Intelligent deflection monitoring and full-field mechanical response evaluation method for large-span truss steel structure

Abstract

The invention discloses a method for intelligent deflection monitoring and full-field mechanical response evaluation of a large-span truss steel structure. An intelligent deflection monitoring deep learning frame of the large-span truss steel structure is provided, and automatic identification and target matching of nodes of the large-span truss steel structure can be achieved. In addition, a proxy model based on a graph neural network GNN is developed, and monitoring displacement data of field finite nodes can be quickly converted into full-field displacement and strain information of the structure. By comparing the full-field displacement and strain data generated by the GNN agent model with a preset standard or a user-defined early warning limit value, real-time guidance can be provided for construction, and danger early warning can be triggered when necessary. Compared with a traditional contact type monitoring means, intelligent identification and non-contact type deflection monitoring of a plurality of key nodes of the large-span truss steel structure can be achieved, and meanwhile convenience, data accuracy and reliability of mechanical response monitoring are remarkably improved.

Description

Technical Field

[0001] The present invention relates to the technical field of deflection monitoring and full-field mechanical response evaluation during the construction process or operation stage of a large-span truss steel structure, and specifically to an intelligent deflection monitoring and full-field mechanical response evaluation method for a large-span truss steel structure. Background Art

[0002] During the entire construction and operation cycle of large-span truss steel structures, accurate monitoring of their deformation state plays a core role. It is not only related to the scientific nature of the construction strategy and the solid foundation of the project quality, but also directly affects the efficiency of the immediate assessment of structural safety and damage identification. However, current measurement technology has encountered multiple challenges in practical applications. Although traditional contact measurement methods are known for their high precision, they are limited by the limitations of the measurement range, the complexity of installation and maintenance, and the lack of adaptability to the changing engineering environment. In contrast, non-contact technologies such as GPS and laser Doppler vibrometers, although they have achieved the convenience of remote monitoring, the lack of vertical accuracy of GPS and the strict requirements of laser Doppler vibrometers on vibration isolation conditions have limited their effectiveness in practical applications.

[0003] In recent years, digital image correlation (DIC) has gradually emerged as a rising star in the field of visual measurement. This technology relies on digital cameras to capture images of structures before and after deformation, and uses sophisticated algorithms to analyze the correlation changes in image sub-regions to achieve high-precision tracking of deformation trajectories. However, when extending it to the field of outdoor large-span truss steel structure monitoring, the following key challenges need to be overcome:

[0004] (1) The challenge of balancing performance and resolution: In complex and ever-changing construction environments, industrial cameras need to balance the relationship between image resolution, frame rate, and storage capacity. Although high resolution improves the ability to capture details, it is also accompanied by problems such as decreased frame rate and increased memory consumption, which in turn restricts the smoothness and response speed of real-time monitoring.

[0005] (2) Insufficient robustness of feature matching: Traditional matching methods based on local feature descriptors (such as SIFT and SURF) are highly sensitive to environmental factors such as illumination changes, rotation, and scaling, and are easily disturbed by noise and occlusion, resulting in frequent mismatches or false matches. In addition, the traditional DIC algorithm's reliance on speckles as displacement markers limits its application flexibility in the absence of a preset target template, increasing the burden of preliminary preparation work.

[0006] (3) The current structural health monitoring system mostly relies on contact sensor data, while the dominant position of optical measurement data in the evaluation framework has not yet been established. Most existing evaluation frameworks are based on preset specification limits, which makes it difficult to fully utilize the rich finite element analysis results for real-time and comprehensive performance evaluation, limiting the intelligence and precision of the monitoring system. Summary of the invention

[0007] In response to the above technical problems, the present invention proposes a method for intelligent deflection monitoring and full-field mechanical response assessment of large-span truss steel structures. The method utilizes an integrated non-contact deflection monitoring and full-field mechanical response assessment system, which can effectively implement non-contact full-field deflection and strain monitoring and assessment of large-span truss steel structures, and has significantly improved the convenience, data accuracy and reliability of mechanical response monitoring, providing strong technical support for safety monitoring and risk assessment of large-span truss steel structures.

[0008] In order to achieve the above technical objectives, the present invention adopts the following technical solutions:

[0009] A method for intelligent deflection monitoring and full-field mechanical response evaluation of a large-span truss steel structure comprises the following steps:

[0010] S1: Obtain the motion trajectory of the key nodes in the mid-span of the long-span truss steel structure;

[0011] S2: Construct a deep learning framework for intelligent deflection monitoring of large-span truss steel structures;

[0012] S21 uses the image of the key nodes in the mid-span of the long-span truss steel structure obtained in step S1 to obtain the pixel coordinates of these key nodes, and calculates the physical distance corresponding to each pixel of each key node in the image, that is, the proportional factor corresponding to each key node in the image;

[0013] S22 selects an upsampling ratio of the diffusion model super-resolution reconstruction algorithm based on the scale factor of each key node obtained in step S21, obtains the scale factor of the reconstructed image, and ensures that the scale factor of the reconstructed image is maintained at an accuracy level of 0.1 to 10 mm / pixel;

[0014] S23 uses the trained YOLO v8 detection algorithm to track the position of each mid-span key node of the long-span truss steel structure in the image sequence taken after loading, passes the target frame coordinate position information output by the YOLO v8 detection algorithm to the LoFTR algorithm, and uses the LoFTR algorithm to calculate the image pixel displacement corresponding to all mid-span key nodes on the long-span truss steel structure, and accurately converts the image displacement of each mid-span key node into a physical displacement based on the scale factor of the reconstructed image obtained in step S22;

[0015] S3 builds a structural evaluation framework based on the graph neural network GNN agent model;

[0016] S31 compares the physical displacement of each key node in the span with the trained GNN proxy model, and uses the GNN proxy model to infer the full-field displacement and strain data of the large-span truss steel structure by matching the corresponding mechanical response state;

[0017] S32 compares the full-field displacement and strain data of the large-span truss steel structure displayed by the GNN agent model with the warning limits specified in the code or defined by the user to provide construction guidance or trigger danger warnings.

[0018] In step S22, the diffusion model super-resolution reconstruction algorithm is first trained on a large-span truss steel structure dataset containing high / low resolution black-and-white image pairs to ensure that the diffusion model super-resolution reconstruction algorithm accurately maps low to high resolution images;

[0019] The super-resolution images generated by the test set are evaluated by the peak signal-to-noise ratio (PSNR) and the structural similarity index (SSIM), and are adjusted to maximize the peak signal-to-noise ratio (PSNR) and the structural similarity index (SSIM) to achieve the best performance.

[0020] In step S23, the YOLO v8 detection algorithm first constructs a black-and-white image dataset containing steel truss node annotations for model training to achieve accurate detection and positioning of steel truss nodes, and improves detection accuracy and overall performance by optimizing the YOLO v8 detection algorithm parameters and hyperparameters.

[0021] In step S23, the LoFTR algorithm relies on the training of the large-span truss steel structure image set containing displacement information, and the YOLO v8 detection algorithm outputs the target frame coordinates to the LoFTR algorithm, which only performs local matching on the pixels within the frame.

[0022] The training of the GNN proxy model in step S3 first collects data by calculating 1,000 finite element models of actual steel trusses. These data are converted into graphical expressions, in which each node of the graph is mapped to the key structural elements of steel nodes, crossbars and vertical bars, and the edges of the graph outline the connection relationship between them; considering that the displacement of steel structures is significantly affected by temperature changes, the surface temperature difference of outdoor steel structures is incorporated into the GNN proxy model as an important feature; the architecture of the GNN proxy model is designed, including graph convolution, attention mechanism and pooling layer, and the prediction task of mechanical response is converted into a node classification or regression problem in the graph domain. The model is optimized and verified by cross-validation and hyperparameter adjustment techniques.

[0023] Step S1 specifically includes:

[0024] During the construction preparation stage, an ultrasonic rangefinder was used to measure the horizontal and vertical distances from the construction site leveling site to the key nodes in the middle of the long-span truss steel structure, and the field of view of the black-and-white industrial camera was estimated based on the pinhole imaging principle.

[0025] The black-and-white industrial camera is used to completely record the motion trajectory of the key nodes in the mid-span of the large-span truss steel structure.

[0026] Step S21 specifically includes:

[0027] The ultrasonic rangefinder is used to accurately measure the distance from the key mid-span nodes of the large-span truss steel structure to the center of the target surface of the black-and-white industrial camera. Subsequently, the pixel coordinates of these key mid-span nodes in the camera view are identified by a computer, and their corresponding scale factors are calculated based on geometric relationships.

[0028] Step S22 specifically includes:

[0029] A black-and-white industrial camera is used to capture a series of images containing key nodes in the mid-span of the large-span truss steel structure during the construction process or operation stage, and an image before loading is selected as the baseline reference image. The research focus is focused on the large-span truss steel structure, and all the collected images are cropped to retain only the main part of the large-span truss steel structure beam. Based on the scale factor calculated in step S3, the upsampling ratio of the diffusion model super-resolution reconstruction algorithm is selected to obtain the scale factor of the reconstructed image, so as to ensure that the scale factor of the reconstructed image is maintained at an accuracy level of 0.1 to 10 mm / pixel.

[0030] The cropping operation is specifically for all images captured by black and white industrial cameras, focusing on the square pixel area occupied by the large-span truss steel structure beams in the image, that is, the area is defined by four coordinate points of the top, bottom, left and right. The union of these square pixel areas in all pictures is accurately cropped and replaced with the original image.

[0031] Beneficial effects:

[0032] First, the present invention provides a method for intelligent deflection monitoring and full-field mechanical response evaluation of large-span truss steel structures, which integrates machine vision and graph neural network algorithms. In terms of recognition technology, a deep learning framework is proposed, which breaks through the limitations of low pixels and the need to install target scattered spots of traditional black-and-white industrial cameras, significantly expands the deployment range of the monitoring system, and realizes automatic recognition and target matching of large-span truss steel structure nodes.

[0033] Second, the present invention further develops a proxy model based on graph neural network, which can quickly infer the full-field displacement and strain information of the structure based on the limited monitoring displacement data of the key nodes in the span on site. At the same time, by comparing the full-field displacement and strain data generated by the proxy model of the graph neural network with the preset specifications or user-defined warning limits, it can provide real-time guidance for construction and trigger danger warnings when necessary.

[0034] Third, compared with traditional contact monitoring methods, the present invention can effectively implement intelligent identification and non-contact deflection and strain monitoring of multiple key nodes of large-span truss steel structures, and can safely, quickly and accurately monitor the changes in mechanical response of large-span truss steel structures during the construction process or operation stage, providing a solid data foundation for structural safety assessment. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 It is a structural schematic diagram of the intelligent deflection monitoring and full-field mechanical response evaluation system for large-span truss steel structures of the present invention;

[0036] Figure 2 This is a flow chart of the intelligent deflection monitoring and full-field mechanical response evaluation method of the large-span truss steel structure of the present invention;

[0037] Figure 3 This is a schematic diagram of the diffusion model training process of the present invention;

[0038] Figure 4 It is a schematic diagram of the local target matching process of the present invention;

[0039] Figure 5 Flowchart of designing the GNN proxy model for the present invention. DETAILED DESCRIPTION

[0040] The present invention will be further explained below in conjunction with the accompanying drawings and specific embodiments. It should be understood that the following specific embodiments are only used to illustrate the present invention and are not used to limit the scope of the present invention.

[0041] To facilitate understanding of the present invention, the present invention will be described more comprehensively below with reference to the relevant drawings. Several embodiments of the present invention are given in the drawings, but the present invention can be implemented in different forms and is not limited to the embodiments described in the text. On the contrary, these embodiments are provided to make the content disclosed in the present invention more thorough and comprehensive.

[0042] For examples, please refer to the attached Figure 1A large-span truss steel structure intelligent deflection monitoring and full-field mechanical response evaluation system is mainly composed of a heavy tripod 1, a black-and-white industrial camera 2, a fixed-focus lens 3, a computer 4, and an ultrasonic rangefinder 5. The black-and-white industrial camera 2 and the fixed-focus lens 3 are fixed on the heavy tripod 1 to shoot the dynamic motion trajectory of the key nodes in the mid-span of the large-span truss steel structure. The computer 4 is connected to the black-and-white industrial camera 2 by data, and the computer 4 controls the camera to shoot the deflection change process at the key nodes in the mid-span of the large-span truss steel structure during the construction process or operation stage.

[0043] The black-and-white industrial camera 2 and the fixed-focus lens 3 are fixed to the ground via a heavy-duty tripod 1 equipped with a pan / tilt head.

[0044] The heavy-duty tripod 1 is a Manfrotto 475B tripod and a 400 gear head.

[0045] The black and white industrial camera 2 is a FLIR GS3-U3-41C6M-C black and white industrial camera.

[0046] The fixed-focus lens 3 is a KOWALM16HC lens.

[0047] The computer 4 is configured with CPU AMD Ryzen 55600X, 32G RAM, 512G memory, and graphics card NVIDIA GeForce RTX 3060.

[0048] The ultrasonic rangefinder 5 is a Xima AR841 handheld ultrasonic rangefinder.

[0049] A method for intelligent deflection monitoring and full-field mechanical response evaluation of a large-span truss steel structure, the method comprising the following steps:

[0050] S1: Obtain the motion trajectory of the key nodes in the mid-span of the long-span truss steel structure;

[0051] During the construction preparation stage, an ultrasonic rangefinder was used to measure the horizontal and vertical distances from the construction site leveling site to the key nodes in the middle of the long-span truss steel structure with an accuracy of ±(1cm+0.5%×distance). Based on the pinhole imaging principle, the field of view of the black-and-white industrial camera was estimated.

[0052] The black-and-white industrial camera is used to completely record the motion trajectory of the key nodes in the mid-span of the large-span truss steel structure;

[0053] S2: Construct a deep learning framework for intelligent deflection monitoring of large-span truss steel structures;

[0054] S21 uses the image of the key nodes in the mid-span of the long-span truss steel structure obtained in step S1 to obtain the pixel coordinates of these key nodes, and calculates the physical distance corresponding to each pixel of each key node in the image, that is, the proportional factor corresponding to each key node in the image, which is specifically:

[0055] The ultrasonic rangefinder is used to accurately measure the distance from the key mid-span nodes of the large-span truss steel structure to the center of the target surface of the black-and-white industrial camera. Subsequently, the pixel coordinates of these key mid-span nodes in the camera view are identified by a computer, and the corresponding scale factors are calculated based on the geometric relationship.

[0056] S22 selects the upsampling ratio of the diffusion model super-resolution reconstruction algorithm based on the scale factor of each mid-span key node obtained in step S21, and obtains the scale factor of the reconstructed image to ensure that the scale factor of the reconstructed image is maintained at an accuracy level of 0.1 to 10 mm / pixel, specifically:

[0057] A black-and-white industrial camera is used to capture a series of images including key mid-span nodes of the long-span truss steel structure during the construction process or operation stage, and an image before loading is selected as a baseline reference image, focusing the research on the long-span truss steel structure, cropping all the collected images, and only retaining the main part of the long-span truss steel structure beam. Based on the scale factor calculated in step S3, the upsampling ratio of the diffusion model super-resolution reconstruction algorithm is selected to obtain the scale factor of the reconstructed image, so as to ensure that the scale factor of the reconstructed image is maintained at an accuracy level of 0.1 to 10 mm / pixel;

[0058] S23 uses the trained YOLO v8 detection algorithm to track the position of each mid-span key node of the long-span truss steel structure in the image sequence taken after loading, passes the target frame coordinate position information output by the YOLO v8 detection algorithm to the LoFTR algorithm, and uses the LoFTR algorithm to calculate the image pixel displacement corresponding to all mid-span key nodes on the long-span truss steel structure, and accurately converts the image displacement of each mid-span key node into a physical displacement based on the scale factor of the reconstructed image obtained in step S22;

[0059] S3: Construct a structural evaluation framework based on the graph neural network (GNN) agent model;

[0060] S31 compares the physical displacement of each key node in the span with the trained GNN proxy model, and uses the GNN proxy model to infer the full-field displacement and strain data of the large-span truss steel structure by matching the corresponding mechanical response state;

[0061] S32 compares the full-field displacement and strain data of the large-span truss steel structure displayed by the GNN agent model with the warning limits specified in the code or defined by the user to provide construction guidance or trigger danger warnings.

[0062] As a further preference of the present invention, in step S22, the image cropping operation is specifically for all images captured by the black and white industrial camera, focusing on the square pixel area occupied by the large-span truss steel structure beam in the image, that is, the area is defined by four coordinate points of the top, bottom, left and right, and the union of these square pixel areas in all pictures is accurately cropped and replaced with the original image.

[0063] As a further preference of the present invention, the implementation of the diffusion model super-resolution reconstruction algorithm primarily relies on a large-span truss steel structure dataset containing high-resolution and low-resolution black-and-white image pairs. This dataset is used to train the diffusion model to ensure that it can accurately grasp the complex mapping relationship from low-resolution images to their high-resolution counterparts. In order to comprehensively evaluate the performance of the model in super-resolution reconstruction, the diffusion model super-resolution reconstruction algorithm is fine-tuned and optimized by calculating the peak signal-to-noise ratio PSNR and structural similarity index SSIM of the super-resolution images generated by the test dataset. The larger the peak signal-to-noise ratio and the structural similarity index, the better, to ensure that its performance reaches the optimal level.

[0064] As a further optimization of the present invention, the YOLO v8 detection algorithm needs to collect a black-and-white image dataset containing steel truss node annotation information in advance. Based on the annotated dataset, the YOLO v8 detection algorithm is trained to achieve accurate detection and positioning of steel truss nodes in the image, and the detection accuracy and overall performance are further optimized by fine-tuning parameters and hyperparameters.

[0065] As a further optimization of the present invention, the operation of the LoFTR algorithm relies on a pre-collected black-and-white image dataset of a large-span truss steel structure, which contains known displacement information; the LoFTR algorithm is trained using an image sequence containing displacement changes. In actual applications, the target frame coordinate position information output by the YOLO v8 detection algorithm is passed to the LoFTR algorithm, and the LoFTR algorithm only performs local matching calculations on the pixels within the range of these target frames.

[0066] As a further preferred embodiment of the present invention, the training of the GNN proxy model first collects data by calculating 1,000 finite element models of actual steel trusses, and these data are converted into graphical expressions, in which each node of the graph is mapped to key structural elements such as steel nodes, crossbars and vertical bars, and the edges of the graph outline the connection relationship between them. Considering that the displacement of steel structures is significantly affected by temperature changes, the surface temperature difference of outdoor steel structures is incorporated into the model as an important feature; the architecture of the GNN proxy model is designed, including graph convolution, attention mechanism and pooling layer, and the prediction task of mechanical response is converted into a node classification or regression problem in the graph domain. The GNN proxy model is optimized and verified by cross-validation and hyperparameter adjustment techniques.

[0067] The above is only an embodiment of the present invention. Although the present invention describes several embodiments, it is not intended to be limited to these embodiments. For those skilled in the art, the present invention can be improved without departing from the principles of the present invention, and all improvements are included in the scope of the claims of the present invention.

Claims

1. A method for intelligent deflection monitoring and full-field mechanical response evaluation of a large-span truss steel structure, characterized in that: The following steps are involved: S1: Obtain the motion trajectory of the key nodes in the mid-span of the long-span truss steel structure; S2: Construct a deep learning framework for intelligent deflection monitoring of large-span truss steel structures; S21 uses the image of the key nodes in the mid-span of the long-span truss steel structure obtained in step S1 to obtain the pixel coordinates of these key nodes, and calculates the physical distance corresponding to each pixel of each key node in the image, that is, the proportional factor corresponding to each key node in the image; S22 selects an upsampling ratio of the diffusion model super-resolution reconstruction algorithm based on the scale factor of each key node obtained in step S21, obtains the scale factor of the reconstructed image, and ensures that the scale factor of the reconstructed image is maintained at an accuracy level of 0.1 to 10 mm / pixel; S23 uses the trained YOLO v8 detection algorithm to track the position of each mid-span key node of the long-span truss steel structure in the image sequence taken after loading, passes the target frame coordinate position information output by the YOLO v8 detection algorithm to the LoFTR algorithm, and uses the LoFTR algorithm to calculate the image pixel displacement corresponding to all mid-span key nodes on the long-span truss steel structure, and accurately converts the image displacement of each mid-span key node into a physical displacement based on the scale factor of the reconstructed image obtained in step S22; S3 builds a structural evaluation framework based on the graph neural network GNN agent model; S31 compares the physical displacement of each key node in the span with the trained GNN proxy model, and uses the GNN proxy model to infer the full-field displacement and strain data of the large-span truss steel structure by matching the corresponding mechanical response state; S32 compares the full-field displacement and strain data of the large-span truss steel structure displayed by the GNN agent model with the warning limits specified in the code or defined by the user to provide construction guidance or trigger danger warnings.

2. According to claim 1, a method for intelligent deflection monitoring and full-field mechanical response evaluation of a large-span truss steel structure is characterized in that: In step S22, the diffusion model super-resolution reconstruction algorithm is first trained on a large-span truss steel structure dataset containing high / low resolution black-and-white image pairs to ensure that the diffusion model super-resolution reconstruction algorithm accurately maps low to high resolution images; The super-resolution images generated by the test set are evaluated by the peak signal-to-noise ratio (PSNR) and the structural similarity index (SSIM), and are adjusted to maximize the peak signal-to-noise ratio (PSNR) and the structural similarity index (SSIM) to achieve the best performance.

3. According to claim 1, a method for intelligent deflection monitoring and full-field mechanical response evaluation of a large-span truss steel structure is characterized in that: In step S23, the YOLO v8 detection algorithm first constructs a black-and-white image dataset containing steel truss node annotations for model training to achieve accurate detection and positioning of steel truss nodes, and improves detection accuracy and overall performance by optimizing the YOLO v8 detection algorithm parameters and hyperparameters.

4. According to claim 1, a method for intelligent deflection monitoring and full-field mechanical response evaluation of a large-span truss steel structure is characterized in that: In step S23, the LoFTR algorithm relies on the training of the large-span truss steel structure image set containing displacement information, and the YOLO v8 detection algorithm outputs the target frame coordinates to the LoFTR algorithm, which only performs local matching on the pixels within the frame.

5. According to claim 1, a method for intelligent deflection monitoring and full-field mechanical response evaluation of a large-span truss steel structure is characterized in that: The training of the GNN proxy model in step S3 first collects data by calculating 1,000 finite element models of actual steel trusses. These data are converted into graphical expressions, in which each node of the graph is mapped to the key structural elements of steel nodes, crossbars and vertical bars, and the edges of the graph outline the connection relationship between them; considering that the displacement of steel structures is significantly affected by temperature changes, the surface temperature difference of outdoor steel structures is incorporated into the GNN proxy model as an important feature; the architecture of the GNN proxy model is designed, including graph convolution, attention mechanism and pooling layer, and the prediction task of mechanical response is converted into a node classification or regression problem in the graph domain. The model is optimized and verified by cross-validation and hyperparameter adjustment techniques.

6. According to claim 1, a method for intelligent deflection monitoring and full-field mechanical response evaluation of a large-span truss steel structure is characterized in that: Step S1 specifically includes: During the construction preparation stage, an ultrasonic rangefinder was used to measure the horizontal and vertical distances from the construction site leveling site to the key nodes in the middle of the long-span truss steel structure, and the field of view of the black-and-white industrial camera was estimated based on the pinhole imaging principle. The black-and-white industrial camera is used to completely record the motion trajectory of the key nodes in the mid-span of the large-span truss steel structure.

7. According to claim 6, a method for intelligent deflection monitoring and full-field mechanical response evaluation of a large-span truss steel structure is characterized in that: Step S21 specifically includes: The ultrasonic rangefinder is used to accurately measure the distance from the key mid-span nodes of the large-span truss steel structure to the center of the target surface of the black-and-white industrial camera. Subsequently, the pixel coordinates of these key mid-span nodes in the camera view are identified by a computer, and their corresponding scale factors are calculated based on geometric relationships.

8. The method for intelligent deflection monitoring and full-field mechanical response evaluation of a large-span truss steel structure according to claim 6 is characterized in that: Step S22 specifically includes: A black-and-white industrial camera is used to capture a series of images containing key nodes in the mid-span of the large-span truss steel structure during the construction process or operation stage, and an image before loading is selected as the baseline reference image. The research focus is focused on the large-span truss steel structure, and all the collected images are cropped to retain only the main part of the large-span truss steel structure beam. Based on the scale factor calculated in step S3, the upsampling ratio of the diffusion model super-resolution reconstruction algorithm is selected to obtain the scale factor of the reconstructed image, so as to ensure that the scale factor of the reconstructed image is maintained at an accuracy level of 0.1 to 10 mm / pixel.

9. The method for intelligent deflection monitoring and full-field mechanical response evaluation of a large-span truss steel structure according to claim 8 is characterized in that: The cropping operation is specifically for all images captured by black and white industrial cameras, focusing on the square pixel area occupied by the large-span truss steel structure beam in the image, that is, the area is defined by four coordinate points of up, down, left and right. The union of these square pixel areas in all pictures is accurately cropped and replaced with the original image.

Images