Centralized tracking and monitoring method for electric power high-altitude constructors

Abstract

The invention discloses an electric power high-altitude constructor centering tracking supervision method, and relates to the technical field of electric power construction, the method adopts a supervision system to work, the supervision system comprises an intelligent monitoring platform and a ground monitoring AI device, the intelligent monitoring platform provides streaming media service, multiple algorithm identification models and data service, and the AI device is connected with the intelligent monitoring platform. A user accesses through a webpage and a mobile phone client, the ground monitoring AI equipment comprises a camera and is horizontally arranged at a construction site by an operator, and the ground monitoring AI equipment comprises a video acquisition module, an image recognition module, a centering algorithm module, a linkage algorithm module and a PTZ control module. The video acquisition module is responsible for acquiring a real-time video stream of a high-altitude operation site, ensuring that the video quality is clear and providing a high-quality data source for subsequent image analysis, and the image recognition module receives the video stream and preprocesses a video frame by using an image processing algorithm.

Description

Technical Field

[0001] This invention relates to the field of power construction technology, specifically a method for centrally tracking and supervising personnel during high-altitude power construction. Background Technology

[0002] With the continuous development of power infrastructure construction, high-altitude power construction operations are increasing. These operations are characterized by complex environments and high risks, placing extremely high demands on the safety and operational standards of construction personnel. Traditional methods of supervising power construction sites mainly rely on manual inspections and limited video surveillance, which suffer from limited coverage, insufficient real-time performance, and difficulty in accurately locating and analyzing the behavior of high-altitude workers. Especially in high-altitude operations, worker displacement or improper operation can lead to serious consequences.

[0003] Currently, while some video analytics-based construction site monitoring systems exist, these systems often focus on detecting abnormal events within a fixed range, such as intrusion detection and fall detection. They remain significantly inadequate for the refined management of workers at heights, particularly in terms of centered tracking and supervision. Existing technologies struggle to accurately identify and continuously track workers at heights, effectively determine whether they are within a pre-set safe zone, and cannot promptly track deviations to ensure they remain in the center. For example, existing technologies, such as (CN119229384B), are primarily used for training and assessment in high-altitude operations, operating in a relatively controlled environment that differs significantly from the dynamic, complex, and unpredictable nature of actual high-altitude work sites. (CN116992080A) focuses on offline video processing and retrieval efficiency rather than real-time proactive tracking and intervention of workers. Furthermore, existing technologies struggle to effectively center and track multiple workers working collaboratively at heights, thus failing to comprehensively guarantee safety and improve management efficiency. Therefore, there is an urgent need for a technical means to accurately, in real time, and automatically track and monitor workers working at heights. Summary of the Invention

[0004] The purpose of this invention is to provide a method for centrally tracking and supervising personnel working at heights in power plants, in order to solve the problems mentioned in the background art.

[0005] To address the aforementioned technical problems, this invention provides the following technical solution: a method for centrally tracking and supervising personnel during high-altitude power construction. This method employs a monitoring system, which includes an intelligent monitoring platform and ground-based AI monitoring equipment. The intelligent monitoring platform provides streaming media services, various algorithm recognition models, and data services, accessible to users via web pages and mobile clients. The ground-based AI monitoring equipment includes cameras, which are horizontally positioned at the construction site by the workers.

[0006] According to the above technical solution, the ground monitoring AI equipment includes a video acquisition module, an image recognition module, a centering algorithm module, a linkage algorithm module, and a PTZ control module. The video acquisition module is responsible for acquiring real-time video streams from the high-altitude work site, ensuring clear video quality and providing a high-quality data source for subsequent image analysis. The image recognition module receives the video stream, preprocesses the video frames using image processing algorithms, including noise reduction and enhancement, and uses deep learning and other technologies to achieve accurate detection and recognition of high-altitude workers. The centering algorithm module calculates the coordinates of the center point of the worker and the combined area of ​​multiple workers in real time, and determines whether the center point is within a preset safety threshold range. When the center point of the worker deviates from the safety threshold, the linkage algorithm module, combined with the worker's coordinates on the image and the current parameters of the PTZ control module, calculates the parameters that the PTZ control module needs to correct using the linkage control algorithm. The PTZ control module, based on the parameters that the PTZ control module needs to adjust calculated by the linkage control algorithm, achieves dynamic PTZ correction, bringing the worker back to the center of the image.

[0007] According to the above technical solution, the following work steps are included:

[0008] S1. Place the ground monitoring AI equipment horizontally, start the system, and the ground monitoring AI equipment automatically adjusts the video monitoring angle to a 30-degree angle with the horizontal plane. The video acquisition module starts to acquire real-time video streams from the high-altitude operation site, and the image recognition module initializes the algorithm model to prepare for human body frame detection.

[0009] S2. Perform human body recognition to determine whether there are people in the current monitoring area. If no one is detected in the monitoring area three times in a row, execute S8 after 30 seconds of sleep. Otherwise, enter the high-altitude area personnel monitoring.

[0010] S3. Human bounding box filtering and optimization: After detecting a person in the supervised area, if multiple people are detected, further filtering and optimization are performed to improve the accuracy of tracking.

[0011] S4. Calculate the coordinates of the center point of the merged area of ​​all high-altitude personnel, calculate the merged area of ​​the human body frames of all identified high-altitude workers, and determine the coordinates of the center point of the merged area. Based on the convex hull algorithm and clustering algorithm, determine a merged area that can represent the activity range of the entire work group.

[0012] S5. Determine if the center point coordinates are within the threshold range;

[0013] S6. Monitoring area perspective adjustment and continuous deviation judgment;

[0014] S7. Judgment of overdue monitoring in high-altitude areas;

[0015] S8. Adjust the monitoring angle and enter the next high-altitude monitoring area.

[0016] According to the above technical solution, the specific filtering rules in S3 include:

[0017] S3-1, Center Point Coordinate Filtering: Filters out human body frames whose center point is lower than a specific coordinate in the image (e.g., image coordinate 800 for a 1920*1080 image). This specific coordinate is determined based on a large amount of data analysis of high-altitude power construction site images, combined with the typical height of high-altitude workers and the image ratio characteristics, and can effectively distinguish high-altitude workers from non-work targets on the ground.

[0018] S3-2, Aspect Ratio Filtering: Filters out human bounding boxes whose height-to-width ratio is less than a specific threshold. This threshold can accurately match the typical morphological features of high-altitude workers and effectively eliminate non-human-shaped misidentifications (such as poles, power lines, tools) caused by occlusion, lighting changes, or background complexity, thereby improving the robustness of high-altitude worker detection. These filtering parameters are adaptive or adjustable and can be dynamically optimized according to different high-altitude working environments (such as pole type, working height, and lighting conditions), further enhancing the system's adaptability in complex environments.

[0019] According to the above technical solution, in S4, the convex hull algorithm is specifically as follows: by calculating the convex hull of the feature points of the human body frame, a minimum convex polygon can be obtained. This polygon can effectively enclose all high-altitude workers, thereby forming a merged region representing their collective activity range. The center point of the merged region is taken as its geometric center or centroid, as follows:

[0020] S4-1. Given a set of N human bounding boxes: B = {B1, B2, ..., B} N}, each bounding box B i From its upper left corner coordinate (x i ,y i ) and width and height (w i ,h i Definition: B i ={(x i ,y i ),w i ,h i};

[0021] S4-2, Each bounding box B i The set of four corner points V i :

[0022] V i =Corners(B i )

[0023] ={(x i ,y i ),(x i +w i ,y i ),(x i +w i ,y i +h i ),(x i ,y i +h i )}

[0024] Union of the corner points of all bounding boxes:

[0025]

[0026] S4-3. A convex hull is defined as the smallest convex set containing all points:

[0027]

[0028] Let the sequence of convex hull vertices be:

[0029] H = {h1,h2,…,h} m}, h j =(x hj ,y hj )

[0030] S4-4, Centroid coordinates of convex hull vertices:

[0031]

[0032] S4-5, Finally, we can obtain a tuple:

[0033] (x c ,y c ,H)

[0034] Where (x) c ,y c ) represents the coordinates of the merged center point, and H represents the sequence of convex hull vertices.

[0035] According to the above technical solution, in S4, the clustering algorithm specifically involves: discovering clusters of arbitrary shapes based on point density and identifying noise points. Core parameters include the neighborhood radius and the minimum number of points required to form a dense region. Based on extensive data analysis of high-altitude power construction site images and combined with typical population distributions of various high-altitude workers, the algorithm optimizes and dynamically adapts these core parameters. This method is particularly suitable for handling non-uniformly distributed groups of workers in high-altitude operations. When multiple high-altitude workers exist, the clustering algorithm can group spatially close bounding boxes that may belong to the same work group, automatically identifying different work groups, as detailed below:

[0036] S4-6. Given a set of N human bounding boxes:

[0037] B = {B1, B2, ..., B} N}

[0038] Each bounding box B i From its upper left corner coordinate (x i ,y i ) and width and height (w i ,h i )definition:

[0039] B i ={(x i ,y i ),w i ,h i}

[0040] Each bounding box B i The center point is:

[0041]

[0042] The set of all center points is:

[0043] P = {C1, C2, ..., C} N}

[0044] S4-7, DBSCAN clustering algorithm core parameters: ε: neighborhood radius, minPts: minimum number of neighboring points. For each point p∈P, its ε-neighborhood is defined as:

[0045] N ε (p)={q∈P∣|pq|≤ε}

[0046] If |N ε If (p)|≥minPts, then p is the core point. Through recursive expansion of the neighborhood of the core point, clusters are formed, and finally P is divided into K clusters and a set of noise points N:

[0047]

[0048] Where P k Let be the set of points in the k-th cluster;

[0049] S4-8. Find the convex hull and centroid of each cluster:

[0050] For each cluster P k Extract all corner points of its corresponding bounding box, denoted as Q. k :

[0051]

[0052] Q k Calculate the convex hull H k :

[0053] H k =ConvexHull(Q k )

[0054] The vertices of the convex hull are

[0055] The centroid of the convex hull of cluster k is:

[0056]

[0057] S4-9, The overall center of a multi-cluster:

[0058] If multiple clusters exist, calculate the centroid center of all clusters:

[0059]

[0060] S4-10, Special Cases:

[0061] If there is no effective clustering (all data is noise), then use the mean of all center points directly:

[0062]

[0063] Ultimately, we can obtain (x) c ,y c ), which is the center point after the merger.

[0064] According to the above technical solution, in step S5, the calculated coordinates of the operator and the center point of the merged area are compared with the preset center point threshold range. The preset center point threshold defines the center safety area of ​​the high-altitude area. For example, for a 1920*1080 image, the threshold range in the horizontal direction is set to 460~1460, and the threshold range in the vertical direction can be set to 280~800. If the center point coordinates are not within the threshold range, step 6 is executed; otherwise, it is considered that the operator is in a safe centered state, and step 7 is executed.

[0065] In step S6, if the center point coordinates are not within the threshold range (i.e., the worker deviates from the center), the system will record the deviation status and determine whether it is outside the range three times (or more) consecutively. This continuous judgment mechanism is optimized to effectively filter out misjudgments and system jitter caused by momentary occlusion, image noise, or brief non-critical movement of the worker, thereby significantly improving the stability, smoothness, and efficiency of PTZ control, avoiding frequent and unnecessary adjustments, and reducing system energy consumption. If there is a continuous deviation, the centering algorithm module, the linkage algorithm module, and the PTZ control module will perform a monitoring area perspective adjustment operation to bring the worker back to the center of the screen. The linkage control algorithm accurately calculates the translation, pitch, and zoom parameters of the PTZ based on the worker's real-time position, the current parameters of the PTZ control module, and the predicted movement trend, ensuring that the worker (or merged area) can return to the center of the screen smoothly, quickly, and accurately, and adapt to the non-linear movement mode of high-altitude workers.

[0066] According to the above technical solution, in step S7, the high-altitude area is continuously monitored, and it is determined whether the preset monitoring timeout time has been reached, such as 10 minutes. If the timeout is exceeded, step 8 is entered; otherwise, step 2 is entered to continuously monitor the current high-altitude area.

[0067] In step S8, the system rotates 60 degrees to the right to enter the next high-altitude monitoring area and restarts from step 2. This is a resource-efficient coverage solution that replaces the high cost and complex deployment of multi-camera systems with the periodic, segmented scanning of a single ground AI device. The 60-degree rotation is optimized based on the camera's field of view and the typical range of high-altitude work areas, ensuring effective connection and seamless coverage between adjacent monitoring areas. The cyclic scanning mechanism works in conjunction with the high-altitude area monitoring timeout judgment in step 7 and the human body recognition in step 2 to ensure that the system can respond quickly and refocus when anomalies occur in non-directly monitored areas. When a specific anomaly occurs in a non-currently monitored area (such as an audible alarm or external sensor trigger), the current scan is immediately interrupted and priority is given to the abnormal area, thereby ensuring real-time response capability under 360-degree full coverage.

[0068] Compared with existing technologies, the beneficial effects achieved by this invention are: This invention enables precise centering tracking and real-time position determination of personnel working at heights, and can be expanded to multi-target management, thereby significantly improving the safety and management efficiency of high-altitude operations. The core of this invention lies in using machine vision technology to perform real-time analysis of high-altitude work scenarios, identify and track personnel, calculate their center point in the image, and adjust the monitoring perspective when personnel deviate from the center to ensure that personnel are always within the effective monitoring range. Attached Figure Description

[0069] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0070] Figure 1 This is a schematic diagram of the overall modular structure of the present invention;

[0071] Figure 2 This is a schematic diagram of the workflow of the present invention. Detailed Implementation

[0072] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0073] Please see Figure 1 and Figure 2 The present invention provides a technical solution: a method for tracking and supervising personnel working at high altitudes in power plants. The method uses a monitoring system, which includes a smart monitoring platform and ground monitoring AI equipment. The smart monitoring platform provides streaming media services, multiple algorithm recognition models and data services, and users can access it through web pages and mobile clients. The ground monitoring AI equipment includes cameras, which are horizontally placed at the construction site by the workers.

[0074] The ground monitoring AI equipment includes a video acquisition module, an image recognition module, a centering algorithm module, a linkage algorithm module, and a PTZ control module. The video acquisition module is responsible for acquiring real-time video streams from the high-altitude work site, ensuring clear video quality and providing a high-quality data source for subsequent image analysis. The image recognition module receives the video stream, uses image processing algorithms to preprocess the video frames, including noise reduction and enhancement, and uses deep learning and other technologies to achieve accurate detection and recognition of high-altitude workers. The centering algorithm module calculates the coordinates of the center point of the worker and the combined area of ​​multiple workers in real time, and determines whether the center point is within the preset safety threshold range. When the center point of the worker deviates from the safety threshold, the linkage algorithm module combines the worker's coordinates on the image and the current parameters of the PTZ control module, and the linkage control algorithm calculates the parameters that the PTZ control module needs to correct. The PTZ control module dynamically corrects the PTZ based on the parameters calculated by the linkage control algorithm, so that the worker returns to the center of the image.

[0075] The work includes the following steps:

[0076] S1. Place the ground monitoring AI equipment horizontally, start the system, and the ground monitoring AI equipment automatically adjusts the video monitoring angle to a 30-degree angle with the horizontal plane. The video acquisition module starts to acquire real-time video streams from the high-altitude operation site, and the image recognition module initializes the algorithm model to prepare for human body frame detection.

[0077] S2. Perform human body recognition to determine whether there are people in the current monitoring area. If no one is detected in the monitoring area three times in a row, execute S8 after 30 seconds of sleep. Otherwise, enter the high-altitude area personnel monitoring.

[0078] S3. Human bounding box filtering and optimization: After detecting a person in the supervised area, if multiple people are detected, further filtering and optimization are performed to improve the accuracy of tracking.

[0079] S4. Calculate the coordinates of the center point of the merged area of ​​all high-altitude personnel, calculate the merged area of ​​the human body frames of all identified high-altitude workers, and determine the coordinates of the center point of the merged area. Based on the convex hull algorithm and clustering algorithm, determine a merged area that can represent the activity range of the entire work group.

[0080] S5. Determine if the center point coordinates are within the threshold range;

[0081] S6. Monitoring area perspective adjustment and continuous deviation judgment;

[0082] S7. Judgment of overdue monitoring in high-altitude areas;

[0083] S8. Adjust the monitoring angle and enter the next high-altitude monitoring area;

[0084] In S3, the specific filtering rules include:

[0085] S3-1, Center Point Coordinate Filtering: Filters out human body frames whose center point is lower than a specific coordinate in the image, such as 800 in the image. For human body frames in a 1920*1080 image, this specific coordinate is determined based on a large amount of data analysis of high-altitude power construction site images, combined with the typical height of high-altitude workers and the characteristics of the image ratio. It can effectively distinguish high-altitude workers from non-work targets on the ground.

[0086] S3-2, Aspect Ratio Filtering: Filters out human bounding boxes whose height-to-width ratio is less than a specific threshold. This threshold can accurately match the typical morphological features of high-altitude workers and effectively eliminate non-human-shaped misidentifications such as poles, power lines, and tools caused by occlusion, lighting changes, or background complexity, thereby improving the robustness of high-altitude worker detection. These filtering parameters are adaptive or adjustable and can be dynamically optimized according to different high-altitude working environments such as pole type, working height, and lighting conditions, further enhancing the system's adaptability in complex environments.

[0087] In S4, the convex hull algorithm works as follows: By calculating the convex hull of the feature points of the human bounding box, a minimal convex polygon can be obtained. This polygon can effectively enclose all the workers at height, thus forming a merged region representing their collective activity range. The center point of the merged region is taken as its geometric center or centroid, as follows:

[0088] S4-1. Given a set of N human bounding boxes: B = {B1, B2, ..., B} N}, each bounding box B i From its upper left corner coordinate (x i ,y i ) and width and height (w i ,h i Definition: B i ={(x i ,y i ),w i ,h i};

[0089] S4-2, Each bounding box B i The set of four corner points V i :

[0090] V i =Corners(B i )

[0091] ={(x i ,y i ),(x i +w i ,y i ),(x i +w i ,y i +h i ),(x i ,y i +h i )}

[0092] Union of the corner points of all bounding boxes:

[0093]

[0094] S4-3. A convex hull is defined as the smallest convex set containing all points:

[0095]

[0096] Let the sequence of convex hull vertices be:

[0097] H = {h1,h2,…,h} m}, h j =(x hj ,yhj )

[0098] S4-4, Centroid coordinates of convex hull vertices:

[0099]

[0100] S4-5, Finally, we can obtain a tuple:

[0101] (x c ,y c ,H)

[0102] Where (x) c ,y c ) represents the coordinates of the merged center point, and H represents the sequence of convex hull vertices;

[0103] In S4, the clustering algorithm specifically discovers clusters of arbitrary shapes based on point density and identifies noise points. Core parameters include the neighborhood radius and the minimum number of points required to form dense regions. Based on extensive data analysis of images from high-altitude power construction sites and combined with typical population distributions of various high-altitude workers, the algorithm optimizes and dynamically adapts these core parameters. This method is particularly suitable for handling non-uniformly distributed groups of workers in high-altitude operations. When multiple high-altitude workers exist, the clustering algorithm can group spatially close bounding boxes that may belong to the same work group, automatically identifying different work groups, as detailed below:

[0104] S4-6. Given a set of N human bounding boxes:

[0105] B = {B1, B2, ..., B} N}

[0106] Each bounding box B i From its upper left corner coordinate (x i ,y i ) and width and height (w i ,h i )definition:

[0107] B i ={(x i ,y i ),w i ,h i}

[0108] Each bounding box B i The center point is:

[0109]

[0110] The set of all center points is:

[0111] P = {C1, C2, ..., C}N}

[0112] S4-7, DBSCAN clustering algorithm core parameters: ε: neighborhood radius, minPts: minimum number of neighboring points. For each point p∈P, its ε-neighborhood is defined as:

[0113] N ε (p)={q∈P∣|pq|≤ε}

[0114] If |N ε If (p)|≥minPts, then p is the core point. Through recursive expansion of the neighborhood of the core point, clusters are formed, and P is finally divided into K clusters and a set of noise points N:

[0115]

[0116] Where P k Let be the set of points in the k-th cluster;

[0117] S4-8. Find the convex hull and centroid of each cluster:

[0118] For each cluster P k Extract all corner points of its corresponding bounding box, denoted as Q. k :

[0119]

[0120] Q k Calculate the convex hull H k :

[0121] H k =ConvexHull(Q k )

[0122] The vertices of the convex hull are

[0123] The centroid of the convex hull of cluster k is:

[0124]

[0125] S4-9, The overall center of a multi-cluster:

[0126] If multiple clusters exist, calculate the centroid center of all clusters:

[0127]

[0128] S4-10, Special Cases:

[0129] If there are no effective clusters and all data is noise, then the mean of all center points is used directly:

[0130]

[0131] Ultimately, we can obtain (x) c ,y c ( ), that is, the center point after the merger;

[0132] In S5, the calculated coordinates of the worker and the center point of the merged area are compared with the preset center point threshold range. The preset center point threshold defines the center safety zone of the high-altitude area. For example, for a 1920*1080 image, the threshold range in the horizontal direction is set to 460~1460, and the threshold range in the vertical direction can be set to 280~800. If the center point coordinates are not within the threshold range, proceed to step 6; otherwise, it is assumed that the worker is in a safe centered state, and proceed to step 7.

[0133] In S6, if the center point coordinates are not within the threshold range, i.e. the operator deviates from the center, the system will record the deviation status and determine whether it is out of range three or more times consecutively. This continuous judgment mechanism is optimized to effectively filter out misjudgments and system jitter caused by momentary occlusion, image noise or brief non-critical movement of the operator, thereby significantly improving the stability, smoothness and efficiency of PTZ control, avoiding frequent and unnecessary adjustments, and reducing system energy consumption. If there is a continuous deviation, the centering algorithm module, linkage algorithm module and PTZ control module will perform the monitoring area perspective adjustment operation to bring the operator back to the center of the screen. The linkage control algorithm accurately calculates the translation, pitch and zoom parameters of PTZ based on the real-time position of the operator, the current parameters of the PTZ control module and the predicted movement trend, to ensure that the operator or merged area can return to the center of the screen smoothly, quickly and accurately, and adapt to the non-linear movement mode of high-altitude operators.

[0134] In S7, continuous monitoring of the high-altitude area is carried out, and it is determined whether the preset monitoring timeout time, such as 10 minutes, has been reached. If the timeout is exceeded, proceed to step 8; otherwise, proceed to step 2 to continue monitoring the current high-altitude area.

[0135] In S8, the system rotates 60 degrees to the right to enter the next high-altitude inspection area and restarts from Step 2. This is a resource-efficient coverage solution that replaces the high cost and complex deployment of multi-camera systems with the periodic, segmented scanning of a single ground AI device. The 60-degree rotation is optimized based on the camera's field of view and the typical range of high-altitude work areas, ensuring effective connection and seamless coverage between adjacent monitoring areas. The cyclic scanning mechanism works in conjunction with the high-altitude area supervision timeout judgment in Step 7 and the human body recognition in Step 2 to ensure that the system can respond quickly and refocus when anomalies occur in non-directly monitored areas. When a specific anomaly occurs in a non-currently monitored area, such as an audible alarm or external sensor trigger, the current scan is immediately interrupted and priority is given to turning to the abnormal area, thereby ensuring real-time response capability under 360-degree full coverage.

[0136] To avoid interference from ground personnel, the 30-degree angle is the optimal initial angle calculated based on the typical working height of high-altitude power construction, the relative position of ground equipment and the work area, and the field of view of the PTZ camera. This angle aims to maximize the visibility of high-altitude workers while minimizing interference from ground personnel and the ground background, thereby improving the accuracy and efficiency of subsequent image recognition.

[0137] The system first determines whether there are people in the current monitored area. During the continuous determination, the system enters a sleep state to save computing resources. This step aims to avoid unnecessary processing of empty areas and improve system efficiency.

[0138] This invention achieves: 1. High-precision, real-time centered tracking: By combining human bounding box detection, center point calculation, and threshold judgment, this invention can accurately and in real-time acquire the position of high-altitude workers in the image and determine whether they are located in the center of a preset high-altitude safety zone. This effectively avoids target loss and ensures continuous and effective monitoring of workers. This overcomes the shortcomings of existing technologies in positioning and tracking accuracy, providing reliable location information for high-altitude operations.

[0139] 2. Establish an efficient automatic correction mechanism: When workers deviate from the safety center area, the system can link with the monitoring equipment to adjust the viewing angle and automatically pull the deviated workers back to the center of the screen, realizing "automatic correction", reducing the need for manual intervention and improving the automation and intelligence level of supervision.

[0140] 3. Supports multi-target refined management: This invention achieves overall centered tracking of multiple workers by calculating the center point of the combined area of ​​all high-altitude workers. This makes it possible to control the overall safety situation of the work area in multi-target collaborative operation scenarios.

[0141] 4. Enhanced adaptability to complex environments: By optimizing the center point and aspect ratio of the human figure frame, this invention effectively eliminates interference from non-human or abnormal targets, improving the accuracy and robustness of identification in complex high-altitude environments (such as towers, power lines, and changes in lighting). This makes the system more environmentally adaptable in practical applications, enabling stable operation and providing reliable monitoring services.

[0142] 5. Resource-Optimized 360° Full Coverage: By periodically rotating and scanning at an optimized angle (60 degrees) using a single ground-based AI device, 360° full coverage of high-altitude areas is achieved. This solution significantly reduces equipment deployment costs and complexity, improves system portability and economy, and ensures continuous and comprehensive safety monitoring in vast high-altitude areas.

[0143] In summary, this invention effectively addresses the pain points in existing high-altitude power construction supervision by providing a central tracking and supervision method for high-altitude power construction workers. It offers an advanced and feasible technical solution for ensuring the safety of high-altitude workers and improving the efficiency of construction site management.

[0144] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0145] Finally, it should be noted that the above descriptions are merely preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for centrally tracking and supervising personnel during high-altitude power construction, characterized in that: This method employs a monitoring system, which includes an intelligent monitoring platform and ground-based AI monitoring equipment. The intelligent monitoring platform provides streaming media services, various algorithm recognition models, and data services, and users can access it through web pages and mobile clients. The ground-based AI monitoring equipment includes cameras, which are horizontally positioned at the construction site by the operators.

2. The method for centrally tracking and supervising personnel during high-altitude power construction according to claim 1, characterized in that: The ground monitoring AI equipment includes a video acquisition module, an image recognition module, a centering algorithm module, a linkage algorithm module, and a PTZ control module. The video acquisition module is responsible for acquiring real-time video streams from the high-altitude work site, ensuring clear video quality and providing a high-quality data source for subsequent image analysis. The image recognition module receives the video stream, preprocesses the video frames using image processing algorithms, including noise reduction and enhancement, and uses deep learning and other technologies to achieve accurate detection and identification of high-altitude workers. The centering algorithm module calculates the coordinates of the center point of the worker and the combined area of ​​multiple workers in real time, and determines whether the center point is within a preset safety threshold range. When the center point of the worker deviates from the safety threshold, the linkage algorithm module, combined with the worker's coordinates on the image and the current parameters of the PTZ control module, calculates the parameters that the PTZ control module needs to correct. The PTZ control module, based on the parameters calculated by the linkage control algorithm, achieves dynamic PTZ correction, bringing the worker back to the center of the image.

3. The method for centrally tracking and supervising personnel during high-altitude power construction according to claim 2, characterized in that: The work includes the following steps: S1. Place the ground monitoring AI equipment horizontally, start the system, and the ground monitoring AI equipment will automatically adjust the video monitoring angle to 30 degrees with the horizontal plane. The video acquisition module will start to acquire the real-time video stream of the high-altitude operation site, and the image recognition module will initialize the algorithm model to prepare for human body frame detection. S2. Perform human body recognition to determine whether there are people in the current monitoring area. If no one is detected in the monitoring area three times in a row, execute S8 after 30 seconds of sleep. Otherwise, enter the high-altitude area personnel monitoring. S3. Human bounding box filtering and optimization: After detecting a person in the supervised area, if multiple people are detected, further filtering and optimization are performed to improve the accuracy of tracking. S4. Calculate the coordinates of the center point of the merged area of ​​all high-altitude personnel, calculate the merged area of ​​the human body frames of all identified high-altitude workers, and determine the coordinates of the center point of the merged area. Based on the convex hull algorithm and clustering algorithm, determine a merged area that can represent the activity range of the entire work group. S5. Determine if the center point coordinates are within the threshold range; S6. Monitoring area perspective adjustment and continuous deviation judgment; S7. Judgment of overdue monitoring in high-altitude areas; S8. Adjust the monitoring angle and enter the next high-altitude monitoring area.

4. The method for centrally tracking and supervising personnel during high-altitude power construction according to claim 3, characterized in that: In S3, the specific filtering rules include: S3-1, Center Point Coordinate Filtering: Filters out human body frames whose center point is lower than a specific coordinate in the image. This specific coordinate is determined based on a large amount of data analysis of high-altitude power construction site images, combined with the typical height of high-altitude workers and the proportion characteristics of the image. S3-2, Aspect Ratio Filtering: Filter out human body frames whose height to width ratio is less than a specific threshold.

5. A method for centrally tracking and supervising personnel during high-altitude power construction, as described in claim 4, characterized in that: In S4, the convex hull algorithm specifically involves calculating the convex hull of the feature points of the human body bounding box to obtain a minimal convex polygon. This polygon encloses all the workers at height, thus forming a merged region representing their collective activity range. The center point of the merged region is taken as its geometric center, as detailed below: S4-1. Given a set of N human bounding boxes: B = {B1, B2, ..., B} N }, each bounding box B i From its upper left corner coordinate (x i ,y i ) and width and height (w i ,h i Definition: B i ={(x i ,y i ),w i ,h i }; S4-2, Each bounding box B i The set of four corner points V i : V i =Corners(B i ) ={(x i ,y i ),(x i +w i ,y i ),(x i +w i ,y i +h i ),(x i ,y i +h i )} Union of the corner points of all bounding boxes: S4-3. A convex hull is defined as the smallest convex set containing all points: Let the sequence of convex hull vertices be: H={h1,h2,…,h m },h j =(x hj ,y hj ) S4-4, Centroid coordinates of convex hull vertices: S4-5, Finally, the tuple is obtained: (x c ,y c ,H) Where (x) c ,y c ) represents the coordinates of the merged center point, and H represents the sequence of convex hull vertices.

6. A method for centrally tracking and supervising personnel during high-altitude power construction, as described in claim 5, is characterized in that: In S4, the clustering algorithm specifically involves: discovering clusters of arbitrary shapes based on point density and identifying noise points. Core parameters include the neighborhood radius and the minimum number of points required to form dense regions. Based on extensive data analysis of high-altitude power construction site images and combined with typical population distributions of various high-altitude workers, the algorithm optimizes and dynamically adapts the core parameters. When multiple high-altitude workers exist, the clustering algorithm groups spatially close bounding boxes that may belong to the same work group, automatically identifying different work groups, as detailed below: S4-6, Each bounding box B i The center point is: The set of all center points is: P={C1,C2,…,C N } S4-7, DBSCAN clustering algorithm core parameters: ε: neighborhood radius, minPts: minimum number of neighborhood points. For each point p∈P, its ε-neighborhood is defined as: N ε (p)={q∈P∣|p-q|≤ε} If |N ε If (p)|≥minPts, then p is the core point. Through recursive expansion of the neighborhood of the core point, clusters are formed, and finally P is divided into K clusters and a set of noise points N: Where P k Let be the set of points in the k-th cluster; S4-8. Find the convex hull and centroid of each cluster: For each cluster P k Extract all corner points of its corresponding bounding box, denoted as Q. k : Q k Calculate the convex hull H k : H k =ConvexHull(Q k ) The vertices of the convex hull are The centroid of the convex hull of cluster k is: S4-9, The overall center of a multi-cluster: If multiple clusters exist, calculate the centroid center of all clusters: S4-10, Special Cases: If there is no effective clustering (all data is noise), then use the mean of all center points directly: Ultimately, we can obtain (x) c ,y c ), which is the center point after the merger.

7. A method for centrally tracking and supervising personnel during high-altitude power construction according to claim 6, characterized in that: In step S5, the calculated coordinates of the operator and the center point of the merged area are compared with the preset center point threshold range. The preset center point threshold defines the central safe area of ​​the high-altitude area. If the center point coordinates are not within the threshold range, step S6 is executed; otherwise, the operator is considered to be in a safe and centered state, and step S7 is executed.

8. A method for centrally tracking and supervising personnel during high-altitude power construction, as described in claim 7, characterized in that: In step S8, the system rotates 60 degrees to the right to enter the next high-altitude inspection area and restarts from step 2. Through periodic and segmented scanning by a single ground AI device, 360-degree scene coverage is achieved.

Images