Structure multi-target deformation monitoring method based on visual segmentation and deblurring enhancement

Abstract

This invention discloses a method for monitoring multi-target deformation of structures based on visual segmentation and deblurring enhancement, relating to the field of engineering structural health monitoring technology. The method includes: acquiring continuous temporal multi-target images of different depth planes of the structure to be monitored; generating a multi-target mask in the first frame by inputting interactive prompts into a pre-trained segmentation model, and removing noise regions based on connected component analysis; cropping using an adaptive multi-scale cropping strategy based on boundary determination to obtain sub-images; enhancing the sub-images using a pre-trained deblurring neural network; extracting the center point and circumscribed rectangle coordinates of the target mask regions within the sub-images and mapping them back to the full image coordinate system; dynamically updating the cropping window of subsequent frames based on the center point and boundary information of the first frame; and converting pixel-level displacement into physical units or three-dimensional coordinate changes by comparing the center point positions of the first and subsequent frames and combining camera calibration parameters, thereby achieving high-precision continuous monitoring.

Description

Technical Field

[0001] This invention relates to the field of engineering structure health monitoring technology, and in particular to a method and device for monitoring multi-target deformation of structures based on visual segmentation and deblurring enhancement. Background Technology

[0002] With the continuous advancement of structural health monitoring technology, machine vision, as a non-contact, information-rich, and easily deployed method, has been widely applied in structural deformation monitoring across various fields such as civil engineering and machinery manufacturing. Compared to traditional contact sensors, visual monitoring systems can not only acquire information on a wide range of structural deformations but also significantly reduce deployment costs and improve the flexibility and reliability of data acquisition.

[0003] In practical engineering monitoring, it is often necessary to monitor the deformation of multiple targets over long periods of time, especially in the monitoring of large-scale structures such as bridges and high-rise buildings. However, due to the limited depth of field of cameras and the constraints of lens parameters and target depth variations, images often contain both sharp and blurred areas simultaneously. This results in some targets becoming out of focus, with blurred edges and difficult to identify. This blurring not only affects the accuracy of target detection and image segmentation but also interferes with subsequent displacement extraction and deformation analysis, becoming one of the key issues restricting the accuracy of visual monitoring.

[0004] Most commonly used image segmentation algorithms rely on clear texture and edge information, which can easily lead to boundary misclassification when processing blurry images or low-contrast regions. Especially under different depth-of-field conditions, multiple targets exhibit significant differences in sharpness, making it difficult for traditional segmentation models to ensure consistent recognition of all targets. Furthermore, existing multi-target processing workflows mostly perform uniform calculations on the entire image, lacking optimization strategies based on local target regions. This not only results in redundant computational resource consumption but also reduces processing efficiency under high frame rates and high concurrency tasks.

[0005] In image quality enhancement, traditional deblurring algorithms are mostly based on physical modeling or filtering methods, which have limited effectiveness in dealing with out-of-focus blur and complex interference from non-uniform blur in real-world scenes. Although deep learning-based deblurring neural networks have made some progress in recent years, they still have significant shortcomings in multi-target dynamic updates and automatic adaptation, especially in multi-target parallel processing and real-time processing capabilities, which are difficult to meet the needs of engineering applications.

[0006] In summary, current structural multi-target visual monitoring technology still faces many technical bottlenecks in handling blurred target recognition under different depth-of-field conditions, accurate cropping and enhancement of multiple targets, and efficient parallel processing. Summary of the Invention

[0007] To address the technical problems of target blurring and defocusing, inaccurate segmentation, and low processing efficiency in existing technologies for multi-depth-of-field and non-coplanar scenes, this invention provides a method and apparatus for monitoring multi-target deformation based on visual segmentation and deblurring enhancement. The technical solution is as follows:

[0008] On the one hand, a method for monitoring multi-target structural deformation based on visual segmentation and deblurring enhancement is provided. This method is implemented by a multi-target structural deformation monitoring device based on visual segmentation and deblurring enhancement, and includes:

[0009] S1. Acquire continuous temporal multi-target images of different depth planes of the structure to be monitored; based on the pre-selected foreground cue points, input the first frame image of the multi-target images into the pre-trained SAM segmentation model, and output a preliminary multi-target segmentation mask;

[0010] S2. Based on the preliminary multi-target segmentation mask, noise is automatically removed using pre-selected foreground cue points and connected component analysis to obtain effective connected regions, and these effective connected regions are numbered.

[0011] S3. Extract the bounding rectangle of the effective connected region, and use an adaptive multi-scale clipping strategy based on boundary conditions to clip it, thereby obtaining a multi-target target region sub-map.

[0012] S4. Input the multi-target target region sub-map into the pre-trained NAFNet network model for enhancement processing to obtain the enhanced multi-target sub-map;

[0013] S5. Extract the center point and circumscribed rectangle coordinates of the target mask region in the enhanced multi-target sub-image and input them into the visual segmentation model for segmentation processing. After segmentation, calculate the center point coordinates of each target mask and map the coordinates in the enhanced multi-target sub-image back to the original full-frame image coordinate system through coordinate conversion. Output the coordinates of the center point and circumscribed rectangle of the first frame image after segmentation.

[0014] S6. Based on the coordinates of the center point and the bounding rectangle of the first frame image, the position of the sub-image of the next frame image is automatically updated using a dynamic update strategy for the cropping region based on mask boundary determination, thereby generating the coordinates of the center point and the bounding rectangle of the next frame image.

[0015] S7. Compare the pixel center positions of each target in the original image coordinate system in the first frame and the next frame, and compare them with the camera calibration parameters to obtain the multi-target deformation monitoring results.

[0016] Optionally, S1, which involves acquiring continuous temporal multi-target images of different depth planes of the structure to be monitored, includes:

[0017] Multiple circular targets are pre-positioned on the surface of the structure to be monitored, distributed in different depth planes. An industrial camera with a fixed focal length and high resolution is used to acquire images, obtaining continuous time-series multi-target images of the surface of the structure to be monitored in different depth planes.

[0018] Optionally, step S2 involves automatic noise removal using pre-selected foreground cue points and connected component analysis, including:

[0019] Based on pre-selected foreground cue points and connected component analysis, the method determines whether the selected points are located inside the connected component. If the pre-selected foreground cue points are located inside the connected component, they are determined to be valid regions; otherwise, they are determined to be noise and replaced with background pixels, thus obtaining a binary mask containing only the target.

[0020] Optionally, the extraction of the circumscribed rectangle and center point of the effective connected region in S3 employs an adaptive multi-scale clipping strategy based on boundary conditions, including:

[0021] S31. Extract the bounding rectangle for each valid connected region, and perform preliminary clipping based on the geometric dimensions of each valid connected region to obtain the clipped sub-image;

[0022] S32. Input the cropped sub-image into the pre-trained SAM segmentation model, determine whether the foreground mask touches the boundary and there is boundary overlap. If there is boundary overlap, gradually expand the cropping boundary in the corresponding direction according to the set ratio until the foreground mask is completely covered in the cropping box without overlap, and then stop cropping.

[0023] Optionally, step S6 automatically updates the sub-region positions of the next frame image based on the coordinates of the center point and the bounding rectangle of the first frame image, using a dynamic update strategy for the cropping region based on mask boundary determination, including:

[0024] S61. Set the initial cropping box size and pixel center coordinates for the i-th target;

[0025] S62. Traverse all foreground pixels in the i-th target mask, calculate the minimum distance from all foreground pixels to the clipping box boundary, determine whether the pixel meets the defined coverage condition, if not, perform the expansion mechanism, expand the boundary according to the set ratio and update the center coordinates until the coverage condition is met.

[0026] Optionally, the coverage condition is expressed by the following formula (1):

[0027] (1)

[0028] in, This represents the x-coordinate of the j-th foreground pixel in the i-th target mask region; This indicates the coordinates of the left boundary of the current cropping box, which is the pixel position of the smallest x-axis on the left. This indicates the right boundary coordinates of the current cropping box, which is the pixel position of the maximum x value on the right. This represents the y-coordinate of the j-th foreground pixel in the i-th target mask region; This indicates the coordinates of the upper boundary of the current cropping box, which is the pixel position of the smallest y-value on the upper side; This indicates the coordinates of the lower boundary of the current cropping box, which is the pixel position of the maximum y-value at the bottom. This represents the distance threshold, used to determine whether the mask touches the clipping box boundary.

[0029] Optionally, step S7 compares the pixel center positions of the target masks in each segmented sub-region of the first frame and the next frame in the original image coordinate system, and compares them in conjunction with camera calibration parameters to obtain multi-target deformation monitoring results, including:

[0030] The pixel center positions of the target mask in each segmented sub-region in the first frame and the next frame are compared in the original image coordinate system. The pixel-level displacement difference of each target center position in the next frame is calculated. Combined with the camera's internal and external parameters and structural dimensions, the pixel displacement difference is converted into physical unit displacement or three-dimensional coordinate change to obtain the multi-target deformation monitoring results.

[0031] On the other hand, a multi-target structural deformation monitoring device based on visual segmentation and deblurring enhancement is provided. This device is applied to a multi-target structural deformation monitoring method based on visual segmentation and deblurring enhancement. The device includes:

[0032] The acquisition unit is used to acquire continuous temporal multi-target images of different depth planes of the structure to be monitored; based on the pre-selected foreground cue points, the first frame image of the multi-target images is input into the pre-trained SAM segmentation model, and a preliminary multi-target segmentation mask is output.

[0033] The processing unit is used to automatically remove noise based on the preliminary multi-target segmentation mask, using pre-selected foreground cue points and connected component analysis to obtain effective connected regions, and to number the effective connected regions.

[0034] The clipping unit is used to extract the bounding rectangle of the effective connected region. It adopts an adaptive multi-scale clipping strategy based on boundary conditions to clip and obtain a multi-target target region sub-map.

[0035] The enhancement unit is used to input the multi-target target region sub-map into the pre-trained NAFNet network model for enhancement processing to obtain the enhanced multi-target sub-map;

[0036] The segmentation unit is used to extract the center point and bounding rectangle coordinates of the target mask region in the enhanced multi-target sub-image and input them into the visual segmentation model for segmentation processing. After segmentation, the center point coordinates of each target mask are calculated, and the coordinates in the enhanced multi-target sub-image are uniformly mapped back to the original full-frame image coordinate system through coordinate conversion. The coordinates of the center point and bounding rectangle of the first frame image after segmentation are output.

[0037] The update unit is used to automatically update the sub-image position of the next frame image based on the coordinates of the center point and the bounding rectangle of the first frame image, using a dynamic update strategy for the cropping region based on mask boundary determination, and to generate the coordinates of the center point and the bounding rectangle of the next frame image.

[0038] The monitoring unit is used to compare the pixel center positions of each target in the original image coordinate system in the first frame and the next frame, and compare them with the camera calibration parameters to obtain the multi-target deformation monitoring results.

[0039] Optionally, acquiring continuous temporal multi-target images of different depth planes of the structure to be monitored includes:

[0040] Multiple circular targets are pre-positioned on the surface of the structure to be monitored, distributed in different depth planes. An industrial camera with a fixed focal length and high resolution is used to acquire images, obtaining continuous time-series multi-target images of the surface of the structure to be monitored in different depth planes.

[0041] Optionally, the processing unit is configured to:

[0042] Based on pre-selected foreground cue points and connected component analysis, the method determines whether the selected points are located inside the connected component. If the pre-selected foreground cue points are located inside the connected component, they are determined to be valid regions; otherwise, they are determined to be noise and replaced with background pixels, thus obtaining a binary mask containing only the target.

[0043] Optionally, the cutting unit is used for:

[0044] Extract the bounding rectangle for each valid connected region, and perform preliminary clipping based on the geometric dimensions of each valid connected region to obtain the clipped subgraph;

[0045] The cropped sub-image is input into the pre-trained SAM segmentation model to determine whether the foreground mask touches the boundary and whether there is boundary overlap. If there is boundary overlap, the cropping boundary in the corresponding direction is gradually expanded according to the set ratio until the foreground mask is completely covered in the cropping box without overlap, and then the cropping stops.

[0046] Optionally, the update unit is used for:

[0047] Let the initial cropping box size and pixel center coordinates of the i-th target be defined;

[0048] Iterate through all foreground pixels in the i-th target mask, calculate the minimum distance from all foreground pixels to the clipping box boundary, and determine whether the pixel meets the defined coverage condition. If not, perform an expansion mechanism to expand the boundary according to the set ratio and update the center coordinates until the coverage condition is met.

[0049] Optionally, the coverage condition is expressed by the following formula (1):

[0050] (1)

[0051] in, This represents the x-coordinate of the j-th foreground pixel in the i-th target mask region; This indicates the coordinates of the left boundary of the current cropping box, which is the pixel position of the smallest x-axis on the left. This indicates the right boundary coordinates of the current cropping box, which is the pixel position of the maximum x value on the right. This represents the y-coordinate of the j-th foreground pixel in the i-th target mask region; This indicates the coordinates of the upper boundary of the current cropping box, which is the pixel position of the smallest y-value on the upper side; This indicates the coordinates of the lower boundary of the current cropping box, which is the pixel position of the maximum y-value at the bottom. This represents the distance threshold, used to determine whether the mask touches the clipping box boundary.

[0052] Optionally, the monitoring unit is used for:

[0053] The pixel center positions of the target mask in each segmented sub-region in the first frame and the next frame are compared in the original image coordinate system. The pixel-level displacement difference of each target center position in the next frame is calculated. Combined with the camera's internal and external parameters and structural dimensions, the pixel displacement difference is converted into physical unit displacement or three-dimensional coordinate change to obtain the multi-target deformation monitoring results.

[0054] On the other hand, a multi-target structural deformation monitoring device based on visual segmentation and deblurring enhancement is provided. The multi-target structural deformation monitoring device based on visual segmentation and deblurring enhancement includes: a processor; a memory, wherein the memory stores computer-readable instructions, and when the computer-readable instructions are executed by the processor, they implement any of the methods described above for multi-target structural deformation monitoring based on visual segmentation and deblurring enhancement.

[0055] On the other hand, a computer-readable storage medium is provided, wherein at least one instruction is stored in the storage medium, the at least one instruction being loaded and executed by a processor to implement any of the above-described methods for monitoring multi-target structural deformation based on visual segmentation and deblurring enhancement.

[0056] The beneficial effects of the technical solutions provided in the embodiments of the present invention include at least the following:

[0057] This invention provides a method for monitoring structural multi-target deformation based on visual segmentation and deblurring enhancement. It enables high-resolution and continuous automated monitoring of multiple targets in complex scenes with varying depths of field, overcoming the problems of inaccurate segmentation and low measurement precision in traditional methods under conditions of perspective distortion and out-of-focus blur. This invention effectively improves the accuracy of multi-target recognition and localization through interactive segmentation and noise removal strategies based on prompts. An adaptive multi-scale cropping method based on boundary conditions dynamically adjusts the cropping window across different frames and depths of field, avoiding truncation or introducing redundant background and ensuring complete target coverage. A pre-trained deblurring neural network model enhances the cropped sub-regions, improving the clarity and segmentation quality of out-of-focus areas. Finally, a GPU parallel computing framework supports pipelined acceleration processing of multiple sub-regions, achieving efficient and scalable computation in high-resolution, large-scale multi-target scenes. This method also supports high-precision pixel-level displacement measurement through unified coordinate mapping from sub-regions to the whole map and continuous tracking across multiple frames. Combined with camera calibration parameters, it enables accurate monitoring in physical units or three-dimensional coordinates. The overall process is highly automated and has strong anti-interference capabilities. It is suitable for applications such as structural health monitoring and industrial multi-target deformation measurement, and has good promotional value and application prospects. Attached Figure Description

[0058] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0059] Figure 1 This is a flowchart of a multi-target structural deformation monitoring method based on visual segmentation and deblurring enhancement provided by an embodiment of the present invention;

[0060] Figure 2 This is a schematic diagram of a multi-target deformation monitoring method based on visual segmentation and deblurring enhancement provided in an embodiment of the present invention;

[0061] Figure 3 This is a block diagram of a multi-target structural deformation monitoring device based on visual segmentation and deblurring enhancement provided in an embodiment of the present invention;

[0062] Figure 4 This is a schematic diagram of a multi-target deformation monitoring device based on visual segmentation and deblurring enhancement provided in an embodiment of the present invention. Detailed Implementation

[0063] The technical solution of the present invention will now be described with reference to the accompanying drawings.

[0064] In embodiments of the present invention, words such as "exemplarily," "for example," etc., are used to indicate that something is an example, illustration, or description. Any embodiment or design described as "exemplary" in the present invention should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of the word "exemplary" is intended to present the concept in a concrete manner. Furthermore, in embodiments of the present invention, the meaning expressed by "and / or" can be both, or either one.

[0065] In the embodiments of this invention, the terms "image" and "picture" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning. Similarly, the terms "of," "corresponding (relevant)," and "corresponding" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning.

[0066] In this embodiment of the invention, sometimes a subscript such as W1 may be written in a non-subscript form such as W1. When the difference is not emphasized, the meaning they express is the same.

[0067] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0068] This invention provides a method for monitoring multi-target structural deformation based on visual segmentation and deblurring enhancement. This method can be implemented using a device for monitoring multi-target structural deformation based on visual segmentation and deblurring enhancement, which can be a terminal or a server. Figure 1 The flowchart shown is for a multi-target structural deformation monitoring method based on visual segmentation and deblurring enhancement. The processing flow of this method may include the following steps:

[0069] S1. Acquire continuous temporal multi-target images of different depth planes of the structure to be monitored; based on the pre-selected foreground cue points, input the first frame image of the multi-target images into the pre-trained SAM segmentation model, and output a preliminary multi-target segmentation mask.

[0070] The foreground cue points are pre-selected through manual interactive selection in the first frame image.

[0071] Optionally, S1 acquires continuous temporal multi-target images of different depth planes of the structure to be monitored, including:

[0072] Multiple circular targets are pre-positioned on the surface of the structure to be monitored, distributed in different depth planes. An industrial camera with a fixed focal length and high resolution is used to acquire images, obtaining continuous time-series multi-target images of the surface of the structure to be monitored in different depth planes.

[0073] In one feasible implementation, the embodiments of the present invention employ a fixed focal length and high resolution industrial camera mounted on a tripod or stabilizing bracket to reduce perspective distortion caused by changes in viewing angle during image acquisition. The resolution of the multi-target images acquired in the embodiments of the present invention is 5120×5120.

[0074] In the first frame image, foreground cue points can be interactively selected at any location within each target area.

[0075] In this embodiment of the invention, the initial multi-target segmentation mask is obtained using a pre-trained SAM segmentation model. Other segmentation models may also be used, and this embodiment of the invention does not impose any limitations.

[0076] S2. Based on the preliminary multi-target segmentation mask, noise is automatically removed using pre-selected foreground cue points and connected component analysis to obtain effective connected regions, and these effective connected regions are numbered.

[0077] Among these, the process of obtaining effective connected regions can improve the accuracy and robustness of the initial segmentation.

[0078] Optionally, S2 employs pre-selected foreground cue points and connected component analysis for automatic noise removal, including:

[0079] Based on pre-selected foreground cue points and connected component analysis, the method determines whether the selected points are located inside the connected component. If the pre-selected foreground cue points are located inside the connected component, they are determined to be valid regions; otherwise, they are determined to be noise and replaced with background pixels, thus obtaining a binary mask containing only the target.

[0080] S3. Extract the bounding rectangle of the effective connected region, and use an adaptive multi-scale clipping strategy based on boundary conditions to clip the region and obtain a sub-map of the multi-target region.

[0081] Among them, the adaptive multi-scale cropping strategy dynamically expands the cropping boundary based on the bounding rectangle information of each effective connected domain, avoiding the loss of edge information due to excessively narrow cropping or the introduction of background noise due to excessively wide cropping, thus achieving optimal coverage, that is, fully preserving the target while suppressing redundant background interference.

[0082] Optionally, the specific implementation process of S3 includes S31-S32:

[0083] S31. Extract the bounding rectangle for each valid connected region, and perform preliminary clipping based on the geometric dimensions of each valid connected region to obtain the clipped sub-image;

[0084] S32. Input the cropped sub-image into the pre-trained SAM segmentation model, determine whether the foreground mask touches the boundary and there is boundary overlap. If there is boundary overlap, gradually expand the cropping boundary in the corresponding direction according to the set ratio until the foreground mask is completely covered in the cropping box without overlap, and then stop cropping.

[0085] S4. Input the multi-target target region sub-map into the pre-trained NAFNet network model for enhancement processing to obtain the enhanced multi-target sub-map.

[0086] In this embodiment of the invention, the sub-regions of each cropped target are input into the pre-trained NAFNet network model, which can improve the clarity and feature representation quality of blurred regions and enhance the accuracy of subsequent segmentation and localization.

[0087] In addition to using a pre-trained NAFNet network model for deblurring and enhancement in this embodiment of the invention, other segmentation models may also be used, and this embodiment of the invention does not impose any limitations.

[0088] S5. Extract the center point and circumscribed rectangle coordinates of the target mask region within the enhanced multi-target sub-image and input them into the visual segmentation model for segmentation. After segmentation, calculate the center point coordinates of each target mask and, through coordinate conversion, map the coordinates within the enhanced multi-target sub-image back to the original full-frame image coordinate system. Output the coordinates of the center point and circumscribed rectangle of the first frame image after segmentation.

[0089] Mapping the coordinates within a sub-region back to the original full-frame image coordinate system enables globally consistent management and visual annotation of multiple targets.

[0090] In this embodiment of the invention, a GPU-based pipelined parallel processing framework is constructed to simultaneously perform deblurring, segmentation, boundary expansion, and coordinate mapping on sub-regions. Load balancing and dynamic memory allocation strategies are employed to improve the speed and resource utilization of large-scale batch processing.

[0091] S6. Based on the coordinates of the center point and the bounding rectangle of the first frame image, the position of the sub-image of the next frame image is automatically updated using a dynamic update strategy for the cropping region based on mask boundary determination, thereby generating the coordinates of the center point and the bounding rectangle of the next frame image.

[0092] In the subsequent temporal frame processing, based on the center point and circumscribed rectangle information of the previous frame, a clipping window update strategy based on mask boundary determination is adopted to achieve adaptive smooth adjustment of the clipping region and maintain the tracking stability between consecutive frames.

[0093] In one feasible implementation, the clipping window update strategy based on mask boundary determination can achieve smooth transition and dynamic adjustment of the clipping window within the time frame, ensuring complete coverage of the target.

[0094] Optionally, the specific process of S6 includes S61-S62:

[0095] S61. Set the initial cropping box size and pixel center coordinates for the i-th target;

[0096] In one feasible implementation, let the initial clipping frame size of the i-th target be... ,in This represents the width of the i-th target. Let represent the height of the i-th target, with the pixel center coordinates being (xi, yi);

[0097] S62. Traverse all foreground pixels in the i-th target mask, calculate the minimum distance from all foreground pixels to the clipping box boundary, determine whether the pixel meets the defined coverage condition, if not, perform the expansion mechanism, expand the boundary according to the set ratio and update the center coordinates until the coverage condition is met.

[0098] Specifically, for each binarized mask corresponding to the cropping region map, if there are pixels in the foreground region that are less than the cropping boundary by a set threshold ε, it is determined that the current cropping region is insufficient to cover the entire target region, and adaptive expansion processing is required.

[0099] Alternatively, the coverage condition can be expressed by the following formula (1):

[0100] (1)

[0101] in, This represents the x-coordinate of the j-th foreground pixel in the i-th target mask region; This indicates the coordinates of the left boundary of the current cropping box, which is the pixel position of the smallest x-axis on the left. This indicates the right boundary coordinates of the current cropping box, which is the pixel position of the maximum x value on the right. This represents the y-coordinate of the j-th foreground pixel in the i-th target mask region; This indicates the coordinates of the upper boundary of the current cropping box, which is the pixel position of the smallest y-value on the upper side; This indicates the coordinates of the lower boundary of the current cropping box, which is the pixel position of the maximum y-value at the bottom. This represents the distance threshold, used to determine whether the mask touches the clipping box boundary.

[0102] S7. Compare the pixel center positions of each target in the original image coordinate system in the first frame and the next frame, and compare them with the camera calibration parameters to obtain the multi-target deformation monitoring results.

[0103] Optionally, the specific implementation process of S7 includes:

[0104] The pixel center positions of the target mask in each segmented sub-region in the first frame and the next frame are compared in the original image coordinate system. The pixel-level displacement difference of each target center position in the next frame is calculated. Combined with the camera's internal and external parameters and structural dimensions, the pixel displacement difference is converted into physical unit displacement or three-dimensional coordinate change to obtain the multi-target deformation monitoring results.

[0105] in, Figure 2 This is a schematic diagram of a structural multi-target deformation monitoring method based on visual segmentation and deblurring enhancement provided by an embodiment of the present invention. In one feasible implementation, a time-series image to be monitored is acquired; the first frame image is input into a pre-trained interactive visual segmentation model to obtain an initial segmentation mask; noise processing is performed based on the initial segmentation mask using pre-selected foreground cue points and connected component analysis to obtain a processed segmentation mask; based on the processed segmentation mask, an adaptive multi-scale cropping strategy based on boundary conditions is used for cropping to obtain sub-images of each target region after cropping; the sub-images of each target region after cropping are input into a pre-trained deblurring enhancement model. In the fuzzy neural network model, sub-images of each target region after enhancement are obtained; the coordinates of the center point and the bounding rectangle of the target mask region within the enhanced multi-target sub-image are calculated and processed by a visual segmentation model to obtain the segmented target region masks; the center point coordinates of each segmented target region mask are calculated, and the sub-region position of the next frame image is automatically updated based on the center point coordinates to obtain the center point coordinates of each target region mask after segmentation in the next frame image; the pixel center point positions of each target in the first frame and the next frame in the original image coordinate system are compared and compared with camera calibration parameters to obtain the multi-target deformation monitoring results.

[0106] The embodiments of the present invention can achieve high-precision displacement extraction and change detection of complex structures under non-coplanar, multi-target and multi-depth conditions. It has the advantages of high degree of automation of processing flow, strong positioning accuracy and high system processing efficiency, and is suitable for structural deformation monitoring scenarios such as bridges, tunnels and tunnel arches.

[0107] This invention provides a method for monitoring structural multi-target deformation based on visual segmentation and deblurring enhancement. It enables high-resolution and continuous automated monitoring of multiple targets in complex scenes with varying depths of field, overcoming the problems of inaccurate segmentation and low measurement precision in traditional methods under conditions of perspective distortion and out-of-focus blur. This invention effectively improves the accuracy of multi-target recognition and localization through interactive segmentation and noise removal strategies based on prompts. An adaptive multi-scale cropping method based on boundary conditions dynamically adjusts the cropping window across different frames and depths of field, avoiding truncation or introducing redundant background and ensuring complete target coverage. A pre-trained deblurring neural network model enhances the cropped sub-regions, improving the clarity and segmentation quality of out-of-focus areas. Finally, a GPU parallel computing framework supports pipelined acceleration processing of multiple sub-regions, achieving efficient and scalable computation in high-resolution, large-scale multi-target scenes. This method also supports high-precision pixel-level displacement measurement through unified coordinate mapping from sub-regions to the whole map and continuous tracking across multiple frames. Combined with camera calibration parameters, it enables accurate monitoring in physical units or three-dimensional coordinates. The overall process is highly automated and has strong anti-interference capabilities. It is suitable for applications such as structural health monitoring and industrial multi-target deformation measurement, and has good promotional value and application prospects.

[0108] Figure 3 This is a block diagram of a structural multi-target deformation monitoring device based on visual segmentation and deblurring enhancement, provided by an embodiment of the present invention. This device is used in a structural multi-target deformation monitoring method based on visual segmentation and deblurring enhancement. (Refer to...) Figure 3 The device includes a first acquisition unit 310, a processing unit 320, a trimming unit 330, an enhancement unit 340, a segmentation unit 350, an update unit 360, and a monitoring unit 370. Wherein:

[0109] The acquisition unit 310 is used to acquire continuous temporal multi-target images of different depth planes of the structure surface to be monitored; based on the pre-selected foreground cue points, the first frame image of the multi-target image is input into the pre-trained SAM segmentation model, and a preliminary multi-target segmentation mask is output.

[0110] The processing unit 320 is used to automatically remove noise based on the preliminary multi-target segmentation mask, using pre-selected foreground cue points and connected component analysis to obtain effective connected regions, and to number the effective connected regions.

[0111] The clipping unit 330 is used to extract the bounding rectangle of the effective connected region. It adopts an adaptive multi-scale clipping strategy based on boundary conditions to clip and obtain a sub-map of the multi-target region.

[0112] The enhancement unit 340 is used to input the multi-target target region sub-map into the pre-trained NAFNet network model for enhancement processing to obtain the enhanced multi-target sub-map;

[0113] The segmentation unit 350 is used to extract the center point and circumscribed rectangle coordinates of the target mask region in the enhanced multi-target sub-image and input them into the visual segmentation model for segmentation processing. After segmentation, the center point coordinates of each target mask are calculated, and the coordinates in the enhanced multi-target sub-image are uniformly mapped back to the original full-frame image coordinate system through coordinate conversion. The coordinates of the center point and circumscribed rectangle of the first frame image after segmentation are output.

[0114] The update unit 360 is used to automatically update the sub-image position of the next frame image based on the coordinates of the center point and the bounding rectangle of the first frame image, using a dynamic update strategy for the cropping region based on mask boundary determination, and to generate the coordinates of the center point and the bounding rectangle of the next frame image.

[0115] The monitoring unit 370 is used to compare the pixel center positions of each target in the original image coordinate system in the first frame and the next frame, and compare them with the camera calibration parameters to obtain the multi-target deformation monitoring results.

[0116] Optionally, acquiring continuous temporal multi-target images of different depth planes of the structure to be monitored includes:

[0117] Multiple circular targets are pre-positioned on the surface of the structure to be monitored, distributed in different depth planes. An industrial camera with a fixed focal length and high resolution is used to acquire images, obtaining continuous time-series multi-target images of the surface of the structure to be monitored in different depth planes.

[0118] Optionally, the processing unit 320 is configured to:

[0119] Based on pre-selected foreground cue points and connected component analysis, the method determines whether the selected points are located inside the connected component. If the pre-selected foreground cue points are located inside the connected component, they are determined to be valid regions; otherwise, they are determined to be noise and replaced with background pixels, thus obtaining a binary mask containing only the target.

[0120] Optionally, the cutting unit 330 is used for:

[0121] Extract the bounding rectangle for each valid connected region, and perform preliminary clipping based on the geometric dimensions of each valid connected region to obtain the clipped subgraph;

[0122] The cropped sub-image is input into the pre-trained SAM segmentation model to determine whether the foreground mask touches the boundary and whether there is boundary overlap. If there is boundary overlap, the cropping boundary in the corresponding direction is gradually expanded according to the set ratio until the foreground mask is completely covered in the cropping box without overlap, and then the cropping stops.

[0123] Optionally, the updating unit 360 is used to:

[0124] Let the initial cropping box size and pixel center coordinates of the i-th target be defined;

[0125] Iterate through all foreground pixels in the i-th target mask, calculate the minimum distance from all foreground pixels to the clipping box boundary, and determine whether the pixel meets the defined coverage condition. If not, perform an expansion mechanism to expand the boundary according to the set ratio and update the center coordinates until the coverage condition is met.

[0126] Optionally, the coverage condition is expressed by the following formula (1):

[0127] (1)

[0128] in, This represents the x-coordinate of the j-th foreground pixel in the i-th target mask region; This indicates the coordinates of the left boundary of the current cropping box, which is the pixel position of the smallest x-axis on the left. This indicates the right boundary coordinates of the current cropping box, which is the pixel position of the maximum x value on the right. This represents the y-coordinate of the j-th foreground pixel in the i-th target mask region; This indicates the coordinates of the upper boundary of the current cropping box, which is the pixel position of the smallest y-value on the upper side; This indicates the coordinates of the lower boundary of the current cropping box, which is the pixel position of the maximum y-value at the bottom. This represents the distance threshold, used to determine whether the mask touches the clipping box boundary.

[0129] Optionally, the monitoring unit 370 is used for:

[0130] The pixel center positions of the target mask in each segmented sub-region in the first frame and the next frame are compared in the original image coordinate system. The pixel-level displacement difference of each target center position in the next frame is calculated. Combined with the camera's internal and external parameters and structural dimensions, the pixel displacement difference is converted into physical unit displacement or three-dimensional coordinate change to obtain the multi-target deformation monitoring results.

[0131] This invention provides a method for monitoring structural multi-target deformation based on visual segmentation and deblurring enhancement. It enables high-resolution and continuous automated monitoring of multiple targets in complex scenes with varying depths of field, overcoming the problems of inaccurate segmentation and low measurement precision in traditional methods under conditions of perspective distortion and out-of-focus blur. This invention effectively improves the accuracy of multi-target recognition and localization through interactive segmentation and noise removal strategies based on prompts. An adaptive multi-scale cropping method based on boundary conditions dynamically adjusts the cropping window across different frames and depths of field, avoiding truncation or introducing redundant background and ensuring complete target coverage. A pre-trained deblurring neural network model enhances the cropped sub-regions, improving the clarity and segmentation quality of out-of-focus areas. Finally, a GPU parallel computing framework supports pipelined acceleration processing of multiple sub-regions, achieving efficient and scalable computation in high-resolution, large-scale multi-target scenes. This method also supports high-precision pixel-level displacement measurement through unified coordinate mapping from sub-regions to the whole map and continuous tracking across multiple frames. Combined with camera calibration parameters, it enables accurate monitoring in physical units or three-dimensional coordinates. The overall process is highly automated and has strong anti-interference capabilities. It is suitable for applications such as structural health monitoring and industrial multi-target deformation measurement, and has good promotional value and application prospects.

[0132] Figure 4 This is a schematic diagram of a multi-target deformation monitoring device based on visual segmentation and deblurring enhancement provided in an embodiment of the present invention, as shown below. Figure 4 As shown, the structural multi-target deformation monitoring device based on visual segmentation and deblurring enhancement may include the above-mentioned... Figure 3 The illustrated multi-target structural deformation monitoring device is based on visual segmentation and deblurring enhancement. Optionally, the multi-target structural deformation monitoring device 410 based on visual segmentation and deblurring enhancement may include a first processor 2001.

[0133] Optionally, the structure multi-target deformation monitoring device 410 based on visual segmentation and deblurring enhancement may also include a memory 2002 and a transceiver 2003.

[0134] The first processor 2001, memory 2002, and transceiver 2003 can be connected via a communication bus.

[0135] The following is combined Figure 4 The components of the multi-target structural deformation monitoring device 410 based on visual segmentation and deblurring enhancement are described in detail below:

[0136] The first processor 2001 is the control center of the multi-target deformation monitoring device 410 based on visual segmentation and deblurring enhancement. It can be a single processor or a collective term for multiple processing elements. For example, the first processor 2001 can be one or more central processing units (CPUs), application-specific integrated circuits (ASICs), or one or more integrated circuits configured to implement embodiments of the present invention, such as one or more digital signal processors (DSPs), or one or more field-programmable gate arrays (FPGAs).

[0137] Optionally, the first processor 2001 can execute various functions of the visual segmentation and deblurring enhancement-based multi-target deformation monitoring device 410 by running or executing software programs stored in the memory 2002 and calling data stored in the memory 2002.

[0138] In a specific implementation, as one example, the first processor 2001 may include one or more CPUs, for example... Figure 4 CPU0 and CPU1 are shown in the diagram.

[0139] In a specific implementation, as one example, the multi-target deformation monitoring device 410 based on visual segmentation and deblurring enhancement may also include multiple processors, for example... Figure 4 The first processor 2001 and the second processor 2004 are shown in the diagram. Each of these processors can be a single-core processor or a multi-core processor. Here, a processor can refer to one or more devices, circuits, and / or processing cores used to process data (such as computer program instructions).

[0140] The memory 2002 is used to store the software program that executes the present invention, and is controlled by the first processor 2001 to execute it. The specific implementation method can be referred to the above method embodiment, and will not be repeated here.

[0141] Optionally, the memory 2002 may be a read-only memory (ROM) or other type of static storage device capable of storing static information and instructions, random access memory (RAM) or other type of dynamic storage device capable of storing information and instructions, or electrically erasable programmable read-only memory (EEPROM), compact disc read-only memory (CD-ROM) or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, digital universal optical discs, Blu-ray discs, etc.), magnetic disk storage media or other magnetic storage devices, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but not limited thereto. The memory 2002 may be integrated with the first processor 2001 or exist independently, and may be connected via the interface circuit of the structure multi-target deformation monitoring device 410 based on visual segmentation and deblurring enhancement. Figure 4 (Not shown in the image) is coupled to the first processor 2001, and this embodiment of the invention does not specifically limit this.

[0142] The transceiver 2003 is used to communicate with network devices or with terminal devices.

[0143] Alternatively, transceiver 2003 may include a receiver and a transmitter. Figure 4 (Not shown separately). The receiver is used to implement the receiving function, and the transmitter is used to implement the transmitting function.

[0144] Optionally, the transceiver 2003 can be integrated with the first processor 2001 or exist independently, and can be connected to the interface circuit of the structure multi-target deformation monitoring device 410 based on visual segmentation and deblurring enhancement. Figure 4 (Not shown in the image) is coupled to the first processor 2001, and this embodiment of the invention does not specifically limit this.

[0145] It should be noted that, Figure 4 The structure of the multi-target deformation monitoring device 410 based on visual segmentation and deblurring enhancement shown in the figure does not constitute a limitation on the router. The actual multi-target deformation monitoring device based on visual segmentation and deblurring enhancement may include more or fewer components than shown, or combine some components, or have different component arrangements.

[0146] Furthermore, the technical effects of the multi-target structural deformation monitoring device 410 based on visual segmentation and deblurring enhancement can be referred to the technical effects of the multi-target structural deformation monitoring method based on visual segmentation and deblurring enhancement described in the above method embodiments, and will not be repeated here.

[0147] It should be understood that the first processor 2001 in the embodiments of the present invention may be a central processing unit (CPU), or it may be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor, or it may be any conventional processor, etc.

[0148] It should also be understood that the memory in the embodiments of the present invention can be volatile memory or non-volatile memory, or may include both volatile and non-volatile memory. The non-volatile memory can be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory. The volatile memory can be random access memory (RAM), which is used as an external cache. By way of example, but not limitation, many forms of random access memory (RAM) are available, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate synchronous DRAM (DDR SDRAM), enhanced synchronous DRAM (ESDRAM), synchronous linked DRAM (SLDRAM), and direct rambus RAM (DR RAM).

[0149] The above embodiments can be implemented, in whole or in part, by software, hardware (such as circuits), firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of the present invention are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more sets of available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. A semiconductor medium can be a solid-state drive.

[0150] It should be understood that the term "and / or" in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. A and B can be singular or plural. Additionally, the character " / " in this article generally indicates an "or" relationship between the preceding and following related objects, but it can also represent an "and / or" relationship. Please refer to the context for a more accurate understanding.

[0151] In this invention, "at least one" means one or more, and "more than one" means two or more. "At least one of the following" or similar expressions refer to any combination of these items, including any combination of a single item or a plurality of items. For example, at least one of a, b, or c can represent: a, b, c, ab, ac, bc, or abc, where a, b, and c can be a single item or multiple items.

[0152] It should be understood that, in various embodiments of the present invention, the sequence number of each process does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0153] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0154] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the devices, apparatuses, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0155] In the several embodiments provided by this invention, it should be understood that the disclosed devices, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another device, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.

[0156] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0157] In addition, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0158] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0159] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for monitoring multi-target structural deformation based on visual segmentation and deblurring enhancement, characterized in that, The method includes: S1. Acquire continuous temporal multi-target images of different depth planes of the structure to be monitored; based on the pre-selected foreground cue points, input the first frame image of the multi-target images into the pre-trained SAM segmentation model, and output a preliminary multi-target segmentation mask; S2. Based on the preliminary multi-target segmentation mask, noise is automatically removed using pre-selected foreground cue points and connected component analysis to obtain effective connected regions, and these effective connected regions are numbered. S3. Extract the bounding rectangle of the effective connected region, and use an adaptive multi-scale clipping strategy based on boundary conditions to clip it, thereby obtaining a multi-target target region sub-map. Specifically, S3, which extracts the circumscribed rectangle and center point of the effective connected region, employs an adaptive multi-scale pruning strategy based on boundary conditions for pruning, including: S31. Extract the bounding rectangle for each valid connected region, and perform preliminary clipping based on the geometric dimensions of each valid connected region to obtain the clipped sub-image. S32. Input the cropped sub-image into the pre-trained SAM segmentation model, determine whether the foreground mask touches the boundary and there is boundary overlap. If there is boundary overlap, gradually expand the cropping boundary in the corresponding direction according to the set ratio until the foreground mask is completely covered in the cropping box and there is no overlap, and stop cropping. S4. Input the multi-target target region sub-map into the pre-trained NAFNet network model for enhancement processing to obtain the enhanced multi-target sub-map; S5. Extract the center point and circumscribed rectangle coordinates of the target mask region within the enhanced multi-target image, and input the enhanced multi-target image into the visual segmentation model for segmentation processing to obtain each target mask; calculate the center point coordinates of each target mask, and map the coordinates within the enhanced multi-target image back to the original full-frame image coordinate system through coordinate conversion, and output the coordinates of the center point and circumscribed rectangle of the first frame image after segmentation; S6. Based on the coordinates of the center point and the bounding rectangle of the first frame image, the position of the sub-image of the next frame image is automatically updated using a dynamic update strategy for the cropping region based on mask boundary determination, thereby generating the coordinates of the center point and the bounding rectangle of the next frame image. S7. Compare the pixel center positions of each target in the original image coordinate system in the first frame and the next frame, and compare them with the camera calibration parameters to obtain the multi-target deformation monitoring results.

2. The method for monitoring multi-target structural deformation based on visual segmentation and deblurring enhancement according to claim 1, characterized in that, The acquisition of continuous temporal multi-target images of different depth planes of the structure to be monitored in step S1 includes: Multiple circular targets are pre-positioned on the surface of the structure to be monitored, distributed in different depth planes. A fixed focal length and high resolution industrial camera is used to acquire images, obtaining continuous time-series multi-target images of the structure surface at different depth planes.

3. The method for monitoring multi-target structural deformation based on visual segmentation and deblurring enhancement according to claim 1, characterized in that, The S2 process employs pre-selected foreground cue points and connected component analysis for automatic noise removal, including: Based on pre-selected foreground cue points and connected component analysis, the method determines whether the selected points are located inside the connected component. If the pre-selected foreground cue points are located inside the connected component, they are determined to be valid regions; otherwise, they are determined to be noise and replaced with background pixels, thus obtaining a binary mask containing only the target.

4. The method for monitoring multi-target structural deformation based on visual segmentation and deblurring enhancement according to claim 1, characterized in that, S6 automatically updates the sub-image position of the next frame image based on the coordinates of the center point and the bounding rectangle of the first frame image, using a dynamic update strategy for the cropping region based on mask boundary determination. This includes: S61. Set the initial cropping box size and pixel center coordinates for the i-th target; S62. Traverse all foreground pixels in the i-th target mask, calculate the minimum distance from all foreground pixels to the clipping box boundary, determine whether the pixel meets the defined coverage condition, if not, perform the expansion mechanism, expand the boundary according to the set ratio and update the center coordinates until the coverage condition is met.

5. The method for monitoring multi-target structural deformation based on visual segmentation and deblurring enhancement according to claim 4, characterized in that, The coverage condition is expressed by the following formula (1): （1） in, This represents the x-coordinate of the j-th foreground pixel in the i-th target mask region; This indicates the coordinates of the left boundary of the current cropping box, which is the pixel position of the smallest x-axis on the left. This indicates the right boundary coordinates of the current cropping box, which is the pixel position of the maximum x-axis on the right. This represents the y-coordinate of the j-th foreground pixel in the i-th target mask region; This indicates the coordinates of the upper boundary of the current cropping box, which is the pixel position of the smallest y-value on the upper side; This indicates the coordinates of the lower boundary of the current cropping box, which is the pixel position of the maximum y-value at the bottom. This represents the distance threshold, used to determine whether the mask touches the clipping box boundary.

6. The method for monitoring multi-target structural deformation based on visual segmentation and deblurring enhancement according to claim 1, characterized in that, S7 compares the pixel center positions of each target in the original image coordinate system in the first frame and the next frame, and compares them in conjunction with camera calibration parameters to obtain multi-target deformation monitoring results, including: The pixel center positions of each target in the first frame and the next frame are compared in the original image coordinate system. The pixel-level displacement difference of each target center position in the next frame is calculated. Combined with the camera's internal and external parameters and structural dimensions, the pixel displacement difference is converted into physical unit displacement or three-dimensional coordinate change to obtain the multi-target deformation monitoring results.

7. A structural multi-target deformation monitoring device based on visual segmentation and deblurring enhancement, wherein the structural multi-target deformation monitoring device based on visual segmentation and deblurring enhancement is used to implement the structural multi-target deformation monitoring method based on visual segmentation and deblurring enhancement as described in any one of claims 1-6, characterized in that, The device includes: The acquisition unit is used to acquire continuous temporal multi-target images of different depth planes of the structure to be monitored; based on the pre-selected foreground cue points, the first frame image of the multi-target images is input into the pre-trained SAM segmentation model, and a preliminary multi-target segmentation mask is output. The processing unit is used to automatically remove noise based on the preliminary multi-target segmentation mask, using pre-selected foreground cue points and connected component analysis to obtain effective connected regions, and to number the effective connected regions. The clipping unit is used to extract the bounding rectangle of the effective connected region. It adopts an adaptive multi-scale clipping strategy based on boundary conditions to clip and obtain a multi-target target region sub-map. The enhancement unit is used to input the multi-target target region sub-map into the pre-trained NAFNet network model for enhancement processing to obtain the enhanced multi-target target region sub-map; The segmentation unit is used to extract the center point and bounding rectangle coordinates of the target mask region within the enhanced multi-target sub-image, and input the enhanced multi-target sub-image into the visual segmentation model for segmentation processing to obtain each target mask; calculate the center point coordinates of each target mask, and map the coordinates within the enhanced multi-target sub-image back to the original full-frame image coordinate system through coordinate transformation, and output the coordinates of the center point and bounding rectangle of the first frame image after segmentation; The update unit is used to automatically update the sub-image position of the next frame image based on the coordinates of the center point and the bounding rectangle of the first frame image, using a dynamic update strategy for the cropping region based on mask boundary determination, and to generate the coordinates of the center point and the bounding rectangle of the next frame image. The monitoring unit is used to compare the pixel center positions of each target in the original image coordinate system in the first frame and the next frame, and to compare them with the camera calibration parameters to obtain the multi-target deformation monitoring results.

8. A multi-target structural deformation monitoring device based on visual segmentation and deblurring enhancement, characterized in that, The structure multi-target deformation monitoring device based on visual segmentation and deblurring enhancement includes: processor; A memory storing computer-readable instructions that, when executed by the processor, implement the method as described in any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium contains program code that can be invoked by a processor to execute the method as described in any one of claims 1 to 6.

Images