Cover beam support deformation detection method and system

Abstract

The invention relates to the technical field of bridge detection, and discloses a cover beam support deformation detection method and system, and the method comprises the steps: firstly obtaining to-be-detected cover beam support image data through an image collection device, and then carrying out the preprocessing, including denoising, enhancement and graying processing, of an image, so as to improve the image quality. Then, an advanced YOLOX neural network framework is adopted to construct a deformation recognition model, the model is optimized especially for the deformation characteristics of the bent cap support, and the deformation characteristics can be accurately extracted and effectively classified and quantified; and finally, inputting the preprocessed image into a model for deformation identification, so as to obtain detailed information such as deformation position, type and degree of the bent cap bracket. According to the method, real-time monitoring and accurate recognition of deformation of the bent cap support are achieved, the construction safety and the quality monitoring level are effectively improved, and powerful technical support is provided for the field of bridge construction.

Description

Technical Field

[0001] The present invention relates to the technical field of bridge detection, and in particular to a method and system for detecting deformation of a cap beam support. Background Art

[0002] In the field of construction engineering, cap beam supports, as important temporary support structures, play a vital role in the bridge construction process. However, due to various complex factors in the construction process, such as load changes, environmental factors, and material aging, the cap beam supports may deform, which not only affects the construction quality and safety, but may also lead to serious engineering accidents. Therefore, it is very important to detect the deformation of the cap beam supports in a timely and accurate manner.

[0003] At present, the deformation detection method of the cap beam support mainly adopts simple measuring tools, which can only measure certain specific points of the support and cannot fully reflect the overall deformation of the support. In addition, these methods can usually only perform passive detection after the deformation occurs, and cannot monitor the deformation process of the support in real time, making it difficult to discover potential safety hazards in time.

[0004] With the rapid development of computer vision and artificial intelligence technology, deformation detection methods based on image processing have gradually emerged. However, there are still some problems with existing image processing methods when applied to deformation detection of cap beam supports. For example, due to the complex construction site environment, images often contain a lot of noise and interference information, which increases the difficulty of deformation identification. At the same time, the deformation types and degrees of deformation of cap beam supports are diverse. How to accurately extract deformation features and effectively classify and quantify them is a problem that needs to be solved urgently. Summary of the invention

[0005] The purpose of the present invention is to provide a method and system for detecting deformation of a cap beam support to solve the problems raised in the above-mentioned background technology.

[0006] To achieve the above object, the present invention provides the following technical solution: a method for detecting deformation of a cap beam support, the method comprising: S1. Using an image acquisition device to obtain image data of the cap beam support to be inspected; S2, preprocessing the collected cap beam support image data, including image denoising, image enhancement and image grayscale; S3. A cap beam support deformation recognition model is constructed using a YOLOX neural network framework. The model is used to identify the deformation of the cap beam support in the image. The model includes a deformation feature extraction module, and the loss function in the original YOLOX neural network is optimized for the deformation characteristics of the cap beam support. S4. Input the cap beam support image preprocessed in step S2 into the trained cap beam support deformation recognition model to identify the deformation of the cap beam support and obtain the deformation information of the cap beam support, including the deformation position, deformation type and deformation degree.

[0007] Preferably, the implementation steps of the deformation feature extraction module include: The convolutional block module is used to preliminarily process the input cap beam support image features, and some features are used as direct paths for subsequent splicing through feature splitting operations; The remaining features are fed into a main path consisting of multiple deformable convolutional layers to capture the complex deformation characteristics of the cap beam support; In the main path, the outputs of each deformable convolutional layer are feature fused to enhance the model’s sensitivity to deformation features. The features of the direct path and the main path are concatenated, and the number of channels is adjusted through the convolution block module to achieve comprehensive feature fusion.

[0008] Preferably, the cap beam support deformation recognition model optimizes the loss function in the original YOLOX neural network for the deformation characteristics of the cap beam support, specifically including: using the improved center point IoU loss to replace the original IoU loss as the positioning loss, which additionally considers the center point distance and aspect ratio factors when calculating the overlap between the prediction box and the actual deformation area, and its algorithm is: in, is the Euclidean distance between the predicted box and the true deformed area, is the diagonal length of the minimum bounding rectangle covering the predicted box and the true deformation area, is a weight function used to adjust the effect of aspect ratio consistency on the loss, It is an indicator to measure the consistency of aspect ratio.

[0009] Preferably, the adjusted focal loss is used to replace the original binary cross entropy loss as the classification loss. The calculation formula of the focal loss is: in, is the probability of predicting a positive sample, is the factor used to balance the weights of positive and negative samples, It is an adjustable focusing parameter used to adjust the degree of attention paid to samples that are difficult to classify.

[0010] Preferably, the Gaussian filtering method is used to eliminate the noise in the image, and the value of each pixel in the image is replaced by the weighted average of the pixel values ​​in its neighborhood. The calculation formula is: Where G(x,y) is the value of the Gaussian function at the two-dimensional coordinate (x,y) of the image, x and y are the coordinates in two-dimensional space, σ is the standard deviation of the Gaussian distribution, e is the base of the natural logarithm, and π is the circumference of a circle.

[0011] Preferably, the steps of implementing image enhancement on the cap beam support image include: Step I: Use the Sobel detection operator to perform convolution on the image and calculate the gradient intensity and direction of each pixel; Step II: traverse each pixel in the image, compare its gradient strength with the gradient strength of the adjacent pixels in its gradient direction, and if the current pixel is not a local maximum, set its gradient strength to zero, thereby eliminating non-contour pixels and reducing identification artifacts; Step III: Set two thresholds: a high threshold and a low threshold. Pixels with gradient strengths higher than the high threshold are considered strong contour points, and pixels with gradient strengths lower than the low threshold are considered non-contour points. For pixels with gradient strengths between the two thresholds, if they are connected to strong contour points, they are considered weak contour points and retained; otherwise, they are considered non-contour points and removed.

[0012] Preferably, the method for grayscale processing of the cap beam support image is: a. Select the weighted method to convert the color image into a grayscale image, and use the formula GRAY = 0.3*R + 0.59*G +0.11*B to calculate the grayscale value. R, G, and B in the formula represent the components of the pixel in the red, green, and blue color domains respectively; b. Read the cap beam support image data, traverse each pixel in the image, and obtain the RGB color value corresponding to each pixel; c. Apply the color value conversion method, call the formula in step a to convert the RGB color value of each pixel obtained in step b to obtain the corresponding grayscale value; d. Apply the converted grayscale value to each pixel to generate a grayscale image of the cap beam support image.

[0013] Preferably, a deformation report is generated based on the deformation information of the cap beam support obtained in step S4, wherein the report includes the precise coordinates of the deformation position, a detailed description of the deformation type, quantitative data of the deformation degree, and recommended repair or replacement measures.

[0014] Preferably, the image acquisition device is a high-definition camera carried by a drone, and the image is acquired by flying the drone above the cap beam support.

[0015] Preferably, a cap beam support deformation detection system comprises: An image acquisition module, used to obtain image data of the cap beam support to be detected; An image preprocessing module, connected to the image acquisition module, for preprocessing the acquired cap beam support image data, including image denoising, image enhancement and image grayscale conversion; A cap beam support deformation recognition model construction module is used to construct a cap beam support deformation recognition model using a YOLOX neural network framework. The model is used to identify the deformation of the cap beam support in an image, including a deformation feature extraction module, and optimizes the loss function in the original YOLOX neural network for the deformation characteristics of the cap beam support; The deformation recognition module is connected to the cap beam support deformation recognition model building module, and is used to input the preprocessed cap beam support image into the trained cap beam support deformation recognition model to identify the deformation of the cap beam support and output the deformation information of the cap beam support, wherein the deformation information includes the deformation position, deformation type and deformation degree.

[0016] Compared with the prior art, the present invention has the following beneficial effects: Strong anti-interference ability: The present invention uses the YOLOX neural network framework to construct a deformation recognition model for the cap beam support, and specially designs a deformation feature extraction module, as well as a loss function optimization for the deformation characteristics of the cap beam support. These innovations enable the model to more accurately identify deformation features in images, effectively distinguish different types of deformation, and accurately quantify the degree of deformation, greatly improving the accuracy and reliability of deformation detection.

[0017] High level of intelligence and automation: The entire detection process is highly automated and intelligent. From image acquisition, preprocessing to deformation recognition and information output, no human intervention is required. This not only reduces the workload of inspection personnel, but also avoids detection errors caused by human factors, thereby improving detection efficiency and quality. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 A step diagram of the cap beam support deformation detection method of the invention; Figure 2 It is the design schematic diagram of the deformation feature extraction module; Figure 3 Schematic diagram of the loss function design. DETAILED DESCRIPTION

[0019] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0020] See also Figure 1-3 The present invention provides a technical solution: a method for detecting deformation of a cap beam support, the method comprising the following steps: S1. Use image acquisition equipment to obtain image data of the cap beam support to be tested First, select an appropriate image acquisition device and start the image acquisition device to obtain image data of the cap beam support to be inspected.

[0021] Image acquisition can be performed using a high-definition camera mounted on a drone. The drone flies to a suitable position above the cap beam support to ensure that the camera can clearly capture the full view of the support. The control system on the drone adjusts the angle and focal length of the camera to obtain high-quality image data.

[0022] S2. Preprocessing the collected cap beam support image data The collected cap beam support image data is imported into the image processing system for preprocessing operations. The preprocessing steps include: image denoising, image enhancement and image grayscale conversion.

[0023] S3. Use YOLOX neural network framework to build a cap beam support deformation recognition model Based on the YOLOX neural network framework, a deformation recognition model for the cap beam support is constructed. The model includes a deformation feature extraction module, which is used to extract the deformation features of the cap beam support from the preprocessed image. At the same time, the loss function in the original YOLOX neural network is optimized for the deformation characteristics of the cap beam support to improve the recognition accuracy and robustness of the model. The optimized loss function can better adapt to the diversity and complexity of the deformation of the cap beam support, ensuring that the model can accurately identify different types of deformation and their degree.

[0024] During the model training process, a large amount of deformation image data of the cap beam support is used as training samples, and the model parameters are continuously adjusted through an iterative optimization algorithm until the model achieves the expected recognition effect.

[0025] S4. Input the preprocessed cap beam support image into the trained cap beam support deformation recognition model to identify the cap beam support deformation. The preprocessed cap beam support image is input into the trained deformation recognition model, and the model will automatically recognize the deformation of the image. Through the deformation feature extraction module, the model can accurately capture the deformation features in the image and judge the deformation of the cap beam support based on these features. Finally, the deformation information of the cap beam support is obtained, including the deformation location, deformation type and deformation degree.

[0026] The present invention will be further described below in conjunction with Examples 1 to 3: Embodiment 1:

[0027] See attached Figure 2 , the following will describe in detail the implementation steps of the deformation feature extraction module in the present invention with reference to specific examples.

[0028] The implementation steps of this module are as follows: Step 1: Preliminary feature processing The input cap beam support image features are preliminarily processed using the Convolutional Block (CBS) module. The CBS module consists of multiple convolutional layers, activation functions, and batch normalization layers, which can effectively extract the preliminary features of the image. In this example, the CBS module contains two 3x3 convolutional layers, each of which is followed by a ReLU activation function and a batch normalization layer. Through the processing of the CBS module, an image feature map with rich preliminary features can be obtained. Through the feature splitting (Split) operation, some features in the preliminary feature map (for example, the features of half the channels) are used as direct paths (sub-gradient branches) for subsequent splicing. The purpose of this step is to retain some of the original features in order to provide direct feature information in subsequent feature fusion.

[0029] Step 2: Capturing complex deformation features The remaining features (i.e., the parts that are not split into direct paths) are input into the main path composed of multiple deformable convolutional layers. A deformable convolutional layer is a special convolutional layer that can adaptively adjust the shape of the convolution kernel according to the local structure of the input feature map, thereby more effectively capturing complex deformation features. In this example, the main path contains three deformable convolutional layers, each of which is followed by a ReLU activation function and a batch normalization layer. Through the processing of the main path, an image feature map containing rich deformation features can be obtained. These feature maps can reflect the deformation of the cap beam support in different positions and directions, providing key information for subsequent deformation recognition.

[0030] Step 3: Feature Fusion In the main path, the outputs of each deformable convolutional layer are feature fused. The purpose of feature fusion is to combine the deformation features of different levels to enhance the model's sensitivity to deformation features. In this example, the outputs of each deformable convolutional layer are fused in a parallel splicing manner. Specifically, the output feature maps of each layer are spliced ​​in the channel dimension to obtain a comprehensive feature map containing deformation features of all levels. The features of the direct path (sub-gradient branch) and the main path (main gradient branch) are spliced. The purpose of this step is to combine the original features with the deformation features so that both features can be used simultaneously in subsequent classification and localization tasks. The spliced ​​feature map is adjusted through another CBS module to match the input requirements of the subsequent network. During the adjustment process, the dimension and size of the feature map can be changed by adjusting the number of channels of the convolutional layer and the size of the convolution kernel.

[0031] Step 4: Feature Output After the above steps, the deformation feature extraction module finally outputs a feature map containing rich deformation features. This feature map will be used as the input of the subsequent deformation recognition model to identify the deformation of the cap beam support. By training and optimizing the deformation recognition model, the deformation of the cap beam support can be accurately identified and located, providing strong support for the safety monitoring and maintenance of the bridge.

[0032] The deformation feature extraction module in the present invention can effectively extract deformation-related features from the cap beam support image by combining the CBS module, the deformable convolution layer and the feature fusion technology.

[0033] Embodiment 2:

[0034] See attached Figure 3 This embodiment is used to describe the optimization method of the cap beam support deformation recognition model for the loss function in the original YOLOX neural network for the cap beam support deformation characteristics, specifically including: For positioning loss, based on the traditional IoU loss, this paper introduces an improved central point IoU (Centralized Intersection over Union) loss to more accurately reflect the difference between the predicted box and the actual deformed area. This loss function not only considers the IoU value, but also introduces two additional factors: central point distance and aspect ratio consistency.

[0035] In specific implementation, first calculate the prediction box The real deformation area The IoU value between them is the ratio of the intersection area to the union area, denoted as .

[0036] Calculate the center point of the prediction box The center point of the actual deformation area The Euclidean distance between , the formula is as follows: At the same time, calculate the diagonal length of the minimum bounding rectangle covering the predicted box and the true deformation area , which can be found by calculating the width and height of the circumscribed rectangle and using the Pythagorean theorem.

[0037] Then, according to the aspect ratio difference between the predicted box and the actual deformed area, the weight function α and the aspect ratio consistency index are calculated. . It can be a function related to the IoU value to adjust the effect of aspect ratio consistency on the loss; It can be calculated by comparing the aspect ratio of the predicted box and the true deformed area, for example, using the absolute value of the difference between the two aspect ratios. Substituting these values ​​into the improved center point IoU loss formula:

[0038] By minimizing the above loss function, the model can learn how to locate the deformation area more accurately.

[0039] To address the problem of imbalance between positive and negative samples in the deformation detection task of cap beam support, the present invention adopts an adjusted focal loss. This loss function improves the generalization ability of the model by adjusting the weights of positive and negative samples and the degree of attention paid to difficult-to-classify samples.

[0040] In specific implementation, first calculate the predicted probability of each sample , that is, the probability that the model predicts that the sample is a positive sample (deformation). For the binary classification problem, the output of the model can be mapped to the interval [0,1] through the sigmoid function. Then, according to the true label y of the sample, determine When the sample is a positive sample (i.e. y=1), Take a smaller value (such as 0.25) to reduce the contribution of positive samples to the loss function; when the sample is a negative sample (ie y=0), Take a larger value (such as 0.75) to increase the contribution of negative samples to the loss function. This can balance the difference in the number of positive and negative samples.

[0041] Setting focus parameters , which is used to adjust the degree of attention paid to samples that are difficult to classify. The larger the value of , the more attention the model pays to difficult-to-classify samples. Substituting these values ​​into the focal loss formula:

[0042] When the sample is a positive sample (i.e. y=1), is the probability that the model predicts a positive sample; when the sample is a negative sample (i.e. y=0), It is 1 minus the probability that the model predicts a positive sample, that is, the probability that the model predicts a negative sample.

[0043] During the training process, by minimizing the above focal loss function, the model can gradually learn how to accurately identify the deformation of the cap beam support and pay more attention to those samples that are difficult to classify.

[0044] The present invention realizes more accurate detection of deformation of the cap beam support through the optimized loss function. In practical applications, the parameters in the loss function can be adjusted according to specific detection requirements and scenarios to obtain the best detection effect.

[0045] Embodiment 3: In order to effectively eliminate the noise in the image and improve the image quality, this embodiment uses Gaussian filtering to process the cap beam support image. Gaussian filtering is a linear smoothing filter. Its core idea is to use the Gaussian function to calculate the weighted average of the pixel values ​​in the neighborhood of each pixel point, thereby replacing the original pixel value to achieve a denoising effect. The following are the specific implementation steps: The parameters of the Gaussian function are mainly the standard deviation σ. The size of σ determines the shape and width of the Gaussian function, which in turn affects the filtering effect. Generally speaking, the larger the σ, the more obvious the filtering effect, but it may also cause the loss of image details; the smaller the σ, the weaker the filtering effect, but it can better preserve image details. According to the characteristics of the cap beam support image and the noise situation, select the appropriate σ value.

[0046] According to the determined Gaussian function parameters, a Gaussian kernel is constructed. The Gaussian kernel is a two-dimensional matrix whose element values ​​are calculated by the Gaussian function. For each pixel point (x, y) in the image, the corresponding Gaussian function value is: Where G(x,y) is the value of the Gaussian function at the two-dimensional coordinate (x,y) of the image, x and y are the coordinates in two-dimensional space, σ is the standard deviation of the Gaussian distribution, e is the base of the natural logarithm, and π is the circumference of a circle.

[0047] Apply the constructed Gaussian kernel to the image for filtering. For each pixel in the image, take a neighborhood with the same size as the Gaussian kernel as the center. Then, multiply the value of each pixel in the neighborhood with the element value of the corresponding position of the Gaussian kernel, and add all the products to get the filtering result of the pixel.

[0048] Traverse each pixel in the image and calculate its pixel value after filtering. Combine the filtering results of all pixels to get the filtered image. At this point, the noise in the image is effectively suppressed and the image quality is improved.

[0049] The steps to implement image enhancement processing include: Application of Sobel detection operator: Apply Sobel detection operator to the original cap beam support image for edge detection. Sobel operator is a discrete differential operator that can calculate the gradient of the grayscale value of an image, thereby highlighting the edges in the image. Specifically, the image is convolved using Sobel filters in the horizontal and vertical directions to obtain the gradient strength G and gradient direction θ of each pixel.

[0050] Non-maximum suppression: Go through each pixel in the image and compare its gradient strength G with the gradient strength of the adjacent pixel in its gradient direction θ. If the gradient strength of the current pixel is not the local maximum (i.e., it is not the largest in its gradient direction), then set its gradient strength to zero. This step aims to eliminate pixels that are not true contours and reduce artifacts in subsequent processing.

[0051] Double threshold detection: Set two thresholds: high threshold T_high and low threshold T_low. Pixels with gradient strength higher than T_high are considered strong contour points; pixels with gradient strength lower than T_low are considered non-contour points and removed. For pixels with gradient strength between T_low and T_high, further check whether they are connected to strong contour points. If connected, they are considered weak contour points and retained; otherwise, they are also considered non-contour points and removed. This step helps retain important edge information in the image while removing unnecessary details.

[0052] The steps to implement image grayscale include: Weighted conversion: Use the weighted method to convert the color image to a grayscale image. For each pixel in the image, use the formula: GRAY=0.3×R+0.59×G+0.11×B Among them, R, G, and B represent the component values ​​of the pixel in the red, green, and blue channels respectively. This formula weights the sensitivity of the human eye to different colors to obtain a grayscale value that is more in line with visual perception.

[0053] Traversal and conversion: Read the original data of the cap beam support image and traverse the image pixel by pixel. For each pixel, extract its RGB color value and apply the above formula to calculate the corresponding grayscale value. Then, assign the calculated grayscale value to the pixel to complete the grayscale conversion of the entire image.

[0054] A deformation report is automatically generated based on the extracted deformation information. The report includes: precise coordinates of the deformation location (based on the image coordinate system), detailed description of the deformation type, quantitative data of the deformation degree (such as deformation amount, deformation ratio, etc.), and repair or replacement suggestions based on the deformation degree and type. The report can be output in electronic or printed form for easy viewing and decision-making by relevant personnel.

[0055] A cap beam support deformation detection system, the system comprising: An image acquisition module, used to obtain image data of the cap beam support to be detected; An image preprocessing module, connected to the image acquisition module, for preprocessing the acquired cap beam support image data, including image denoising, image enhancement and image grayscale conversion; A cap beam support deformation recognition model construction module is used to construct a cap beam support deformation recognition model using a YOLOX neural network framework. The model is used to identify the deformation of the cap beam support in an image, including a deformation feature extraction module, and optimizes the loss function in the original YOLOX neural network for the deformation characteristics of the cap beam support; The deformation recognition module is connected to the cap beam support deformation recognition model building module, and is used to input the preprocessed cap beam support image into the trained cap beam support deformation recognition model to identify the deformation of the cap beam support and output the deformation information of the cap beam support, wherein the deformation information includes the deformation position, deformation type and deformation degree.

[0056] The implementation of the system refers to the above-mentioned embodiment and will not be described in detail in the specification.

[0057] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0058] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for detecting deformation of a cap beam support, characterized in that: The method comprises: S1. Using an image acquisition device to obtain image data of the cap beam support to be inspected; S2, preprocessing the collected cap beam support image data, including image denoising, image enhancement and image grayscale; S3. A cap beam support deformation recognition model is constructed using a YOLOX neural network framework. The model is used to identify the deformation of the cap beam support in the image. The model includes a deformation feature extraction module, and the loss function in the original YOLOX neural network is optimized for the deformation characteristics of the cap beam support. S4. Input the cap beam support image preprocessed in step S2 into the trained cap beam support deformation recognition model to identify the deformation of the cap beam support and obtain the deformation information of the cap beam support, including the deformation position, deformation type and deformation degree.

2. The cap beam support deformation detection method according to claim 1, characterized in that: The implementation steps of the deformation feature extraction module include: The convolutional block module is used to preliminarily process the input cap beam support image features, and some features are used as direct paths for subsequent splicing through feature splitting operations; The remaining features are fed into a main path consisting of multiple deformable convolutional layers to capture the complex deformation characteristics of the cap beam support; In the main path, the outputs of each deformable convolutional layer are feature fused to enhance the model’s sensitivity to deformation features; The features of the direct path and the main path are concatenated, and the number of channels is adjusted through the convolution block module to achieve comprehensive feature fusion.

3. The cap beam support deformation detection method according to claim 1, characterized in that: The cap beam support deformation recognition model optimizes the loss function in the original YOLOX neural network for the deformation characteristics of the cap beam support, specifically including: using the improved center point IoU loss to replace the original IoU loss as the positioning loss. When calculating the overlap between the prediction box and the actual deformation area, the loss additionally considers the center point distance and aspect ratio factors. The algorithm is: in, is the Euclidean distance between the predicted box and the true deformed area, is the diagonal length of the minimum bounding rectangle covering the predicted box and the true deformation area, is a weight function used to adjust the effect of aspect ratio consistency on the loss. It is an indicator to measure the consistency of aspect ratio.

4. The cap beam support deformation detection method according to claim 3, characterized in that: The adjusted focal loss is used to replace the original binary cross entropy loss as the classification loss. The calculation formula of the focal loss is: in, The probability of predicting a positive sample, is the factor used to balance the weights of positive and negative samples, It is an adjustable focusing parameter used to adjust the degree of attention paid to samples that are difficult to classify.

5. The cap beam support deformation detection method according to claim 1, characterized in that: Gaussian filtering is used to eliminate noise in the image. The value of each pixel in the image is replaced by the weighted average of the pixel values ​​in its neighborhood. The calculation formula is: Where G(x,y) is the value of the Gaussian function at the two-dimensional coordinate (x,y) of the image, x and y are the coordinates in two-dimensional space, σ is the standard deviation of the Gaussian distribution, e is the base of the natural logarithm, and π is the circumference of a circle.

6. The cap beam support deformation detection method according to claim 1, characterized in that: The steps for implementing image enhancement on the cap beam support image include: Step I: Use the Sobel detection operator to perform convolution on the image and calculate the gradient intensity and direction of each pixel; Step II: traverse each pixel in the image, compare its gradient strength with the gradient strength of the adjacent pixels in its gradient direction, and if the current pixel is not a local maximum, set its gradient strength to zero, thereby eliminating non-contour pixels and reducing identification artifacts; Step III: Set two thresholds: a high threshold and a low threshold. Pixels with gradient strengths higher than the high threshold are considered strong contour points, and pixels with gradient strengths lower than the low threshold are considered non-contour points. For pixels with gradient strengths between the two thresholds, if they are connected to strong contour points, they are considered weak contour points and retained; otherwise, they are considered non-contour points and removed.

7. The cap beam support deformation detection method according to claim 1, characterized in that: The method for grayscale processing of the cap beam support image is: a. Select the weighted method to convert the color image into a grayscale image, and use the formula GRAY = 0.3*R + 0.59*G +0.11*B to calculate the grayscale value. R, G, and B in the formula represent the components of the pixel in the red, green, and blue color domains respectively; b. Read the cap beam support image data, traverse each pixel in the image, and obtain the RGB color value corresponding to each pixel; c. Apply the color value conversion method, call the formula in step a to convert the RGB color value of each pixel obtained in step b to obtain the corresponding grayscale value; d. Apply the converted grayscale value to each pixel to generate a grayscale image of the cap beam support image.

8. The cap beam support deformation detection method according to claim 1, characterized in that: Based on the deformation information of the cap beam support obtained in step S4, a deformation report is generated, which includes the precise coordinates of the deformation position, a detailed description of the deformation type, quantitative data on the degree of deformation, and recommended repair or replacement measures.

9. The cap beam support deformation detection method according to claim 1, characterized in that: The image acquisition device is a high-definition camera carried by a drone, which is used to capture images by flying the drone above the cap beam support.

10. A cap beam support deformation detection system, characterized in that: The system comprises: An image acquisition module, used to obtain image data of the cap beam support to be detected; An image preprocessing module, connected to the image acquisition module, for preprocessing the acquired cap beam support image data, including image denoising, image enhancement and image grayscale conversion; A cap beam support deformation recognition model construction module is used to construct a cap beam support deformation recognition model using a YOLOX neural network framework. The model is used to identify the deformation of the cap beam support in an image, including a deformation feature extraction module, and optimizes the loss function in the original YOLOX neural network for the deformation characteristics of the cap beam support; The deformation recognition module is connected to the cap beam support deformation recognition model building module, and is used to input the preprocessed cap beam support image into the trained cap beam support deformation recognition model to identify the deformation of the cap beam support and output the deformation information of the cap beam support, wherein the deformation information includes the deformation position, deformation type and deformation degree.

Images