Intelligent positioning analysis method and system for safety tools

Abstract

The invention relates to the technical field of wireless positioning, in particular to an intelligent positioning analysis method and system for a safety tool, and the method comprises the steps: collecting the video data and voice data of a high-altitude operation site in real time through a video collection module worn by an operator; the method comprises the following steps: acquiring far-end high-altitude operation video data through a ground camera, and when the far-end high-altitude operation video data is lower than a preset value, acquiring three-dimensional position data of personnel in real time through a wireless positioning module; establishing a three-dimensional electronic fence model on the monitoring platform, and synchronizing the three-dimensional position data to the three-dimensional electronic fence model; when the monitoring platform detects that the three-dimensional position data exceed the three-dimensional electronic fence, early warning is triggered; after early warning is triggered, video data and voice data of the aerial work site in the corresponding time period are obtained, personnel posture information in the video data is analyzed, and abnormal types are recognized in combination with voice keywords. Therefore, personnel supplementary positioning under the condition of low visibility can be realized, and the reliability of the positioning monitoring system is improved.

Description

Technical Field

[0001] The present invention relates to the field of wireless positioning technology, and in particular to an intelligent positioning analysis method and system for safety tools. Background Art

[0002] Aerial work safety monitoring refers to the process of real-time monitoring and assessment of the work environment, work processes, and worker status using various technical means and equipment to ensure the safety of workers and prevent accidents during high-altitude operations. This includes, but is not limited to, the use of sensors to monitor meteorological conditions and structural stability, wearable devices to track worker health and location, and video surveillance systems to monitor operational compliance. Effective aerial work safety monitoring can promptly identify potential risks and provide early warnings, enabling appropriate measures to prevent accidents and ensure smooth operations.

[0003] Existing risk warning systems mostly use cameras for positioning, but in low visibility conditions, such as fog, snow, and heavy rain, visual observation is limited, thereby reducing the effectiveness of positioning. Summary of the Invention

[0004] The purpose of the present invention is to provide an intelligent positioning analysis method and system for safety tools, aiming to achieve supplementary positioning of personnel in low visibility conditions to improve the reliability of the positioning monitoring system.

[0005] To achieve the above objectives, in a first aspect, the present invention provides an intelligent positioning and analysis method for safety tools, comprising collecting video data and voice data of a high-altitude operation site in real time through a video acquisition module worn by an operator;

[0006] The ground camera collects remote high-altitude operation video data. When the remote high-altitude operation video data is lower than the preset value, the wireless positioning module obtains the personnel's three-dimensional position data in real time;

[0007] Establish a 3D electronic fence model on the monitoring platform and synchronize the 3D location data to the 3D electronic fence model;

[0008] When the monitoring platform detects that the 3D position data exceeds the 3D electronic fence, an alarm is triggered;

[0009] After the warning is triggered, the video data and voice data of the high-altitude operation site in the corresponding time period are obtained to analyze the posture information of the personnel in the video data and identify the abnormal type in combination with the voice keywords.

[0010] The specific steps of collecting video data and voice data of the aerial work site in real time through the video acquisition module worn by the operator include:

[0011] Install the video acquisition module and the audio acquisition module on the operator;

[0012] Obtain video and sound data from aerial work sites;

[0013] Send video and sound data to the monitoring center.

[0014] The specific steps of collecting remote high-altitude operation video data through a ground camera and obtaining the three-dimensional position data of the personnel in real time through a wireless positioning module when the remote high-altitude operation video data is lower than a preset value include:

[0015] Ground cameras collect ground video data of aerial operations;

[0016] Calculate image clarity based on ground video data;

[0017] When the image clarity is lower than the preset value, a positioning request is sent to the wireless positioning module worn by the aerial workers;

[0018] The three-dimensional position data of the high-altitude workers is calculated using radio positioning technology based on the positioning request.

[0019] The specific steps of calculating the three-dimensional position data of the high-altitude worker using radio positioning technology based on the positioning request include:

[0020] Equip workers working at height with UWB tags and install multiple UWB anchor points in the work area;

[0021] The UWB tag periodically sends short pulse signals to all anchor points, and each anchor point records the timestamp of the received signal;

[0022] The distance between the tag and each anchor point is calculated based on the time difference of receiving the signal at each anchor point, and multiple sets of distance data are obtained;

[0023] The three-dimensional position data of the UWB tag is calculated using the least squares method by combining multiple sets of distance data.

[0024] The specific steps of establishing a three-dimensional electronic fence model on the monitoring platform and synchronizing the three-dimensional position data into the three-dimensional electronic fence model include:

[0025] Obtain BIM model files for construction projects and set up electronic fences in key areas based on safety requirements for aerial work;

[0026] Draw the electronic fence area based on the key area of ​​the electronic fence and set the trigger conditions of the electronic fence area;

[0027] Convert 3D position data into local coordinate data used by BIM models;

[0028] Synchronize local coordinate data to the geo-fence area.

[0029] The specific steps of obtaining the BIM model file of the construction project and setting up the key areas of the electronic fence according to the safety requirements of the high-altitude operation include:

[0030] Select the ground range to define the two-dimensional boundary;

[0031] Set vertical boundaries based on the actual height requirements of high-altitude operations;

[0032] For irregular areas, import pre-defined boundary files to obtain key areas.

[0033] The specific steps of triggering an early warning when the monitoring platform detects that the three-dimensional position data exceeds the three-dimensional electronic fence include:

[0034] Obtain the two-dimensional coordinate data x and y of the aerial worker;

[0035] Get a list of all vertex coordinates in the key area;

[0036] Select the horizontal right direction as the ray direction and set the intersection counter;

[0037] For each edge of the polygon, check if it intersects the ray. If so, update the intersection counter based on the exact location of the intersection: if the intersection is strictly above the ray, ignore it; if the intersection falls exactly at the starting point of the ray, the point is considered to be on the boundary. In other cases, increase the counter value;

[0038] If the count value is odd, it means the point is inside the critical area; if it is even, the point is outside the critical area;

[0039] If the point is within the critical area, the height z of the point is further checked to see if it meets the height limit of the geo-fence. If it does not meet the height limit, an early warning is triggered.

[0040] Among them, after the warning is triggered, the specific steps of obtaining the high-altitude work site video data and voice data of the corresponding time period, analyzing the personnel posture information in the video data, and identifying the abnormal type in combination with the voice keywords include:

[0041] Extract the video and audio data of the aerial work site in the corresponding time period according to the timestamp of the warning trigger;

[0042] Define the normal posture range according to the standard operating procedures for working at heights;

[0043] Use human posture estimation algorithms to detect the posture information of people in aerial work site video data;

[0044] Compare the current posture with the normal posture range to detect abnormal behavior;

[0045] Use automatic speech recognition technology to convert speech signals into text, define a set of keywords related to anomalies, and search for these keywords in the converted text to determine whether there are any anomalies.

[0046] Generate corresponding anomaly types based on detected abnormal behaviors and abnormal conditions.

[0047] The specific steps of using the human posture estimation algorithm to detect the posture information of people in the aerial work site video data include:

[0048] Obtain video streams or recordings from cameras at aerial work sites and decompose the video into a series of single-frame images;

[0049] Use Faster R-CNN to locate the people in each frame and obtain the people data;

[0050] The OpenPose algorithm is applied based on personnel data to identify the key points of each human body and obtain posture information.

[0051] In a second aspect, the present invention further provides an intelligent positioning and analysis system for safety tools, which adopts the intelligent positioning and analysis method for safety tools.

[0052] The present invention provides an intelligent positioning and analysis method and system for safety tools. At the high-altitude work site, workers wear video capture modules equipped with intelligent cameras and microphones to capture high-definition video data and voice information during the work process in real time. Simultaneously, a remote high-altitude work camera, mounted on the ground or in a fixed location, continuously captures the entire high-altitude work area. Through multi-angle, multi-device video capture, comprehensive coverage of the entire work area is achieved, enhancing the monitoring system's perception capabilities. When the video clarity captured by the remote camera falls below a preset threshold (e.g., image blur or severe occlusion) due to altitude, weather, or other factors, the system automatically switches to the wireless positioning module. By deploying technologies such as UWB, Bluetooth beacons, or Beidou / GPS, the three-dimensional spatial coordinate information of workers and their tools is acquired in real time, ensuring accurate positioning even in visually restricted conditions. Within the backend monitoring platform, a three-dimensional electronic fence model is constructed using BIM (Building Information Modeling) or GIS (Geographic Information System) technology, and the real-time three-dimensional position data acquired is dynamically synchronized into the model. The three-dimensional electronic fence can be flexibly configured according to the actual work range, setting different levels of safety boundaries. When the system detects that an employee or their tools have exceeded the boundaries of a pre-defined three-dimensional electronic fence, an alert is immediately triggered. This alert is notified to on-site management and the remote command center via platform pop-up windows, audio and visual alarms, and SMS / WeChat push notifications, enabling rapid response. Once the alert is triggered, the system automatically backtracks and extracts on-site video and audio data corresponding to the time period of the alert to identify and analyze anomalies. This allows for additional personnel positioning in low-visibility situations, improving the reliability of the positioning monitoring system. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0054] Figure 1 The present invention is a flowchart of an intelligent positioning analysis method for safety tools.

[0055] Figure 2 This is a flow chart of the present invention for collecting video data and voice data of a high-altitude operation site in real time through a video acquisition module worn by an operator.

[0056] Figure 3This is a flow chart of the present invention for collecting remote high-altitude work video data through a ground camera, and when the remote high-altitude work video data is lower than a preset value, obtaining the three-dimensional position data of the personnel in real time through a wireless positioning module.

[0057] Figure 4 This is a flow chart of the present invention for calculating the three-dimensional position data of a person working at height using radio positioning technology based on a positioning request.

[0058] Figure 5 This is a flow chart of the present invention for establishing a three-dimensional electronic fence model on a monitoring platform and synchronizing three-dimensional position data into the three-dimensional electronic fence model.

[0059] Figure 6 This is a flowchart of the present invention for obtaining a BIM model file of a construction project and setting up key areas of an electronic fence according to the safety requirements of high-altitude operations.

[0060] Figure 7 This is a flowchart of the present invention for triggering an early warning when the monitoring platform detects that the three-dimensional position data exceeds the three-dimensional electronic fence.

[0061] Figure 8 This is a flowchart of the present invention that obtains high-altitude work site video data and voice data of the corresponding time period after triggering the early warning, analyzes the personnel posture information in the video data, and combines voice keywords to identify the abnormal type. DETAILED DESCRIPTION

[0062] The following describes embodiments of the present invention in detail, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present invention, and are not to be construed as limiting the present invention.

[0063] First embodiment

[0064] See also Figures 1 to 8 The present invention also provides an intelligent positioning and analysis method for safety tools, comprising:

[0065] S101 collects real-time video and voice data from aerial work sites through the video acquisition module worn by the operator;

[0066] The specific steps include:

[0067] S201: Install the video acquisition module and the audio acquisition module on the operator;

[0068] Before the operation begins, the operator wears the integrated video acquisition equipment (such as a high-definition camera, panoramic camera or wearable recorder) and audio acquisition equipment (such as a high-sensitivity microphone or pickup) on the appropriate part of the body, such as the head, chest or shoulder. It can also be integrated into protective equipment such as safety helmets and work vests.

[0069] S202 acquires video data and sound data of the aerial work site;

[0070] When the operator enters the high-altitude work area and starts operating, the video acquisition module captures high-definition video images of the work site in real time, covering the operator's operating actions, location, work object and surrounding environment information; at the same time, the audio acquisition module synchronously collects sound signals at the site, including communication between operators, equipment operation sounds, sudden abnormal sounds, etc.

[0071] S203 sends the video data and sound data to the monitoring center.

[0072] After being compressed and converted by a local processing unit, the collected audio and video data is uploaded to a remote monitoring center in real time via a wireless transmission module (such as a 4G / 5G communication module, Wi-Fi module, or satellite communication module). Encrypted communication protocols are used during transmission to ensure data security and integrity. Once received, the monitoring center displays the data in real time on a large monitor or workstation for management review, analysis, and emergency response.

[0073] S102 collects remote high-altitude work video data through ground cameras. When the remote high-altitude work video data is lower than the preset value, the wireless positioning module obtains the three-dimensional position data of the personnel in real time;

[0074] The specific steps include:

[0075] The S301 ground camera collects ground video data for aerial work;

[0076] Deploy multiple high-definition cameras (such as PTZ cameras, panoramic cameras, or infrared thermal imaging cameras) around the aerial work area and install them on fixed ground supports or high-point locations such as tower cranes to provide remote, wide-angle coverage of the aerial work area. The cameras capture real-time video footage of the work site, including the aerial work platform, personnel posture, tool usage status, and surrounding conditions.

[0077] S302 calculates picture clarity based on ground video data;

[0078] The captured video data is transmitted to the image processing unit, where the Sobel operator is selected to identify edge information in the image. The selected edge detection algorithm is applied to each frame of the video image to generate an edge image. In this process, significant boundaries in the image are highlighted, while the background or other irrelevant parts are weakened. A clarity index is calculated based on the generated edge image, specifically by counting the number of non-zero pixels in the edge image or calculating the sum of the gradient amplitudes of the edge pixels. Higher values ​​generally indicate higher clarity.

[0079] S303 sends a positioning request to the wireless positioning module worn by the aerial worker when the image clarity is lower than a preset value;

[0080] If the clarity of the captured image at a certain moment falls below a preset threshold (for example, due to weather changes, poor lighting conditions, severe occlusion, etc.), the system automatically determines that "effective monitoring is impossible" and immediately triggers the auxiliary positioning mechanism. At this time, the system sends a positioning request signal to the wireless positioning module worn by the high-altitude worker, requesting it to feedback its current location information.

[0081] S304 calculates the three-dimensional position data of the aerial worker using radio positioning technology based on the positioning request.

[0082] The specific steps include:

[0083] S401 equips workers working at height with UWB tags and installs multiple UWB anchor points in the work area;

[0084] In this embodiment, ultra-wideband (UWB) positioning technology is used. Each aerial worker wears a low-power, lightweight UWB tag. At the same time, at least four UWB anchor points are deployed at key locations in the work area (such as steel structure supports, platform corners, tower crane bases, etc.) to form a three-dimensional positioning network.

[0085] The S402 UWB tag periodically sends short pulse signals to all anchor points, and each anchor point records the timestamp of the received signal;

[0086] UWB tags transmit short, high-frequency pulses at regular intervals (e.g., 100 milliseconds). Each anchor receives the signal and precisely records the moment it arrives. Because UWB signals have nanosecond-level time resolution, they enable extremely high-precision time synchronization and measurement.

[0087] S403 calculates the distance between the tag and each anchor point based on the time difference of receiving the signal at each anchor point, and obtains multiple sets of distance data;

[0088] Using the time-of-flight method, the straight-line distance between the tag and each anchor point is calculated based on the signal propagation speed and the timestamp difference recorded at the anchor point. This yields several sets of relative distance information between the tag and the anchor point.

[0089] S404 combines multiple sets of distance data and uses the least square method to calculate the three-dimensional position data of the UWB tag.

[0090] The above-mentioned multiple sets of distance data are substituted into a geometric model in a three-dimensional coordinate system and optimized using the least squares method to eliminate the deviation caused by measurement errors. The final result is the spatial coordinates (x, y, z) of the tag (i.e., the operator). This three-dimensional position data can be uploaded to the monitoring center for functional modules such as personnel positioning, trajectory mapping, and electronic fence crossing alarm.

[0091] S103 establishes a three-dimensional electronic fence model on the monitoring platform and synchronizes the three-dimensional position data to the three-dimensional electronic fence model;

[0092] The specific steps include:

[0093] S501 obtains the BIM model file of the construction project and sets up the key areas of the electronic fence according to the safety requirements of the high-altitude operation;

[0094] The specific steps include:

[0095] S601 selects a ground range to define a two-dimensional boundary;

[0096] In the BIM model, the ground activity areas corresponding to aerial work are first identified, such as lifting areas, scaffolding foundation areas, and construction elevator stops. The system uses graphical tools to manually or automatically outline these areas, forming a two-dimensional boundary that serves as the base plane range for the electronic fence.

[0097] S602 sets vertical boundaries based on the actual height requirements of the aerial work;

[0098] After completing the setting of the ground boundary, the vertical boundary of the electronic fence is extended upward according to the operating height of different types of high-altitude operations (such as 3 meters, 10 meters, and above 30 meters), combined with factors such as the floor elevation, steel beam distribution, and work platform height in the building structure, thereby constructing a three-dimensional fence space to prevent workers from entering dangerous or unrestricted height areas.

[0099] S603 imports a predefined boundary file for the irregular area to obtain a key area.

[0100] For building areas with complex shapes and non-standard geometries (such as curved roofs, special-shaped towers, and cantilevered structures), you can import boundary definition files converted from CAD drawings (such as IFC, STL, or custom JSON boundary configuration files). After the system parses the file, it embeds it into the BIM model, accurately restoring the 3D shape of irregular areas and constructing critical operating areas and restricted areas that better reflect actual working conditions.

[0101] S502 draws an electronic fence area based on the key area of ​​the electronic fence and sets the triggering conditions of the electronic fence area;

[0102] After identifying all critical areas, the system visually marks these areas as electronic fences and sets corresponding safety trigger logic.

[0103] For example, a level one alarm is triggered when a worker enters a restricted area; a level two alarm is triggered if they are not wearing a safety harness or exceed the specified stay time; a boundary crossing alarm is triggered when a person signals outside the electronic fence; and an emergency response mechanism is activated if a fall risk is detected, based on biosensor data. These trigger conditions can be dynamically adjusted based on different types of work, weather conditions, and operation phases, enhancing the system's flexibility and practicality.

[0104] S503 converts the three-dimensional position data into local coordinate data used by the BIM model;

[0105] The 3D position data of people obtained from the UWB positioning module usually uses a global coordinate system (such as the WGS84 geographic coordinate system or the device's customized relative coordinates), while the BIM model operates in its own unique local coordinate system. Therefore, coordinate conversion processing is required. Specifically, this includes establishing a mapping relationship between global coordinates and BIM local coordinates; unifying the 3D position data into the BIM model coordinate system through a coordinate transformation matrix (translation, rotation, and scaling); and performing error correction on the coordinate conversion process to ensure that positioning accuracy meets centimeter-level requirements.

[0106] S504 synchronizes the local coordinate data to the electronic fence area.

[0107] The coordinate-converted worker location information is overlaid in real time onto the defined geo-fence model, ensuring a match between the spatial location and the pre-defined safety zone. The system continuously updates the worker's location status and compares it with the geo-fence attributes of the area in which they are located. If a worker deviates from the planned path, enters a restricted area, or remains stationary for an extended period, the system immediately issues an alert to the monitoring center. This alert is then integrated with multi-channel response mechanisms, including video surveillance screen switching, voice broadcast notifications, and mobile device push notifications, enabling proactive safety management.

[0108] S104 triggers an alarm when the monitoring platform detects that the three-dimensional position data exceeds the three-dimensional electronic fence;

[0109] The specific steps include:

[0110] S701 obtains the two-dimensional coordinate data x and y of the person working at height;

[0111] The horizontal X (east) and Y (north) coordinates are extracted from the three-dimensional position information of the person acquired by the UWB positioning module to form the two-dimensional coordinate point P(x,y) of the person's current position. These coordinates are converted and mapped to the local coordinate system used by the BIM model to ensure spatial consistency with the geo-fence model.

[0112] S702 obtains a list of all vertex coordinates in the key area;

[0113] Retrieve the boundary vertex information of the key operation area or restricted area preset in the electronic fence. This information is stored in the system database in the form of polygons, which contains the coordinates of multiple vertices connected in sequence to form a closed two-dimensional geometric figure. For example, the vertex set is represented as Polygon = [(x1, y1), (x2, y2), ..., (x n ,y n )].

[0114] S703 selects the horizontal rightward direction as the ray direction and sets the intersection counter;

[0115] The ray intersection method is used to determine whether a point is inside a polygon. Specifically, an infinitely long ray is extended horizontally to the right from the point to be tested, and an intersection counter is initialized to 0. This method is widely used in computer graphics and exhibits good robustness and efficiency.

[0116] S704 checks each edge of the polygon to see if it intersects the ray. If so, updates the intersection counter based on the specific location of the intersection: if the intersection is strictly above the ray, ignore it; if the intersection falls exactly at the starting point of the ray, the point is considered to be on the boundary. In other cases, increment the counter;

[0117] Traverse the edge set of the polygon one by one Edge=[(x1,y1)-(x2,y2),(x2,y2)-(x3,y3),...,(x n ,y n )-(x1,y1)], and perform the following judgment and processing on each edge:

[0118] If the edge has no intersection with the ray, skip it;

[0119] If there is an intersection, the relative position relationship between the intersection and the ray is further determined:

[0120] If the intersection point is strictly above the ray (i.e., its y value is greater than the y value of point P), then ignore the intersection point;

[0121] If the intersection point falls exactly at the starting point of the ray (that is, the intersection point coincides with point P), then the point is determined to be on the boundary of the polygon;

[0122] In other cases, the intersection counter Count is increased by 1.

[0123] S705: If the count value is an odd number, it indicates that the point is inside the key area; if it is an even number, the point is outside the key area;

[0124] After traversing all edges, the ray intersection method is used to determine whether the ray crosses the polygon an odd number of times, indicating that point P is inside the polygon. If Count is even, point P is outside the polygon. This judgment result is used to determine whether the person working at height is within the designated safe working area.

[0125] S706 If the point is within the critical area, further check whether the height z of the point meets the height limit of the electronic fence. If it does not meet the height limit, trigger an early warning.

[0126] Even if the XY coordinates are within the critical area, the Z-axis height coordinate must still be verified to ensure it meets the vertical range requirements set by the geo-fence. For example, certain operations are only permitted between +3 meters and +10 meters above floor level. If the Z value is within the permitted range, the operator is considered fully compliant; if the Z value is below or above the set threshold, it is considered a violation of the hazardous height zone.

[0127] After S105 triggers the warning, the video data and voice data of the aerial work site in the corresponding time period are obtained, the posture information of the personnel in the video data is analyzed, and the abnormal type is identified in combination with the voice keywords.

[0128] The specific steps include:

[0129] S801 extracts the video data and audio data of the aerial work site corresponding to the time period according to the timestamp of the warning trigger;

[0130] When the system issues an alert, it records the current timestamp, T0, and uses that time as a benchmark to retrieve the corresponding video and audio data segments from a local storage device or cloud server for a set period of time (e.g., 30 or 60 seconds). This period should be sufficient to cover the entire process before and after the abnormal event.

[0131] The extracted data includes: close-up video from wearable cameras; panoramic video from ground-mounted cameras; and ambient voice data collected by microphones. This data will serve as the basic input for subsequent intelligent analysis.

[0132] S802 defines the normal posture range according to the standard operating procedures for working at heights;

[0133] Based on industry standards, company safety regulations, and historical experience, a reference model for "normal posture" during aerial work is established. For example: safety belts are correctly fastened; the body's center of gravity is stable; both feet are on a stable platform; and hand movements do not exceed the body's controllable balance.

[0134] S803 uses a human posture estimation algorithm to detect posture information of people in the video data of the aerial work site;

[0135] The specific steps include:

[0136] Obtain video streams or recordings from cameras at aerial work sites and decompose the video into a series of single-frame images;

[0137] Real-time video streams are acquired from fixed or mobile cameras deployed at aerial work sites, or historical video recording files are read from storage devices. Continuous video streams are decomposed into a series of image frames to form a static image sequence that can be processed frame by frame by the algorithm. Necessary preprocessing operations are performed on each frame, including but not limited to grayscale conversion, resizing, image enhancement, and noise removal, to improve the accuracy of subsequent target detection and posture estimation.

[0138] Use Faster R-CNN to locate people in each frame and obtain their data. Use the pre-trained Faster R-CNN object detection model in each frame to detect all people in the image. The detection results output the bounding box coordinates and confidence score of each person in the image, which are used to identify all areas in the image that may contain people. Strategies such as multi-scale detection and non-maximum suppression (NMS) can be introduced to further optimize detection results and avoid duplicate detection and false detection.

[0139] The OpenPose algorithm is applied based on personnel data to identify the key points of each human body and obtain posture information.

[0140] Based on the human region detected by Faster R-CNN, the corresponding local image is cropped within each detection frame and input into the OpenPose pose estimation algorithm;

[0141] OpenPose is a multi-person 2D keypoint detection algorithm based on convolutional neural networks. It can simultaneously detect key points of body parts (such as the head, neck, shoulders, elbows, wrists, hips, knees, and ankles) and their connections (skeleton lines). The algorithm predicts the coordinates of multiple key points of the human body within each detection area and constructs a complete posture skeleton based on these coordinates. It outputs structured data, such as the coordinates (x, y) of each person's joint points and their confidence levels, to facilitate subsequent behavioral analysis and posture pattern recognition.

[0142] S804 compares the current posture with the normal posture range to detect whether there is abnormal behavior;

[0143] The system compares the posture information detected in each frame with a preset "normal posture range" to determine whether there are any deviations from safety standards, such as: not wearing a safety rope; leaning significantly; supporting oneself with one hand or one foot; sudden falls or rapid descents. Once such posture states are detected, they are marked as potentially risky behaviors, and their duration and frequency are counted.

[0144] S805 uses automatic speech recognition technology to convert speech signals into text, defines a set of keywords related to anomalies, and searches for these keywords in the converted text to determine whether there are any anomalies.

[0145] Perform speech recognition processing on the synchronously extracted audio data and use technology such as DeepSpeech to convert the speech into text strings.

[0146] The system then defines a set of keywords based on typical abnormal semantics in high-altitude work scenarios, such as "slipped," "fell," "help," "hold on," and "call for help." Using text matching algorithms (such as regular expressions, keyword filters, and NLP semantic similarity analysis), the system searches for these keywords in the identified text. If such urgent or dangerous statements are detected in the speech, the system further confirms the possibility of an abnormality.

[0147] S806 generates a corresponding abnormality type based on the detected abnormal behavior and abnormal situation.

[0148] Combining the posture analysis results with the speech recognition results, the system performs a multimodal fusion judgment and outputs the final abnormality type label, as shown in Table 1.

[0149] Table 1

[0150]

[0151] At the same time, the system can package abnormal events and related video, voice, and positioning data and archive them into the event database for subsequent manual review, evidence analysis, and safety training.

[0152] Second embodiment

[0153] The present invention further provides an intelligent positioning and analysis system for safety tools, which adopts the intelligent positioning and analysis method for safety tools described in the first embodiment.

[0154] The above disclosure is only a preferred embodiment of the present invention, and certainly cannot be used to limit the scope of the rights of the present invention. Ordinary technicians in this field can understand that all or part of the processes of the above embodiment and equivalent changes made in accordance with the claims of the present invention are still within the scope of the invention.

Claims

1. An intelligent positioning and analysis method for safety tools, characterized in that: include: The video acquisition module worn by the operator collects real-time video and voice data from the aerial work site; The ground camera collects remote high-altitude operation video data. When the remote high-altitude operation video data is lower than the preset value, the wireless positioning module obtains the personnel's three-dimensional position data in real time; Establish a 3D electronic fence model on the monitoring platform and synchronize the 3D location data to the 3D electronic fence model; When the monitoring platform detects that the 3D position data exceeds the 3D electronic fence, an alarm is triggered; After the warning is triggered, the video data and voice data of the high-altitude operation site in the corresponding time period are obtained to analyze the posture information of the personnel in the video data and identify the abnormal type in combination with the voice keywords.

2. The intelligent positioning analysis method for safety tools according to claim 1, characterized in that: The specific steps of collecting video data and voice data of the aerial work site in real time by the video acquisition module worn by the operator include: Install the video acquisition module and the audio acquisition module on the operator; Obtain video and sound data from aerial work sites; Send video and sound data to the monitoring center.

3. The intelligent positioning and analysis method for safety tools according to claim 2, characterized in that: The specific steps of collecting remote high-altitude operation video data through the ground camera and obtaining the personnel's three-dimensional position data in real time through the wireless positioning module when the remote high-altitude operation video data is lower than a preset value include: Ground cameras collect ground video data of aerial operations; Calculate image clarity based on ground video data; When the image clarity is lower than the preset value, a positioning request is sent to the wireless positioning module worn by the aerial workers; The three-dimensional position data of the high-altitude workers is calculated using radio positioning technology based on the positioning request.

4. The intelligent positioning analysis method for safety tools according to claim 3, characterized in that: The specific steps of calculating the three-dimensional position data of the high-altitude worker using the radio positioning technology based on the positioning request include: Equip workers working at height with UWB tags and install multiple UWB anchor points in the work area; The UWB tag periodically sends short pulse signals to all anchor points, and each anchor point records the timestamp of the received signal; The distance between the tag and each anchor point is calculated based on the time difference of receiving the signal at each anchor point, and multiple sets of distance data are obtained; The three-dimensional position data of the UWB tag is calculated using the least squares method by combining multiple sets of distance data.

5. The intelligent positioning analysis method for safety tools according to claim 4, characterized in that: The specific steps of establishing a three-dimensional electronic fence model on the monitoring platform and synchronizing the three-dimensional position data to the three-dimensional electronic fence model include: Obtain BIM model files for construction projects and set up electronic fences in key areas based on safety requirements for aerial work; Draw the electronic fence area based on the key area of ​​the electronic fence and set the trigger conditions of the electronic fence area; Convert 3D position data into local coordinate data used by BIM models; Synchronize local coordinate data to the geo-fence area.

6. The intelligent positioning and analysis method for safety tools according to claim 5, characterized in that: The specific steps of obtaining the BIM model file of the construction project and setting up the key areas of the electronic fence according to the safety requirements of the high-altitude operation include: Select the ground range to define the two-dimensional boundary; Set vertical boundaries based on the actual height requirements of high-altitude operations; For irregular areas, import pre-defined boundary files to obtain key areas.

7. The intelligent positioning analysis method for safety tools according to claim 6, characterized in that: The specific steps of triggering an early warning when the monitoring platform detects that the three-dimensional position data exceeds the three-dimensional electronic fence include: Obtain the two-dimensional coordinate data x and y of the aerial worker; Get a list of all vertex coordinates in the key area; Select the horizontal right direction as the ray direction and set the intersection counter; For each edge of the polygon, check if it intersects the ray. If so, update the intersection counter based on the exact location of the intersection: if the intersection is strictly above the ray, ignore it; if the intersection falls exactly at the starting point of the ray, the point is considered to be on the boundary. In other cases, increase the counter value; If the count value is odd, it means the point is inside the critical area; if it is even, the point is outside the critical area; If the point is within the critical area, the height z of the point is further checked to see if it meets the height limit of the geo-fence. If it does not meet the height limit, an early warning is triggered.

8. The intelligent positioning and analysis method for safety tools according to claim 7, characterized in that: After the warning is triggered, the specific steps of obtaining the video data and voice data of the aerial work site in the corresponding time period, analyzing the personnel posture information in the video data, and identifying the abnormal type in combination with the voice keywords include: Extract the video and audio data of the aerial work site in the corresponding time period according to the timestamp of the warning trigger; Define the normal posture range according to the standard operating procedures for working at heights; Use human posture estimation algorithms to detect the posture information of people in aerial work site video data; Compare the current posture with the normal posture range to detect abnormal behavior; Use automatic speech recognition technology to convert speech signals into text, define a set of keywords related to anomalies, and search for these keywords in the converted text to determine whether there are any anomalies. Generate corresponding anomaly types based on detected abnormal behaviors and abnormal conditions.

9. The intelligent positioning and analysis method for safety tools according to claim 8, characterized in that: The specific steps of using the human posture estimation algorithm to detect the posture information of people in the aerial work site video data include: Obtain video streams or recordings from cameras at aerial work sites and decompose the video into a series of single-frame images; Use Faster R-CNN to locate the people in each frame and obtain the people data; The OpenPose algorithm is applied based on personnel data to identify the key points of each human body and obtain posture information.

10. An intelligent positioning and analysis system for safety tools, characterized in that: An intelligent positioning and analysis method for a safety tool is adopted according to any one of claims 1 to 9.

Images