pixels are evenly redistributed to each grayscale level to obtain the grayscale histogram after contrast limiting. Small patch histogram equalization: For each small patch histogram that has been contrast-limited, calculate its cumulative distribution function, and find the corresponding new gray value through the cumulative distribution function; Bilinear interpolation stitching: The small blocks after histogram equalization are stitched together. For pixels in the steel wheel laser speckle interferometry image that are not on the boundary of the small blocks, the gray value is determined by bilinear interpolation. According to the position of the pixel in the four adjacent small blocks, the weighting coefficient in each small block is calculated. According to the weighting coefficient, the gray values ​​of the four adjacent small blocks are weighted and averaged to obtain the final gray value of the pixel. Output Enhanced Image: After the above steps, the enhanced laser speckle interference image of the steel wheel is obtained.

4. The method for detecting scratches on the surface of a steel wheel according to claim 1, characterized in that, The step S30, which involves constructing and training a deep learning-based steel wheel surface scratch feature extraction model, includes: Construction of a feature extraction model for scratches on steel wheel surfaces: A convolutional neural network (CNN) model is constructed as the scratch feature extractor. This model includes convolutional layers, pooling layers, and fully connected layers. The convolutional layers are used to extract local features of the scratch images on the steel wheel surface, the pooling layers are used to reduce the dimensionality of the feature maps and reduce the amount of computation, and the fully connected layers are used to classify the extracted features. Training of the steel wheel surface scratch feature extraction model: Collect laser speckle interferometry images of steel wheels with scratch annotations as the training set, and train the constructed steel wheel surface scratch feature extraction model. During the training process, the cross-entropy loss function is used as the optimization objective, and the stochastic gradient descent algorithm is used to update the model parameters and adjust the model weights. Feature extraction: Input the laser speckle interferometric image of the steel wheel processed in step S20 into the trained steel wheel surface scratch feature extraction model. The model outputs the scratch feature vector of the image. The output scratch feature vector includes the shape, size and location information of the scratch.

5. A steel wheel surface scratch detection system, characterized in that, The steel wheel surface scratch detection system includes: Steel wheel laser speckle interferometric image acquisition module: used to construct a laser speckle interferometric detection system to acquire steel wheel laser speckle interferometric images. The system includes a laser, beam expander, reflector, imaging lens and high-speed camera; Steel wheel laser speckle interferometric image preprocessing module: The wavelet threshold denoising algorithm is used to denoise the acquired steel wheel laser speckle interferometric image, and the contrast-limited adaptive histogram equalization algorithm is used to enhance the denoised image. Steel wheel surface scratch feature extraction model construction and training module: Construct and train a deep learning-based steel wheel surface scratch feature extraction model. Input the enhanced steel wheel laser speckle interferometry image into the trained steel wheel surface scratch feature extraction model, and the model outputs the scratch feature vector of the steel wheel laser speckle interferometry image. Steel wheel surface scratch detection module: By statistically analyzing the scratch features on the steel wheel surface, a scratch detection threshold is determined. The scratch feature vector of the model output image is compared with the set threshold. When the feature vector is greater than the threshold, it is determined that there is a scratch in the corresponding area of ​​the image; when the feature vector is less than or equal to the threshold, it is determined that there is no scratch in the corresponding area of ​​the image. For areas with scratches, the type of scratch is identified and classified according to the scratch feature vector. The laser speckle interferometry image acquisition module for steel wheels includes a laser speckle interferometry detection system for acquiring laser speckle interferometry images of steel wheels, comprising: Construct a laser speckle interferometry detection system: including a laser, a beam expander, a reflector, an imaging lens, and a high-speed camera. The laser beam emitted by the laser is expanded by the beam expander and then reflected by the reflector to illuminate the surface of a steel wheel. The scattered light from the surface of the steel wheel is imaged onto the photosensitive surface of the high-speed camera by the imaging lens. Laser speckle interferometry image acquisition of steel wheel: By applying vibration excitation to the surface of the steel wheel to cause deformation, a high-speed camera is used to acquire laser speckle interferometry images of the steel wheel surface before and after vibration. The frequency and amplitude of vibration excitation are adjusted according to the material and size of the steel wheel. The steel wheel surface scratch detection module determines the scratch detection threshold by statistically analyzing the scratch characteristics of the steel wheel surface, including: Determine the scratch judgment index: Select the scratch length, width and gray scale change degree among the scratch features of the steel wheel surface as scratch judgment index. Extract the scratch feature vector of all steel wheel laser speckle interferometric images in the training set, classify them according to key indexes, and form a separate data set. Mean and standard deviation calculation: The mean μ and standard deviation σ of each data set are calculated using statistical methods. Threshold determination: Based on the actual detection situation, a multiple of standard deviation k is set to adjust the leniency of the threshold. The calculated mean is added to the standard deviation of the selected multiple to obtain the scratch detection threshold T, T=μ+kσ.

6. A steel wheel surface scratch detection device, characterized in that, The steel wheel surface scratch detection device includes: A memory, a processor, and a steel wheel surface scratch detection program stored in the memory and executable on the processor, wherein the steel wheel surface scratch detection program, when executed by the processor, implements a steel wheel surface scratch detection method as described in any one of claims 1 to 4.

7. A computer program product, characterized in that, The computer program product includes a steel wheel surface scratch detection program, which, when executed by a processor, implements a steel wheel surface scratch detection method as described in any one of claims 1 to 4.

Images