Aerial work safety monitoring method and device and storage medium
By training object detection and semantic segmentation models, rectangular and polygonal boxes are detected in high-altitude operation scenarios, solving the problem of low recognition rate in existing technologies and realizing comprehensive and reliable monitoring of the status of high-altitude operations.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- GUANGDONG POWER GRID CO LTD
- Filing Date
- 2023-10-24
- Publication Date
- 2026-05-15
AI Technical Summary
Existing methods for monitoring safety during high-altitude operations are easily affected by objective factors such as sample database and lighting conditions, resulting in low recognition rates.
By collecting data on high-altitude operation scenarios, and after data annotation, target detection models and semantic segmentation models are trained to detect rectangular boxes and polygonal boxes respectively, and the high-altitude operation status is determined based on the number and degree of overlap of the boxes.
It effectively improves the accuracy and comprehensiveness of high-altitude operation identification, and can accurately monitor whether workers are wearing safety helmets, work clothes and safety belts, etc., reducing the impact of factors such as sample library and lighting.
Smart Images
Figure CN117253136B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of safety detection technology, and in particular to a method, device and storage medium for high-altitude operation safety monitoring. Background Technology
[0002] Substations are a crucial component of power grids, playing a vital role in voltage transformation, current collection, and energy distribution. Therefore, proper maintenance of substations ensures stable and safe operation. Substation transformer and tower maintenance inevitably involve working at heights. Properly worn safety harnesses and effective safety measures must be implemented during these operations.
[0003] Existing methods for monitoring safety at heights typically employ database comparison, which involves comparing real-time photos with standard photos of high-altitude operations in a database to determine whether such activities have occurred. However, this database comparison method is susceptible to the influence of objective factors such as the sample library and lighting conditions on the recognition rate, resulting in a low effectiveness in monitoring high-altitude operations. Summary of the Invention
[0004] This invention provides a method, device, and storage medium for monitoring safety during high-altitude operations, in order to address the problem that some high-altitude operation safety monitoring methods are easily affected by objective factors such as sample databases and lighting, resulting in low recognition rates and low monitoring effectiveness for high-altitude operations.
[0005] This invention provides a method for monitoring safety during high-altitude operations, comprising:
[0006] Collect data on high-altitude work scenarios;
[0007] The high-altitude operation scene data is annotated to obtain rectangular box annotation data and polygon annotation data;
[0008] The high-altitude operation scene data and the rectangular bounding box annotation data are input into a preset initial target detection model for training to obtain the target detection model;
[0009] The high-altitude operation scene data and the polygon annotation data are input into a preset initial semantic segmentation model for training to obtain a semantic segmentation model;
[0010] The image of the operation to be monitored is input into the target detection model to obtain a rectangular bounding box, and the image of the operation to be monitored is input into the semantic segmentation model to obtain a polygonal bounding box; the rectangular bounding box includes the area where the worker is located, the safety helmet area, and the work clothes area; the polygonal bounding box includes the horizontal structure area, the vertical structure area, and the safety belt area;
[0011] The high-altitude operation status is determined based on the number and overlap of the rectangular and polygonal frames.
[0012] Furthermore, the methods for collecting high-altitude operation scene data include at least one of monitoring and shooting, mobile phone shooting, web crawling, and manual simulation.
[0013] Furthermore, the step of annotating the high-altitude operation scene data to obtain rectangular bounding box annotation data and polygon annotation data includes:
[0014] Based on a preset target detection standard, the high-altitude operation scene data is labeled to obtain rectangular bounding box labeled data; based on a preset semantic segmentation standard, the high-altitude operation scene data is labeled to obtain polygon labeled data.
[0015] Furthermore, determining the high-altitude operation status based on the number and overlap data of the rectangular and polygonal frames includes:
[0016] If the number of areas where the operator is located is greater than 0, and the ordinate of the center point of the area where the operator is located is less than the coordinate of the upper left point of the vertical structure area, then it is determined that there is a high-altitude operation.
[0017] Furthermore, determining the high-altitude operation status based on the number and overlap data of the rectangular and polygonal frames also includes:
[0018] After determining that there is a high-altitude operation, if the number of safety helmet areas detected is greater than 0, and the overlap value between the area where the worker is located and the safety helmet area is greater than a first preset threshold, then it is determined that the worker is wearing a safety helmet; otherwise, a safety helmet wearing alarm is initiated.
[0019] Furthermore, determining the high-altitude operation status based on the number and overlap data of the rectangular and polygonal frames also includes:
[0020] After determining that there is a high-altitude operation, if the number of the work clothes area is greater than 0 and the overlap value between the worker and the work clothes area is higher than the second preset threshold, it is determined that the worker is wearing work clothes; otherwise, a work clothes wearing alarm is initiated.
[0021] Furthermore, determining the high-altitude operation status based on the number and overlap data of the rectangular and polygonal frames also includes:
[0022] After determining that there is a high-altitude operation, if the number of safety belt areas is greater than 0, and the overlap value between the area where the worker is located and the safety belt area is greater than a third preset threshold, then it is determined that the worker is wearing a safety belt; otherwise, a safety belt wearing alarm is initiated.
[0023] Furthermore, determining the high-altitude operation status based on the number and overlap data of the rectangular and polygonal frames also includes:
[0024] After determining that there is a high-altitude operation, if the number of safety belt areas is greater than 0 and the overlap value between the longitudinal structure area and the safety belt area is greater than the fourth preset threshold, it is determined that the current worker has fixed the safety belt to the longitudinal structure; otherwise, a safety belt fixing alarm is initiated.
[0025] Scene data acquisition module, used to collect high-altitude operation scene data;
[0026] The data annotation module is used to annotate the high-altitude operation scene data to obtain rectangular box annotation data and polygon annotation data;
[0027] The target detection model training module is used to input the high-altitude operation scene data and the rectangular box annotation data into a preset initial target detection model for training, so as to obtain the target detection model;
[0028] The semantic segmentation model training module is used to input the high-altitude operation scene data and the polygon annotation data into a preset initial semantic segmentation model for training, so as to obtain a semantic segmentation model.
[0029] The region detection module is used to input the image of the operation to be monitored into the target detection model to obtain a rectangular bounding box, and input the image of the operation to be monitored into the semantic segmentation model to obtain a polygonal bounding box; the rectangular bounding box includes the area where the worker is located, the safety helmet area, and the work clothes area; the polygonal bounding box includes the horizontal structure area, the vertical structure area, and the safety belt area;
[0030] The work status recognition module is used to determine the high-altitude work status based on the number and overlap of the rectangular and polygonal frames.
[0031] The present invention also provides a storage medium comprising a stored computer program, wherein, when the computer program is executed, it controls the device containing the storage medium to perform a high-altitude operation safety monitoring method as described above.
[0032] This invention provides a target detection model and a semantic segmentation model through model training. These models detect rectangular and polygonal boxes in the image of the operation to be monitored, and determine the status of the high-altitude operation based on the number and overlap of the rectangular and polygonal boxes. The recognition rate is not affected by objective factors such as sample library and lighting, thus effectively improving the recognition effect of high-altitude operations.
[0033] Furthermore, in this embodiment of the invention, the status of high-altitude operations is determined based on the number and overlap of the rectangular and polygonal frames, including whether there is high-altitude work activity, whether a safety helmet is worn, whether work clothes are worn, whether a safety belt is worn, and whether the safety belt is secured. This comprehensively monitors various conditions of high-altitude operations, effectively improving the comprehensiveness and reliability of high-altitude operation monitoring. Attached Figure Description
[0034] Figure 1 This is a flowchart illustrating a high-altitude operation safety monitoring method provided in an embodiment of the present invention;
[0035] Figure 2 This is a schematic diagram of data annotation provided in an embodiment of the present invention;
[0036] Figure 3 This is a schematic diagram of a rectangular frame provided in an embodiment of the present invention;
[0037] Figure 4 This is a schematic diagram of a polygonal frame provided in an embodiment of the present invention;
[0038] Figure 5 This is a flowchart illustrating a high-altitude operation safety monitoring method provided in an embodiment of the present invention. Detailed Implementation
[0039] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.
[0040] In the description of this application, it should be understood that the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this application, unless otherwise stated, "multiple" means two or more.
[0041] In the description of this application, it should be noted that, unless otherwise expressly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection between two components. Those skilled in the art can understand the specific meaning of the above terms in this application based on the specific circumstances.
[0042] Please see Figure 1 The present invention provides a method for monitoring the safety of high-altitude operations, comprising:
[0043] S1. Collect data on high-altitude operation scenarios;
[0044] In this embodiment of the invention, the methods for collecting high-altitude operation scene data include at least one of monitoring and shooting, mobile phone shooting, web crawling and artificial simulation.
[0045] S2. Label the high-altitude operation scene data to obtain rectangular box annotation data and polygon annotation data;
[0046] Please see Figure 2 In this embodiment of the invention, the rectangular box annotation data includes annotation data such as personnel, wearing a safety helmet, not wearing a safety helmet, wearing work clothes, and not wearing work clothes; the polygon annotation data includes annotation data such as safety belt, horizontal structure, and vertical structure.
[0047] S3. Input the high-altitude operation scene data and the rectangular box annotation data into the preset initial target detection model for training to obtain the target detection model;
[0048] In this invention, the preset initial object detection model can be a YOLOv7-based object detection model. The object detection model includes one of the following: Faster R-CNN, YOLOv5, Ppyoloe, etc.
[0049] S4. Input the high-altitude operation scene data and polygon annotation data into the preset initial semantic segmentation model for training to obtain the semantic segmentation model;
[0050] In this embodiment of the invention, the initial semantic segmentation model can be a semantic segmentation model based on HRNet. Semantic segmentation models include one of the following: OCRNet, SegNet, Deeplabv3, etc.
[0051] In this embodiment of the invention, in order to improve the robustness of the algorithm, data augmentation processing is performed on the training data, including image blurring, image rotation, CLAHE, Mosaic and affine transformations.
[0052] S5. Input the image of the operation to be monitored into the target detection model to obtain a rectangular box, and input the image of the operation to be monitored into the semantic segmentation model to obtain a polygonal box; the rectangular box includes the area where the worker is located, the safety helmet area, and the work clothes area; the polygonal box includes the horizontal structure area, the vertical structure area, and the safety belt area;
[0053] In this embodiment of the invention, the area where the worker is located is Rect_person, the safety helmet area is Rect_helment, and the work clothes area is Rect_uniforms; the polygonal frame includes a horizontal structure area called Mask_transerse, a vertical structure area called Mask_longitudinal, and a safety belt area called Mask_belt.
[0054] Please see Figure 3 This is a schematic diagram of a rectangular frame provided in an embodiment of the present invention.
[0055] Please see Figure 4 This is a schematic diagram of a polygonal frame provided in an embodiment of the present invention.
[0056] S6. Determine the high-altitude operation status based on the number and overlap of rectangular and polygonal frames.
[0057] In this embodiment of the invention, the intersection-union ratio (IOU) is used as the overlap data to represent the degree of overlap between boxes. The greater the degree of overlap, the larger the IOU value.
[0058] In this embodiment of the invention, the intersection-union ratio (Rect_IOU) of the rectangles is used to measure the degree of overlap between two rectangles. The greater the overlap, the larger the IOU value. For example, assuming the two rectangles are Rect1 and Rect2, the coordinates of the top-left corner and bottom-right corner of Rect1 are [x1, y1] and [x2, y2] respectively, and the coordinates of the top-left corner and bottom-right corner of Rect2 are [x3, y3] and [x3, y4] respectively, the formula for calculating Rect_IOU is as follows:
[0059] x inter1 =max(x1,x3)
[0060] y inter1 =max(y1,y3)
[0061] x inter2 =max(x2,x4)
[0062] y inter2 =max(y2,y4)
[0063] width inter =abs(x inter2 -x inter1 )
[0064] height inter =abs(y inter2 -y inter1 )
[0065] area inter =widthinter *height inter
[0066] width box1 =abs(x2-x1)
[0067] height box1 =abs(y2-y1)
[0068] width box2 =abs(x4-x3)
[0069] height box2 =abs(y4-y3)
[0070] area box1 =width box1 *height box1
[0071] area box2 =width box2 *height box2
[0072] Rect_IOU = area inter / (area box1 +area box2 -area inter
[0073] Among them, the Max function is a function that outputs the maximum value of the parameter value, such as max(y1,y3) which outputs the maximum value of y1 and y3; the Abs function is a function that outputs the absolute value of the parameter value.
[0074] In this embodiment of the invention, the intersection-union ratio (Mask_IOU) between a rectangular bounding box and a polygonal bounding box is used to measure the degree of overlap between them. The greater the overlap, the larger the Mask_IOU value. For example, assuming the rectangular bounding box and the polygonal bounding box are Rect and Polygon respectively, when traversing the image, if a pixel is not only within the Rect range but also within the Polygon range, then num... inter Increment the count by 1. If the pixel is within the Rect range but not within the Polygon range, or if the pixel is within the Polygon range but not within the Rect range, then increment the count by 1. area Increment the count by 1. After the traversal is complete, then...
[0075] Mask_IOU = num_inter / (num inter +num area ).
[0076] The embodiments of the present invention obtain target detection model and semantic segmentation model through model training, respectively detect rectangular boxes and polygonal boxes of the image to be monitored, and determine the high-altitude operation status based on the number and overlap of rectangular boxes and polygonal boxes. The recognition rate is not affected by objective factors such as sample library and lighting, thereby effectively improving the recognition effect of high-altitude operations.
[0077] In one embodiment, step S2, labeling the high-altitude operation scene data to obtain rectangular bounding box labeling data and polygon labeling data, includes:
[0078] Based on preset target detection standards, high-altitude operation scene data is labeled to obtain rectangular bounding box labeled data; based on preset semantic segmentation standards, high-altitude operation scene data is labeled to obtain polygon labeled data.
[0079] In one embodiment, step S6, determining the high-altitude operation status based on the number and overlap data of rectangular and polygonal frames, further includes:
[0080] S61. If the number of Rect_persons in the area where the worker is located is greater than 0, and the ordinate of the center point of Rect_person in the area where the worker is located is less than the coordinate of the upper left point of the vertical structure area, then it is determined that there is a high-altitude operation.
[0081] In one embodiment, step S6, determining the high-altitude operation status based on the number and overlap data of rectangular and polygonal frames, further includes:
[0082] S62. After determining that there is a high-altitude operation, if the number of safety helmet areas detected is greater than 0, and the overlap value between the area where the worker is located and the safety helmet area is greater than the first preset threshold, then it is determined that the worker is currently wearing a safety helmet; otherwise, a safety helmet wearing alarm is initiated.
[0083] In this embodiment of the invention, preset thresholds can be set and adjusted according to actual needs.
[0084] This invention determines whether a safety helmet is present at the work site based on the number of safety helmet areas. Furthermore, it can further determine whether high-altitude work is taking place based on the overlap value between the worker's area and the safety helmet area. This method is unaffected by objective factors such as sample databases and lighting conditions, thus effectively obtaining monitoring results for high-altitude work.
[0085] In one embodiment, step S6, determining the high-altitude operation status based on the number and overlap data of rectangular and polygonal frames, further includes:
[0086] S63. After determining that there is a high-altitude operation, if the number of work clothes areas detected is greater than 0, and the overlap value between the worker and the work clothes area is higher than the second preset threshold, then it is determined that the worker is currently wearing work clothes; otherwise, a work clothes wearing alarm is initiated.
[0087] In one embodiment, step S6, determining the high-altitude operation status based on the number and overlap data of rectangular and polygonal frames, further includes:
[0088] S64. After determining that there is a high-altitude operation, if the number of safety belt areas is greater than 0 and the overlap value between the area where the worker is located and the safety belt area is greater than the third preset threshold, then it is determined that the worker is wearing a safety belt; otherwise, a safety belt wearing alarm is initiated.
[0089] In one embodiment, step S6, determining the high-altitude operation status based on the number and overlap data of rectangular and polygonal frames, further includes:
[0090] S65. After determining that there is a high-altitude operation, if the number of safety belt areas is greater than 0 and the overlap value between the longitudinal structure area and the safety belt area is greater than the fourth preset threshold, it is determined that the current worker has fixed the safety belt to the longitudinal structure; otherwise, a safety belt fixing alarm is initiated.
[0091] Implementing the embodiments of the present invention has the following beneficial effects:
[0092] The embodiments of the present invention obtain target detection model and semantic segmentation model through model training, respectively detect rectangular boxes and polygonal boxes of the image to be monitored, and determine the high-altitude operation status based on the number and overlap of rectangular boxes and polygonal boxes. The recognition rate is not affected by objective factors such as sample library and lighting, thereby effectively improving the recognition effect of high-altitude operations.
[0093] Furthermore, in this embodiment of the invention, the status of high-altitude operations is determined based on the number and overlap of rectangular and polygonal frames, including whether high-altitude work is being carried out, whether safety helmets, work clothes, safety belts, and whether safety belts are being secured. This comprehensively monitors various conditions of high-altitude operations, effectively improving the comprehensiveness and reliability of high-altitude operation monitoring.
[0094] Please see Figure 5 Based on the same inventive concept as the above embodiments, the present invention also provides a high-altitude operation safety monitoring device, comprising:
[0095] Scene data acquisition module 10 is used to collect high-altitude operation scene data;
[0096] Data annotation module 20 is used to annotate high-altitude operation scene data to obtain rectangular box annotation data and polygon annotation data;
[0097] The target detection model training module 30 is used to input high-altitude operation scene data and rectangular box annotation data into the preset initial target detection model for training, so as to obtain the target detection model;
[0098] The semantic segmentation model training module 40 is used to input high-altitude operation scene data and polygon annotation data into the preset initial semantic segmentation model for training to obtain the semantic segmentation model;
[0099] The region detection module 50 is used to input the image of the operation to be monitored into the target detection model to obtain a rectangular box, and input the image of the operation to be monitored into the semantic segmentation model to obtain a polygonal box; the rectangular box includes the area where the worker is located, the safety helmet area, and the work clothes area; the polygonal box includes the horizontal structure area, the vertical structure area, and the safety belt area;
[0100] The work status recognition module 60 is used to determine the high-altitude work status based on the number and overlap of rectangular and polygonal boxes.
[0101] In one embodiment, the methods for collecting data on high-altitude work scenarios include at least one of surveillance video recording, mobile phone photography, web crawling, and artificial simulation.
[0102] In one embodiment, the data annotation module 20 is further configured to:
[0103] Based on preset target detection standards, high-altitude operation scene data is labeled to obtain rectangular bounding box labeled data; based on preset semantic segmentation standards, high-altitude operation scene data is labeled to obtain polygon labeled data.
[0104] In one embodiment, the job status identification module 60 is further configured to:
[0105] If the number of areas where workers are located is greater than 0, and the vertical coordinate of the center point of the area where workers are located is less than the coordinate of the top left point of the vertical structure area, then it is determined that there is currently a high-altitude operation.
[0106] In one embodiment, the job status identification module 60 is further configured to:
[0107] After determining that there is a high-altitude operation, if the number of safety helmet areas detected is greater than 0, and the overlap between the area where the worker is located and the safety helmet area is greater than the first preset threshold, then it is determined that the worker is wearing a safety helmet; otherwise, a safety helmet wearing alarm is initiated.
[0108] In one embodiment, the job status identification module 60 is further configured to:
[0109] After determining that there is a high-altitude operation, if the number of work clothes areas detected is greater than 0 and the overlap value between the worker and the work clothes area is higher than the second preset threshold, it is determined that the worker is wearing work clothes; otherwise, a work clothes wearing alarm is initiated.
[0110] In one embodiment, the job status identification module 60 is further configured to:
[0111] After determining that there is a high-altitude operation, if the number of safety belt areas is greater than 0 and the overlap between the area where the worker is located and the safety belt area is greater than the third preset threshold, then it is determined that the worker is wearing a safety belt; otherwise, a safety belt wearing alarm is initiated.
[0112] In one embodiment, the job status identification module 60 is further configured to:
[0113] After determining that there is a high-altitude operation, if the number of safety belt areas is greater than 0 and the overlap value between the longitudinal structure area and the safety belt area is greater than the fourth preset threshold, it is determined that the current worker has fixed the safety belt to the longitudinal structure; otherwise, a safety belt fixing alarm is initiated.
[0114] The present invention also provides a storage medium, the storage medium including a stored computer program, wherein, when the computer program is running, it controls the device where the storage medium is located to execute a high-altitude operation safety monitoring method as described above.
[0115] The above are preferred embodiments of the present invention. It should be noted that, for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications are also considered to be within the scope of protection of the present invention.
Claims
1. A method for monitoring safety during high-altitude operations, characterized in that, include: Collect data on high-altitude work scenarios; The high-altitude operation scene data is annotated to obtain rectangular box annotation data and polygon annotation data; The high-altitude operation scene data and the rectangular bounding box annotation data are input into a preset initial target detection model for training to obtain the target detection model; The high-altitude operation scene data and the polygon annotation data are input into a preset initial semantic segmentation model for training to obtain a semantic segmentation model; The image of the operation to be monitored is input into the target detection model to obtain a rectangular bounding box, and the image of the operation to be monitored is input into the semantic segmentation model to obtain a polygonal bounding box; the rectangular bounding box includes the area where the worker is located, the safety helmet area, and the work clothes area; the polygonal bounding box includes the horizontal structure area, the vertical structure area, and the safety belt area; The high-altitude operation status is determined based on the number and overlap of the rectangular and polygonal frames; wherein, the determination of the high-altitude operation status based on the number and overlap of the rectangular and polygonal frames includes: if the number of areas where the operator is located is greater than 0, and the ordinate of the center point of the area where the operator is located is less than the coordinate of the upper left point of the vertical structure area, then it is determined that there is a high-altitude operation. Specifically, after determining that there is a high-altitude operation, if the number of safety belt areas is greater than 0, and the overlap between the area where the worker is located and the safety belt area is greater than a third preset threshold, then it is determined that the worker is wearing a safety belt; otherwise, a safety belt wearing alarm is initiated. After determining that there is a high-altitude operation, if the number of safety belt areas is greater than 0, and the overlap between the longitudinal structure area and the safety belt area is greater than a fourth preset threshold, then it is determined that the worker has secured the safety belt to the longitudinal structure; otherwise, a safety belt securing alarm is initiated.
2. The high-altitude operation safety monitoring method as described in claim 1, characterized in that, The methods for collecting high-altitude operation scene data include at least one of the following: surveillance video, mobile phone video, web crawler, and manual simulation.
3. The high-altitude operation safety monitoring method as described in claim 1, characterized in that, The step of annotating the high-altitude operation scene data to obtain rectangular bounding box annotation data and polygon annotation data includes: The high-altitude operation scene data is labeled with rectangular bounding boxes based on a preset target detection standard; the high-altitude operation scene data is labeled with polygons based on a preset semantic segmentation standard.
4. The high-altitude operation safety monitoring method as described in claim 1, characterized in that, The step of determining the high-altitude operation status based on the number and overlap data of the rectangular and polygonal frames also includes: After determining that there is a high-altitude operation, if the number of safety helmet areas detected is greater than 0, and the overlap value between the area where the worker is located and the safety helmet area is greater than a first preset threshold, then it is determined that the worker is wearing a safety helmet; otherwise, a safety helmet wearing alarm is initiated.
5. The high-altitude operation safety monitoring method as described in claim 1, characterized in that, The step of determining the high-altitude operation status based on the number and overlap data of the rectangular and polygonal frames also includes: After determining that there is a high-altitude operation, if the number of the work clothes area is greater than 0 and the overlap value between the worker and the work clothes area is higher than the second preset threshold, it is determined that the worker is wearing work clothes; otherwise, a work clothes wearing alarm is initiated.
6. A high-altitude work safety monitoring device, characterized in that, include: Scene data acquisition module, used to collect high-altitude operation scene data; The data annotation module is used to annotate the high-altitude operation scene data to obtain rectangular box annotation data and polygon annotation data; The target detection model training module is used to input the high-altitude operation scene data and the rectangular box annotation data into a preset initial target detection model for training, so as to obtain the target detection model; The semantic segmentation model training module is used to input the high-altitude operation scene data and the polygon annotation data into a preset initial semantic segmentation model for training, so as to obtain a semantic segmentation model. The region detection module is used to input the image of the operation to be monitored into the target detection model to obtain a rectangular bounding box, and input the image of the operation to be monitored into the semantic segmentation model to obtain a polygonal bounding box; the rectangular bounding box includes the area where the worker is located, the safety helmet area, and the work clothes area; the polygonal bounding box includes the horizontal structure area, the vertical structure area, and the safety belt area; The work status recognition module is used to determine the high-altitude work status based on the number and overlap of the rectangular boxes and the polygonal boxes; wherein, determining the high-altitude work status based on the number and overlap of the rectangular boxes and the polygonal boxes includes: if the number of areas where the worker is located is greater than 0, and the ordinate of the center point of the area where the worker is located is less than the coordinate of the upper left point of the vertical structure area, then it is determined that there is a high-altitude work activity. Specifically, after determining that there is a high-altitude operation, if the number of safety belt areas is greater than 0, and the overlap between the area where the worker is located and the safety belt area is greater than a third preset threshold, then it is determined that the worker is wearing a safety belt; otherwise, a safety belt wearing alarm is initiated. After determining that there is a high-altitude operation, if the number of safety belt areas is greater than 0, and the overlap between the longitudinal structure area and the safety belt area is greater than a fourth preset threshold, then it is determined that the worker has secured the safety belt to the longitudinal structure; otherwise, a safety belt securing alarm is initiated.
7. A storage medium, characterized in that, The storage medium includes a stored computer program, wherein, when the computer program is executed, it controls the device where the storage medium is located to perform a high-altitude operation safety monitoring method as described in any one of claims 1 to 5.