Laser stripe detection method and system for steel components based on artificial intelligence algorithm

Abstract

The present invention discloses a steel component laser stripe detection method and system based on an artificial intelligence algorithm, and the method includes the following steps: S1, using a laser projection device to project a stripe pattern onto the surface of a steel component to be detected; S2, capturing a laser stripe image modulated on the surface of the steel component from at least two different angles; S3, preprocessing the captured laser stripe image to eliminate noise and enhance image contrast; S4, inputting the preprocessed image into a trained artificial intelligence model; S5, the artificial intelligence model outputs the defect detection result of the steel component surface; S6, generating a quality assessment report of the steel component. The present invention can automatically and efficiently identify and analyze various defects on the surface of steel components, which is of great significance for improving production efficiency, reducing labor costs, and ensuring product quality. Through image preprocessing, feature extraction, defect recognition and classification steps, high-precision detection and analysis of surface defects of steel components can be achieved.

Description

Technical Field

[0001] The present invention belongs to the field of laser stripe detection, and more specifically, it particularly relates to a steel component laser stripe detection method based on an artificial intelligence algorithm. At the same time, the present invention also relates to a steel component laser stripe detection system based on an artificial intelligence algorithm. Background Art

[0002] In modern manufacturing, quality control of steel components is a key link to ensure structural safety and functional reliability. Although traditional steel component inspection methods such as magnetic particle inspection and ultrasonic inspection can detect surface and internal defects to a certain extent, they have problems such as complex operation, high operator skill requirements, and low inspection efficiency. With the rapid development of artificial intelligence technology, especially the application of deep learning theory, new solutions have been provided for the automated inspection of steel components.

[0003] Laser stripe detection technology has shown unique advantages in steel component surface quality inspection due to its non-contact nature and high precision. By projecting laser stripes onto the surface of steel components and analyzing the surface flatness and defects based on the changes in the stripes, this method can quickly and accurately identify tiny surface defects. However, most existing laser stripe detection systems rely on simple image processing technology, which is difficult to cope with complex surface conditions and diverse defect types, and lack intelligence in data processing and result interpretation, which limits the further improvement of detection efficiency and accuracy. Therefore, we propose a laser stripe detection method and system for steel components based on artificial intelligence algorithms. Summary of the invention

[0004] The purpose of the present invention is to solve the shortcomings existing in the prior art, and the proposed steel component laser stripe detection method and system based on artificial intelligence algorithm can automatically and efficiently identify and analyze various defects on the surface of steel components, which is of great significance for improving production efficiency, reducing labor costs, and ensuring product quality. Combining laser stripe technology with deep learning models, through image preprocessing, feature extraction, defect recognition and classification steps, high-precision detection and analysis of surface defects of steel components can be achieved.

[0005] To achieve the above object, the present invention provides the following technical solutions:

[0006] The laser stripe detection method for steel components based on artificial intelligence algorithm includes the following steps:

[0007] S1. Use laser projection equipment to project a stripe pattern onto the surface of the steel component to be inspected;

[0008] S2, capturing the laser stripe images after being modulated on the surface of the steel component from at least two different angles;

[0009] S3, preprocessing the captured laser stripe image to eliminate noise and enhance image contrast;

[0010] S4, inputting the preprocessed image into a trained artificial intelligence model, which is a convolutional neural network based on deep learning, for identifying and analyzing anomalies and defects in the laser stripe pattern;

[0011] S5. The artificial intelligence model outputs the defect detection results on the surface of the steel component and classifies the defect type, location and size;

[0012] S6. Generate a quality assessment report for steel components based on the test results.

[0013] Preferably, step S1 specifically comprises:

[0014] S11, selecting a laser projection device that can produce high-quality laser stripe patterns;

[0015] S12. Clean the surface of the steel structure to ensure that there is no dust, oil or impurities that affect the projection of laser stripes;

[0016] S13. According to the size and shape of the steel structure, adjust the focal length and projection range of the laser projection equipment to ensure that the stripe pattern can be clearly projected onto the surface of the steel structure, and set the density and spacing of the laser stripes. These parameters will vary depending on the type and size of the defects to be detected;

[0017] S14. Turn on the laser projection equipment to project the stripe pattern onto the surface of the steel structure, ensuring that the stripe pattern is evenly distributed and covers all areas that need to be inspected. Project multiple stripe patterns from different angles or positions to obtain more comprehensive surface coverage.

[0018] Preferably, step S3 pre-processes the captured laser stripe image, specifically:

[0019] S31, using bilateral filtering to preprocess the image, smoothing the image while keeping the edge clear, thereby reducing noise interference;

[0020] S32, replacing the pixel value with the median value in the pixel neighborhood can effectively remove isolated noise points;

[0021] S33, by calculating the grayscale histogram of the image, the brightness distribution of the entire image is obtained, and then the image is segmented respectively, and finally an accurate laser stripe area is obtained;

[0022] S34. Use the Gaussian fitting method to adjust the brightness distribution of the laser stripes so that the difference between the laser stripes and the background is more significant.

[0023] Preferably, the artificial intelligence model described in step S4 is composed of several layers, each layer contains several feature maps, and each feature map is connected to the feature map of the previous layer through a convolution kernel, and the expression is:

[0024]

[0025] Where M l and N l Respectively represent the size of the feature map of the lth layer, w ijmn is the convolution kernel weight connecting the mth and nth nodes in the (l-1)th layer with the ith and jth nodes in the lth layer, b ij is the bias term, tanh is the activation function, v ij are the nodes of the i-th and j-th feature maps of the l-th layer.

[0026] Preferably, during the calculation process, the artificial intelligence model is forward propagated to the output layer through continuous convolution, nonlinear activation, and pooling operations. For the classification task, each node in the output layer corresponds to the predicted probability of a category, and the calculation formula is:

[0027]

[0028] Where ac is the forward propagation result of the cth node in the output layer, C is the total number of categories, W and b are the weight and bias parameters of the network respectively, P(Y=c|X;W,b) is the probability that the output variable Y is equal to a specific value c given the input X and the model parameters W and b, and ak is the score calculated based on the input X and the model parameters W and b;

[0029] During the training of the artificial intelligence model, the network parameters are updated to minimize the loss by calculating the gradient of the loss function with respect to the weights and biases. The expression is:

[0030]

[0031] In the above formula, w old and b old are the weight and bias parameters before updating, w new and b new is the updated parameter, α is the learning rate, and They are the gradients of the loss function L with respect to the weight w and b bias respectively;

[0032] In a multivariable function, the gradient is a vector whose direction is the direction in which the function value increases most rapidly and whose magnitude is equal to the directional derivative in that direction.

[0033] A laser stripe detection system for steel components based on an artificial intelligence algorithm, the system is used to implement the above method, and the system comprises:

[0034] The laser projection module is responsible for projecting high-quality stripe patterns onto the surface of the steel component to be inspected; this module includes the functions of selecting the laser projection equipment, cleaning the surface of the steel component, adjusting the focal length and projection range, and actually projecting the stripe pattern;

[0035] An image capture module captures the laser stripe images modulated on the surface of the steel component from at least two different angles; this module ensures that the image can fully cover the area to be inspected and adapts to the shooting requirements at different angles;

[0036] The preprocessing module performs noise elimination and contrast enhancement preprocessing operations on the captured laser stripe image; specifically, it includes bilateral filtering, median filtering, histogram calculation and segmentation, and Gaussian fitting steps;

[0037] The deep learning analysis module inputs the pre-processed images into a trained artificial intelligence model; this model is a deep learning-based convolutional neural network specially designed to identify and analyze anomalies and defects in laser stripe patterns.

[0038] Preferably, the system further comprises:

[0039] Defect detection and classification module: The artificial intelligence model outputs the defect detection results of the steel component surface and classifies the defect type, location and size. This module uses the model's forward propagation algorithm and loss function minimization to achieve defect detection and classification.

[0040] The quality assessment report generation module automatically generates a quality assessment report for steel components based on the test results; the report reflects the defects of the steel components in detail, providing a basis for subsequent quality judgment and processing.

[0041] Preferably, the laser projection module includes a multi-wavelength laser source, which can select the most suitable laser wavelength according to the material of the steel component and the expected detection requirements;

[0042] The laser projection module also integrates adjustable optical elements, allowing the operator to adjust the focus and coverage of the laser beam according to the specific shape and size of the steel structure. The adjustment mechanism ensures that the laser stripes are evenly distributed on steel structures with complex geometries, improving the accuracy and flexibility of detection.

[0043] Preferably, the image capture module is composed of several groups of high-resolution cameras, which are installed on a robotic arm with multiple degrees of freedom. The robotic arm can move freely in space, so that the camera can shoot the steel structure from multiple angles and positions. By precisely controlling the movement of the robotic arm, the system can automatically plan the optimal shooting path to ensure that the laser stripe pattern is fully and evenly captured without human intervention. The coordinated image stitching algorithm can generate a panoramic image of the steel structure in real time, providing complete visual information for subsequent defect detection.

[0044] The steel component laser stripe detection method and system based on artificial intelligence algorithm provided by the present invention have the following advantages compared with the prior art:

[0045] High efficiency and high accuracy. The deep learning model can quickly and accurately identify and classify various defects on the surface of steel components, greatly improving the detection efficiency and accuracy;

[0046] Non-contact detection: laser stripe detection is non-contact in nature, which avoids secondary damage to the surface of steel components and reduces the error caused by contact detection;

[0047] High degree of automation: From laser stripe projection to final defect identification and classification, the entire process is automated, reducing dependence on manual operations and reducing labor costs.

[0048] Strong adaptability: the system can adapt to steel components of different shapes, sizes and materials, and has good versatility and adaptability;

[0049] Continuous optimization supported by data: the system can continuously optimize the deep learning model based on the detection data, and continuously improve the accuracy and robustness of detection;

[0050] In addition, by introducing artificial intelligence algorithms, the patented steel component laser stripe detection method and system have significant technical advantages and market potential in improving production efficiency, reducing costs and improving product quality, providing a new solution for quality control in modern manufacturing. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 This is a flow chart of the laser stripe detection method for steel components based on artificial intelligence algorithm of the present invention;

[0052] Figure 2 This is a flow chart of preprocessing a captured laser stripe image in an embodiment of the present invention;

[0053] Figure 3 This is the architecture diagram of the steel structure laser stripe detection system based on artificial intelligence algorithm. DETAILED DESCRIPTION

[0054] In order to make the purpose, technical scheme and advantages of the present invention clearer, the present invention is further described in detail below in conjunction with specific embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0055] The present invention provides a laser stripe detection method and system for steel components based on artificial intelligence algorithms, which can automatically and efficiently identify and analyze various defects on the surface of steel components, and is of great significance for improving production efficiency, reducing labor costs, and ensuring product quality. Combining laser stripe technology with deep learning models, through image preprocessing, feature extraction, defect recognition and classification steps, high-precision detection and analysis of surface defects of steel components can be achieved;

[0056] like Figure 1 As shown, the method specifically includes the following steps:

[0057] S1, using a laser projection device to project a stripe pattern onto the surface of the steel component to be inspected; step S1 is specifically as follows:

[0058] S11, selecting a laser projection device that can produce high-quality laser stripe patterns;

[0059] S12. Clean the surface of the steel structure to ensure that there is no dust, oil or impurities that affect the projection of laser stripes;

[0060] S13. According to the size and shape of the steel structure, adjust the focal length and projection range of the laser projection equipment to ensure that the stripe pattern can be clearly projected onto the surface of the steel structure, and set the density and spacing of the laser stripes. These parameters will vary depending on the type and size of the defects to be detected;

[0061] S14. Turn on the laser projection equipment to project the stripe pattern onto the surface of the steel structure, ensuring that the stripe pattern is evenly distributed and covers all areas that need to be inspected. Project multiple stripe patterns from different angles or positions to obtain more comprehensive surface coverage.

[0062] S2, capturing the laser stripe images after being modulated on the surface of the steel component from at least two different angles;

[0063] For example, after the laser stripe patterns are projected onto the surface of the steel component, images of these patterns need to be acquired from multiple angles by an image capture device to ensure that the system receives information from different perspectives, thereby more comprehensively evaluating the surface condition of the steel component. When operating this step, the following key points are included:

[0064] Angle selection: When selecting shooting angles, you must ensure that each angle can cover different sides of the steel component, especially those hidden parts that are difficult to observe. Generally, it is recommended to select the basic angles of front view, side view and top view, as well as other specific angles that may be required according to the shape of the component.

[0065] Lighting conditions: ensure that lighting conditions remain consistent during the shooting process; using additional light sources can help maintain image brightness consistency and reduce image quality differences caused by lighting changes;

[0066] Image clarity: adjust the camera’s focus and aperture to ensure that each image is clear enough for subsequent analysis; blurry images may affect the accuracy of defect detection;

[0067] Image synchronization: if multiple cameras are used to shoot synchronously, it is necessary to ensure the time synchronization between images to facilitate subsequent image stitching and comparative analysis;

[0068] Data storage: The amount of data from high-resolution images captured may be large, requiring effective data management and storage solutions to ensure data security and fast access.

[0069] S3, preprocessing the captured laser stripe image to eliminate noise and enhance image contrast; Figure 2 As shown, step S3 pre-processes the captured laser stripe image, specifically:

[0070] S31, using bilateral filtering to preprocess the image, smoothing the image while keeping the edge clear, thereby reducing noise interference;

[0071] S32, replacing the pixel value with the median value in the pixel neighborhood can effectively remove isolated noise points;

[0072] S33, by calculating the grayscale histogram of the image, the brightness distribution of the entire image is obtained, and then the image is segmented respectively, and finally an accurate laser stripe area is obtained;

[0073] S34. Use the Gaussian fitting method to adjust the brightness distribution of the laser stripes so that the difference between the laser stripes and the background is more significant.

[0074] S4, input the preprocessed image into a trained artificial intelligence model, which is a convolutional neural network based on deep learning, used to identify and analyze anomalies and defects in laser stripe patterns; the artificial intelligence model described in step S4 is composed of several layers, each layer contains several feature maps, and each feature map is connected to the feature map of the previous layer through a convolution kernel, and the expression is:

[0075]

[0076] Where M l and N l Respectively represent the size of the feature map of the lth layer, w ijmn is the convolution kernel weight connecting the mth and nth nodes in the (l-1)th layer with the ith and jth nodes in the lth layer, b ij is the bias term, tanh is the activation function, v ij are the nodes of the i-th and j-th feature graphs of the l-th layer;

[0077] During the calculation process of the artificial intelligence model, forward propagation is passed to the output layer through continuous convolution, nonlinear activation, and pooling operations. For classification tasks, each node in the output layer corresponds to the predicted probability of a category, and the calculation formula is:

[0078]

[0079] Where ac is the forward propagation result of the cth node in the output layer, C is the total number of categories, W and b are the weight and bias parameters of the network respectively, P(Y=c|X;W,b) is the probability that the output variable Y is equal to a specific value c given the input X and the model parameters W and b, and ak is the score calculated based on the input X and the model parameters W and b;

[0080] During the training of the artificial intelligence model, the network parameters are updated to minimize the loss by calculating the gradient of the loss function with respect to the weights and biases. The expression is:

[0081]

[0082] In the above formula, w old and b old are the weight and bias parameters before updating, w new and b new is the updated parameter, α is the learning rate, and They are the gradients of the loss function L with respect to the weight w and b bias respectively;

[0083] In a multivariable function, the gradient is a vector whose direction is the direction in which the function value increases fastest and whose magnitude is equal to the directional derivative in that direction;

[0084] Exemplarily, the artificial intelligence model includes the following hierarchical structure:

[0085] Convolutional layer 1: uses 32 5×5 kernels, ReLU activation function, stride 1, and no padding;

[0086] Pooling layer 1: 2×2 maximum pooling is used to reduce the feature map size with a step size of 2;

[0087] Convolutional layer 2: uses 64 3×3 convolution kernels and the activation function is ReLU;

[0088] Pooling layer 2: 2×2 maximum pooling is used again;

[0089] Fully connected layer 1: has 128 neurons and the activation function is ReLU;

[0090] Fully connected layer 2: output layer, the number of nodes is equal to the total number of defect categories, and the Softmax activation function is used to output the probability of each category;

[0091] Based on the above, the input image is a 28×28 single-channel image, which we can represent as I, where the size of I is 28×28;

[0092] In the calculation of convolution layer 1, the size of each convolution kernel is 5×5, and the number is 32, which is expressed as:

[0093]

[0094] K is the convolution kernel weight, b is the bias term, and O is the output feature map; since the stride is 1 and there is no padding, the size of the output feature map will become:

[0095] (28-5+1)×(28-5+1)=24×24(28-5+1)×(28-5+1)=24×24;

[0096] The calculation of activation layer 1 uses the ReLU activation function, and the calculation formula is:

[0097] ReLU(x)=max(0,x);

[0098] Apply the ReLU activation function to each convolution output to obtain the activated feature map;

[0099] The calculation of pooling layer 1 is 2×2 maximum pooling with a step size of 2 and no padding, expressed as:

[0100]

[0101] The output size after pooling is (24 / 2)×(24 / 2)=12×12(24 / 2)×(24 / 2)=12×12;

[0102] Calculation of the output layer, assuming that the output layer is a fully connected layer, the output of the pooling layer is flattened and connected to a layer with multiple neurons. The output of each neuron is expressed as:

[0103]

[0104] Among them, w i is the weight, xi is the input, b is the bias term, and N is the number of input features;

[0105] Finally, the Softmax activation function is used to calculate the final classification probability:

[0106]

[0107] In the above, i, j, m, n are variables used for indexing, usually used to specify the position in a matrix or array. In our problem, they are used to specify a specific position in the input image I and the convolution kernel K;

[0108] l represents the current level, that is, the convolutional layer, K m+2,n+2 It is an element in the convolution kernel matrix, and its specific position is determined by m+2, n+2;

[0109] represents the value of the output feature map of the lth convolutional layer at position (i, j), obtained by sliding the convolution kernel K over the input image I and calculating the dot product;

[0110] is the output of the pooling layer, which is obtained by applying the maximum pooling operation to the output of the convolutional layer. In the above, the maximum value of the 2×2 region is taken, and then the stride of this 2×2 region is set to 2;

[0111] y k is the neuron output of the fully connected layer, which is calculated by converting the input vector x i With weight w i Multiply, and add bias b;

[0112] S(y) k It is the output of the Softmax function and is used for multi-class classification problems.

[0113] S5. The artificial intelligence model outputs the defect detection results on the surface of the steel component and classifies the defect type, location and size;

[0114] Specifically, after image preprocessing and deep learning model analysis, the model will output the defect detection results of the steel component surface, specifically:

[0115] Result interpretation: The output of the model includes the type of defects, including cracks, scratches, and dents; location and length, width, and depth;

[0116] Accuracy verification: Verify the output of the model to ensure its accuracy. Compare and analyze the results with those of manual inspection or use samples with known defects for cross-validation;

[0117] Real-time feedback: When running the model in real time on the production line, the test results need to be fed back to the operator or production management system in a timely manner so that necessary adjustments or repair measures can be taken immediately.

[0118] Model iteration: Based on the feedback from the test results, the AI ​​model is continuously optimized and adjusted to improve its detection accuracy and reliability.

[0119] S6. Generate a quality assessment report for steel components based on the test results;

[0120] Specifically, the report will provide a documentary basis for the quality control of steel components and is also an important reference for subsequent engineering applications;

[0121] Report content: The report should record in detail the date, time, tester, equipment and software version used for each test, as well as specific information about the defects, including type, location, and size. In addition, it should include an assessment of the overall quality of the steel structure, as well as recommendations and remedial measures for any problems found.

[0122] Data visualization: The report includes charts, images and other visualization elements to intuitively show the distribution and severity of defects. Use heat maps to show the distribution of defects on steel components;

[0123] Report distribution: Ensure that reports are distributed to all relevant parties in a timely and secure manner, including quality control departments, production management and customers; digital reports can reduce the use of paper documents and improve the efficiency of information sharing.

[0124] The present invention also proposes a steel component laser stripe detection system based on artificial intelligence algorithm, which is used to implement the above method, such as Figure 3 As shown, specifically including:

[0125] The laser projection module is responsible for projecting high-quality stripe patterns onto the surface of the steel component to be inspected. This module includes the selection of laser projection equipment, cleaning of the steel component surface, adjustment of the focal length and projection range, and the actual projection of the stripe pattern. The laser projection module includes a multi-wavelength laser source, which can select the most appropriate laser wavelength according to the material of the steel component and the expected inspection requirements.

[0126] The laser projection module also integrates adjustable optical elements, allowing the operator to adjust the focus and coverage of the laser beam according to the specific shape and size of the steel structure. The adjustment mechanism ensures that the laser stripes are evenly distributed on steel structures with complex geometries, improving the accuracy and flexibility of detection.

[0127] The image capture module captures the laser stripe image after modulating the surface of the steel component from at least two different angles; the module ensures that the image can fully cover the area to be inspected and adapts to the shooting requirements at different angles; the image capture module is composed of several groups of high-resolution cameras, which are installed on a robotic arm with multiple degrees of freedom. The robotic arm can move freely in space, so that the camera can shoot the steel component from multiple angles and positions. By precisely controlling the movement of the robotic arm, the system can automatically plan the optimal shooting path to ensure that the laser stripe pattern is fully and evenly captured without human intervention. The coordinated image stitching algorithm can generate a panoramic image of the steel component in real time, providing complete visual information for subsequent defect detection.

[0128] The preprocessing module performs noise elimination and contrast enhancement preprocessing operations on the captured laser stripe image; specifically, it includes bilateral filtering, median filtering, histogram calculation and segmentation, and Gaussian fitting steps;

[0129] A deep learning analysis module that feeds the pre-processed images into a trained AI model; the model is a deep learning-based convolutional neural network specifically designed to identify and analyze anomalies and defects in laser stripe patterns;

[0130] Defect detection and classification module: The artificial intelligence model outputs the defect detection results of the steel component surface and classifies the defect type, location and size. This module uses the model's forward propagation algorithm and loss function minimization to achieve defect detection and classification. The defect detection and classification module is the application end of the system. It detects, classifies and locates defects on the surface of steel components based on the output results of the deep learning model.

[0131] Defect detection: The model can identify whether there are defects and locate the defective areas by learning the surface characteristics of steel components;

[0132] Defect classification: The system can not only detect the presence of defects, but also classify them according to their shape, size, and texture characteristics, such as cracks, scratches, and dents;

[0133] Dimension measurement: For detected defects, the system can measure their specific dimensions to provide accurate data support for subsequent repair work;

[0134] Optimized results: Through continuous iterative training and model optimization, the defect detection and classification module can adapt to the surfaces of steel components of different types and complexities, improving the accuracy and reliability of detection;

[0135] The quality assessment report generation module automatically generates a quality assessment report for steel components based on the test results; the report reflects the defects of the steel components in detail, providing a basis for subsequent quality judgment and processing;

[0136] The quality assessment report generation module is the output end of the system, which automatically generates a detailed quality assessment report based on the test results;

[0137] Report content: The report records the defects of the steel structure in detail, including the type, location, size, possible impact and recommended repair measures of the defects;

[0138] Data visualization: The report contains charts and images to intuitively display the distribution and severity of defects, making it easier for users to understand and make decisions;

[0139] Report format: The report supports multiple output formats, such as PDF and Word, for easy archiving and transmission;

[0140] Customized service: According to different user needs, the report can be customized with specific content and format to meet application requirements in different scenarios.

[0141] In summary, the present invention has the following advantages:

[0142] Advantages include:

[0143] High efficiency and high accuracy. The deep learning model can quickly and accurately identify and classify various defects on the surface of steel components, greatly improving the detection efficiency and accuracy;

[0144] Non-contact detection: laser stripe detection is non-contact in nature, which avoids secondary damage to the surface of steel components and reduces the error caused by contact detection;

[0145] High degree of automation: From laser stripe projection to final defect identification and classification, the entire process is automated, reducing dependence on manual operations and reducing labor costs.

[0146] Strong adaptability: the system can adapt to steel components of different shapes, sizes and materials, and has good versatility and adaptability;

[0147] Continuous optimization supported by data: the system can continuously optimize the deep learning model based on the detection data, and continuously improve the accuracy and robustness of detection;

[0148] In addition, by introducing artificial intelligence algorithms, the patented steel component laser stripe detection method and system have significant technical advantages and market potential in improving production efficiency, reducing costs and improving product quality, providing a new solution for quality control in modern manufacturing.

[0149] Finally, it should be noted that the above is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, it is still possible for those skilled in the art to modify the technical solutions described in the aforementioned embodiments, or to make equivalent substitutions for some of the technical features therein. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A laser stripe detection method for steel components based on artificial intelligence algorithm, characterized in that: The steps include: S1. Use laser projection equipment to project a stripe pattern onto the surface of the steel component to be inspected; S2, capturing the laser stripe images after being modulated on the surface of the steel component from at least two different angles; S3, preprocessing the captured laser stripe image to eliminate noise and enhance image contrast; S4, input the preprocessed image into a trained artificial intelligence model, which is a convolutional neural network based on deep learning, used to identify and analyze anomalies and defects in laser stripe patterns; the artificial intelligence model described in step S4 is composed of several layers, each layer contains several feature maps, and each feature map is connected to the feature map of the previous layer through a convolution kernel, and the expression is: Where M l and N l Respectively represent the size of the feature map of the lth layer, w ijmn is the convolution kernel weight connecting the mth and nth nodes in the (l-1)th layer with the ith and jth nodes in the lth layer, b ij is the bias term, tanh is the activation function, v ij are the nodes of the i-th and j-th feature graphs of the l-th layer; During the calculation process, the artificial intelligence model is forward propagated to the output layer through continuous convolution, nonlinear activation, and pooling operations. For classification tasks, each node in the output layer corresponds to the predicted probability of a category, and the calculation formula is: Where ac is the forward propagation result of the cth node in the output layer, C is the total number of categories, W and b are the weight and bias parameters of the network respectively, P(Y=c|X;W,b) is the probability that the output variable Y is equal to a specific value c given the input X and the model parameters W and b, and ak is the score calculated based on the input X and the model parameters W and b; During the training of the artificial intelligence model, the network parameters are updated to minimize the loss by calculating the gradient of the loss function with respect to the weights and biases. The expression is: In the above formula, w old and b old are the weight and bias parameters before updating, w new and b new is the updated parameter, α is the learning rate, and They are the gradients of the loss function L with respect to the weight w and b bias respectively; In a multivariable function, the gradient is a vector whose direction is the direction in which the function value increases fastest and whose magnitude is equal to the directional derivative in that direction; S5. The artificial intelligence model outputs the defect detection results on the surface of the steel component and classifies the defect type, location and size; S6. Generate a quality assessment report for steel components based on the test results.

2. The steel component laser stripe detection method based on artificial intelligence algorithm according to claim 1 is characterized in that: Step S1 is specifically as follows: S11, selecting a laser projection device that can produce high-quality laser stripe patterns; S12. Clean the surface of the steel structure to ensure that there is no dust, oil or impurities that affect the projection of laser stripes; S13. According to the size and shape of the steel structure, adjust the focal length and projection range of the laser projection equipment to ensure that the stripe pattern can be clearly projected onto the surface of the steel structure, and set the density and spacing of the laser stripes. These parameters will vary depending on the type and size of the defects to be detected; S14. Turn on the laser projection equipment to project the stripe pattern onto the surface of the steel structure, ensuring that the stripe pattern is evenly distributed and covers all areas that need to be inspected. Project multiple stripe patterns from different angles or positions to obtain more comprehensive surface coverage.

3. The steel component laser stripe detection method based on artificial intelligence algorithm according to claim 1 is characterized in that: Step S3 preprocesses the captured laser stripe image, specifically: S31, using bilateral filtering to preprocess the image, smoothing the image while keeping the edge clear, thereby reducing noise interference; S32, replacing the pixel value with the median value in the pixel neighborhood can effectively remove isolated noise points; S33, by calculating the grayscale histogram of the image, the brightness distribution of the entire image is obtained, and then the image is segmented respectively, and finally an accurate laser stripe area is obtained; S34. Use the Gaussian fitting method to adjust the brightness distribution of the laser stripes so that the difference between the laser stripes and the background is more significant.

4. A laser stripe detection system for steel components based on an artificial intelligence algorithm, the system being used to implement the method described in any one of claims 1 to 3, characterized in that: The system comprises: The laser projection module is responsible for projecting high-quality stripe patterns onto the surface of the steel component to be inspected; this module includes the functions of selecting the laser projection equipment, cleaning the surface of the steel component, adjusting the focal length and projection range, and actually projecting the stripe pattern; An image capture module captures the laser stripe images modulated on the surface of the steel component from at least two different angles; this module ensures that the image can fully cover the area to be inspected and adapts to the shooting requirements at different angles; The preprocessing module performs noise elimination and contrast enhancement preprocessing operations on the captured laser stripe image; specifically, it includes bilateral filtering, median filtering, histogram calculation and segmentation, and Gaussian fitting steps; The deep learning analysis module inputs the pre-processed images into a trained artificial intelligence model; this model is a deep learning-based convolutional neural network specially designed to identify and analyze anomalies and defects in laser stripe patterns.

5. The steel component laser stripe detection system based on artificial intelligence algorithm according to claim 4 is characterized in that: The system further comprises: Defect detection and classification module: The artificial intelligence model outputs the defect detection results of the steel component surface and classifies the defect type, location and size. This module uses the model's forward propagation algorithm and loss function minimization to achieve defect detection and classification. The quality assessment report generation module automatically generates a quality assessment report for steel components based on the test results; the report reflects the defects of the steel components in detail, providing a basis for subsequent quality judgment and processing.

6. The steel component laser stripe detection system based on artificial intelligence algorithm according to claim 4 is characterized in that: The laser projection module includes a multi-wavelength laser source, which can select the most suitable laser wavelength according to the material of the steel component and the expected detection requirements; The laser projection module also integrates adjustable optical elements, allowing the operator to adjust the focus and coverage of the laser beam according to the specific shape and size of the steel structure. The adjustment mechanism ensures that the laser stripes are evenly distributed on steel structures with complex geometries, improving the accuracy and flexibility of detection.

7. The steel component laser stripe detection system based on artificial intelligence algorithm according to claim 4 is characterized in that: The image capture module is composed of several groups of high-resolution cameras, which are installed on a robotic arm with multiple degrees of freedom. The robotic arm can move freely in space, so that the camera can shoot steel components from multiple angles and positions. By precisely controlling the movement of the robotic arm, the system can automatically plan the optimal shooting path to ensure that the laser stripe pattern is captured comprehensively and evenly without human intervention. The coordinated image stitching algorithm can generate panoramic images of steel components in real time, providing complete visual information for subsequent defect detection.

Images