A ranging module guided point cloud clustering identity recognition and risk prevention method

Abstract

This invention relates to the field of power equipment maintenance technology and discloses an identity recognition and risk prevention method for point cloud clustering guided by a ranging module. This method integrates a high-precision ranging and positioning network with the three-dimensional lidar point cloud perception depth, uses the high-precision coordinates provided by the ranging module as spatial prior anchor points to guide the lidar to perform focused local clustering, binds the positioning identity information with the point cloud clustering results, and combines motion trend prediction algorithms to achieve forward-looking early warning. This effectively solves the problems of target identity ambiguity, redundant global point cloud calculation, and passive safety protection lag in traditional lidar solutions.

Description

Technical Field

[0001] This invention relates to the field of power equipment maintenance technology, specifically to a method for identity recognition and risk prevention in point cloud clustering guided by a ranging module. Background Technology

[0002] When conducting inspections, maintenance, or large-scale machinery hoisting operations at high-voltage locations such as substations, workers and vehicles are prone to unintentionally approaching live equipment, which could lead to serious safety accidents such as electric arc shock or equipment short circuits.

[0003] Traditional safety management methods rely on manual monitoring, signage warnings, and insulation isolation, which suffer from problems such as strong reliance on experience, insufficient monitoring continuity, and poor data traceability, making it difficult to meet the high-precision real-time risk prevention and control needs of modern power scenarios. Although the industry has introduced 3D LiDAR technology to achieve target positioning and boundary crossing alarms through scene point cloud acquisition and cluster analysis, it still faces two major technical bottlenecks: First, the processing mode based on global point cloud data requires segmentation and clustering of massive environmental information, resulting in high algorithm complexity and poor real-time performance, which is prone to missed detections and false detections in multi-target dynamic scenarios; Second, existing solutions can only achieve general alarms when a target approaches, lacking the ability to bind point cloud features to the identity of specific operating individuals (personnel / vehicles), making it impossible for the system to accurately determine the boundary crossing subject and failing to support the refined safety management needs of assigning responsibility to individuals.

[0004] The aforementioned shortcomings make it difficult for existing technology systems to effectively build proactive defense mechanisms under complex operating conditions. Therefore, it is urgent to make breakthroughs in key technologies such as multimodal perception fusion and target identity association recognition in order to achieve intelligent early warning and precise prevention and control of near-electric risks. Summary of the Invention

[0005] (a) Technical problems to be solved To address the shortcomings of existing technologies, this invention provides a method for identity recognition and risk prevention through point cloud clustering guided by a ranging module. This method has the advantages of multi-source heterogeneous data collaborative perception, target identity-space dual-dimensional association, lightweight point cloud intelligent analysis, and advanced risk situation prediction. It solves the problems of target identity ambiguity, redundant global point cloud calculation, and lag in passive security protection that exist in traditional LiDAR solutions.

[0006] (II) Technical Solution To achieve the above objectives, the present invention provides the following technical solution: a method for identity recognition and risk prevention in point cloud clustering guided by a ranging module, comprising the following steps: S1. Deploy a ranging device containing a three-dimensional lidar and a single built-in ranging module at the substation work site, and deploy three external ranging modules around the work area. Place a high reflectivity target at the geometric center of the outer shell of the ranging module. At the same time, equip the workers or special vehicles with identification tags, and the tags have built-in ranging modules. S2. Activate the 3D lidar within the ranging device to scan the work scene and acquire global 3D point cloud data containing the high-reflectivity target; based on a preset reflection intensity threshold, segment the point cloud clusters of the ranging module from the global 3D point cloud, and obtain the coordinates of the external ranging module in the lidar coordinate system. Based on the hardware structure size parameters, the offset of the antenna phase center of the ranging module in the lidar coordinate system is converted into coordinates. ; S3. A ranging module that acquires known coordinates in real time, and a ranging module based on a tag worn on the target, use these four distance measurements to perform initial calculations using the principle of polygonal positioning, obtaining a rough three-dimensional spatial coordinate of the target tag as its initial position estimate. ; S4. Due to the inherent error in distance measurement, the accuracy of the initial calculated coordinates is limited. Therefore, it is necessary to adjust the initial coordinates... Iterative optimization is performed to obtain high-precision positioning; S5. High-precision coordinates using a tag ranging module. Using spatial prior anchor points, the real-time point cloud of the lidar is clustered to obtain the point cloud of the monitored object. ; S6. Based on the charged body point cloud selected by the operator in the scene point cloud. ,calculate and The shortest Euclidean distance; S7. When within a safe distance, the point closest to the charged body in the previous frame is used as the reference. The point closest to the charged body in the current frame For the j-th monitored object, make a prediction to determine whether there is a risk of near-electric shock.

[0007] Preferably, the three external ranging modules in S1 include a first external base station, a second external base station, and a third external base station. The first external base station is fixedly installed using a high-stability bracket deployed in the upper left corner of the site; the second external base station is fixedly installed using an insulating support rod deployed in the lower right corner of the site; and the third external base station is temporarily set up and quickly calibrated using a mobile tripod deployed in front of the site.

[0008] Preferably, S1 includes: embedding the ranging module in a lightweight tag shell and fixing it to the top of the helmet or a conspicuous position on the vehicle body, as a tag ranging module.

[0009] Preferably, the ranging device in S2 is an integrated safety early warning host, which has a three-dimensional laser radar and a built-in ranging module internally connected. The external ranging module is an independently deployed auxiliary base station, which maintains time consistency with the ranging device through a wired or wireless clock synchronization mechanism, together forming a high-precision ranging and positioning network.

[0010] Preferably, step S2 involves obtaining the coordinates of the external ranging modules in the lidar coordinate system by extracting the centroid coordinates of the point cloud of the high-reflectivity target at the center of the outer shell of each of the three external ranging modules, which are then used as the three-dimensional coordinates of the first, second, and third external base stations in the lidar coordinate system. .

[0011] Preferably, the initial coordinates in S4 The iterative optimization process is as follows: S4.1, Let the first... The known coordinates of the external ranging modules are: The actual measured distance between the tag ranging module and the external ranging module is Establish the following system of nonlinear equations: S4.2. Linearize the nonlinear equation system using the first-order Taylor series of multivariable functions, i.e.: in, The theoretical true distance from the actual coordinates to the base station, i.e., the measured distance. ; This represents the coordinates of the current iteration. And substitute it into the equation to calculate the distance; S4.3, the initial estimated value Set as the initial value for iteration Taylor series expansion method is introduced for nonlinear optimization and refinement; let the observation point be... The Jacobian matrix of the distance observation equation is The position increment is updated according to the following iterative criteria. : in, The residual vector represents the difference between the measured distance and the theoretical distance calculated based on the current iterative coordinates.

[0012] Preferably, in S4.3, when the position increment... Less than the convergence threshold The iteration terminates when the time is reached, and the high-precision spatial coordinates of the j-th tag ranging module in the global three-dimensional coordinate system are output. ; The expression for the Jacobian matrix A is as follows: in, , representing the coordinates based on the current iteration Calculate the theoretical spatial distance between the obtained tag and the i-th external ranging module; When position increment Greater than or equal to the convergence threshold Automatically update the current coordinate estimate: And order The updated coordinates are then substituted into the recalculated Jacobian matrix. With residual vector Then, proceed to the next iteration.

[0013] Preferably, in S5, with Centered on, with a preset radius Within a neighborhood, all laser points within the spatial sphere are extracted from the real-time 3D point cloud. Then, Euclidean distance clustering or region growing clustering is performed on the extracted local point cloud. After removing outlier noise points, the largest connected cluster is taken as the point cloud cluster corresponding to the j-th monitoring object. .

[0014] Preferably, in step S6, the calculation and The shortest Euclidean distance is calculated using the following formula: in, Point cloud of the j-th monitoring object , Point cloud of charged body The point in the middle, and determine whether it is less than the safe distance. ,when At that time, the point closest to the charged body in the k-th frame of the current frame is automatically recorded as... And trigger the danger crossing alarm for the j-th monitored object.

[0015] Preferably, the process for S7 to predict whether the monitored object has a near-electricity hazard is as follows: S7.1, Predict the velocity of the j-th monitored object in frame k+1 as follows: in, The time interval between two frames; S7.2, Predict the position of the j-th monitored object in frame k+1 as follows: S7.3, Calculation Point cloud of charged body closest distance And determine whether it is less than the safe distance. ,when When the predicted over-limit alarm for the j-th monitored object is triggered, the alarm will be triggered.

[0016] Compared with existing technologies, this invention provides a method for identity recognition and risk prevention in point cloud clustering guided by a ranging module, which has the following beneficial effects: 1. This invention installs high-reflectivity targets on external ranging modules (base stations), enabling lidar to identify their precise locations and establish a unified spatial coordinate system between the ranging network and the point cloud. Simultaneously, by equipping each worker or vehicle with a unique identification tag containing a built-in ranging module, and utilizing the ranging network for continuous high-precision positioning and identification, the identity represented by these positioning coordinates is fused and correlated with the target point cloud obtained by lidar scanning and clustering around those coordinates. This eliminates the traditional approach that only identifies "a target is approaching" but cannot distinguish "who is approaching," completely solving the problem of point cloud data identification. This allows the method to accurately determine who has crossed the boundary in complex scenarios involving multiple people and vehicles, achieving refined safety management by alerting whoever has crossed the boundary. Ultimately, this significantly improves the targeting of alarms and the traceability of on-site control.

[0017] 2. This invention utilizes the target label coordinates provided by a high-precision ranging module as spatial prior anchor points, guiding the point cloud processing algorithm of the lidar to perform focused local clustering only in the neighborhood of this anchor point. This transforms blind search into precise guidance, significantly reducing computational complexity from the scale of the entire scene to the scale of a fixed neighborhood. While ensuring clustering accuracy, this architecture avoids redundant calculations at the source, greatly reducing computational load and processing latency, and ultimately significantly improving the real-time response capability of the system. This ensures that the method of this invention can still operate stably and efficiently in the complex dynamic environment of substations with high point cloud density and a large number of targets, effectively reducing the risk of missed detections and false detections under complex operating conditions. It fundamentally changes the inefficient computation mode of traditional solutions that require indiscriminate global clustering of massive point clouds across the entire scene.

[0018] 3. This invention introduces intelligent optimization and decision-making mechanisms at both the distance measurement data processing and distance judgment levels. On the one hand, a nonlinear iterative optimization algorithm based on Taylor expansion is used for the initial positioning coordinates obtained by the distance measurement network. By constructing the Jacobian matrix and performing Gaussian-Newton refinement, the distance measurement noise and multipath interference can be effectively suppressed, improving the positioning accuracy from the meter level to the centimeter level or even the millimeter level, providing highly reliable input data for subsequent clustering and judgment. On the other hand, based on the real-time shortest Euclidean distance calculation, a velocity extrapolation prediction algorithm based on historical positions is further introduced to dynamically predict the target's trajectory at the next moment and its predicted distance to the charged body. When the predicted distance triggers the safety threshold, the system can issue an advance warning before the actual boundary crossing occurs, upgrading the safety protection mechanism from post-event alarm to pre-event prevention, thereby gaining valuable intervention time for on-site safety supervisors and fundamentally improving the proactive safety protection capability of substation electrical operations. Attached Figure Description

[0019] Figure 1 This is a flowchart illustrating the execution process of the method of the present invention. Detailed Implementation

[0020] 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.

[0021] Please see Figure 1 An identity recognition and risk prevention method guided by a ranging module for point cloud clustering is proposed. This method integrates a high-precision ranging and positioning network with the depth perception of a 3D LiDAR point cloud. Using the high-precision coordinates provided by the ranging module as spatial prior anchor points, it guides the LiDAR to perform focused local clustering. Simultaneously, it binds positioning identity information with the point cloud clustering results and combines this with a motion trend prediction algorithm to achieve proactive early warning. This effectively addresses the shortcomings of existing technologies in terms of identity recognition, computational efficiency, and early warning timeliness. The method includes the following steps: S1. Deploy a ranging device at the substation work site, including a 3D lidar and a single built-in ranging module. Set up three external ranging modules around the work area (including a first external base station, a second external base station, and a third external base station; the first external base station is fixedly installed using a high-stability bracket deployed in the upper left corner of the site; the second external base station is fixedly installed using an insulating support rod deployed in the lower right corner of the site; and the third external base station is temporarily erected and quickly calibrated using a mobile tripod deployed directly in front of the site). Place a high-reflectivity target at the geometric center of the ranging module's outer shell. Simultaneously, equip workers or special vehicles with identification tags, embedding the ranging module within a lightweight tag shell and fixing it to the top of a safety helmet or a conspicuous location on the vehicle body, serving as a tag ranging module. By equipping each target with a uniquely coded tag ranging module, embedding a high-reflectivity target in the center of the external base station shell, and fixing the 3D lidar and the built-in ranging module to the same ranging device base, a joint perception foundation of "unified spatial coordinate reference and binding of identity and point cloud" is established, providing a hardware prerequisite for subsequent guided clustering and accurate alarm. S2. Activate the ranging device (this ranging device is an integrated safety early warning host, internally connected to a 3D LiDAR and a built-in ranging module; the external ranging module is an independently deployed auxiliary base station, maintaining time consistency with the ranging device through a wired or wireless clock synchronization mechanism, together forming a high-precision ranging and positioning network). The 3D LiDAR within the device scans the operational scene to acquire global 3D point cloud data containing highly reflective targets. Based on a preset reflection intensity threshold, the point cloud clusters of the ranging modules are segmented from the global 3D point cloud, and the coordinates of the external ranging modules in the LiDAR coordinate system are obtained (the centroid coordinates of the point cloud of the high-reflectivity targets at the center of the shells of the three external ranging modules are extracted respectively, serving as the 3D coordinates of the first, second, and third external base stations in the LiDAR coordinate system). Based on the hardware structure size parameters, the offset of the antenna phase center of the ranging module in the lidar coordinate system is converted into coordinates. ; The advantages are: by extracting the centroid of the high-reflectivity target to complete the external base station coordinate calibration, by using the known installation offset to calculate the coordinates of the built-in module, and by incorporating the four known coordinate ranging modules into a unified spatial reference, the purpose of providing four stable anchor point coordinates for multilateral positioning calculation is achieved, ensuring that subsequent tag positioning has a fixed reference that can be referenced. S3, a distance measurement module that acquires known coordinates in real time ( The corresponding module and the tag ranging module worn on the target (person / vehicle), based on these four distance measurements, use the principle of polygonal positioning to perform initial calculations to obtain a rough three-dimensional spatial coordinate of the target tag, which serves as its initial position estimate. ; The advantages are: by combining the four known coordinate anchor points with the measured distance of the tag ranging module into a spherical intersection equation system, using the least squares closed-form solution method to perform a one-time coordinate calculation, and using the closed-form solution as the initial value for subsequent nonlinear iterations, the goal of obtaining the approximate global position with a low computational cost is achieved, providing a reliable iterative starting point for high-precision refinement. S4. Due to the inherent error in distance measurement, the accuracy of the initial calculated coordinates is limited. Therefore, it is necessary to adjust the initial coordinates... Iterative optimization is performed to obtain high-precision positioning. The specific process is as follows: S4.1, Let the first... The known coordinates of the external ranging modules are: The actual measured distance between the tag ranging module and the external ranging module is Establish the following system of nonlinear equations: S4.2. Linearize the nonlinear equation system (the aforementioned nonlinear functions) using the first-order Taylor series of multivariable functions, i.e.: in, The theoretical true distance from the actual coordinates to the base station, i.e., the measured distance. ; This represents the coordinates of the current iteration. And substitute it into the equation to calculate the distance; S4.3, the initial estimated value Set as the initial value for iteration Taylor series expansion method is introduced for nonlinear optimization and refinement; let the observation point be... The Jacobian matrix of the distance observation equation is The position increment is updated according to the following iterative criteria. : in, The residual vector representing the measured distance and the theoretical distance calculated based on the current iteration coordinates; When position increment Less than the convergence threshold The iteration terminates when the time is reached, and the high-precision spatial coordinates of the j-th tag ranging module in the global three-dimensional coordinate system are output. ; The expression for the Jacobian matrix A is as follows: in, , representing the coordinates based on the current iteration Calculate the theoretical spatial distance between the obtained tag and the i-th external ranging module; When the position increment is greater than or equal to the convergence threshold , the current coordinate estimated value is automatically updated: and let , and substitute the updated coordinates into the recalculation of the Jacobian matrix and the residual vector , then proceed to the next iteration; The advantages are: by introducing Taylor expansion to linearize the nonlinear distance equations, constructing a Jacobian matrix for Gauss-Newton iteration refinement, and setting a convergence threshold to control the iteration termination timing, the effect of improving the positioning result with initial meter-level accuracy to centimeter-level or even millimeter-level high precision is achieved, which effectively suppresses the influence of ranging noise and multipath interference on position estimation, and provides sufficiently accurate spatial prior anchor points for the subsequent "position first then cluster" architecture; S5. Using the high-precision coordinates of the tag ranging module as spatial prior anchor points, cluster the real-time point cloud of LiDAR to obtain the point cloud of the monitored object ; Specifically, with as the center and the preset radius as the neighborhood range, all laser points in the spatial sphere are extracted from the real-time three-dimensional point cloud, then Euclidean distance clustering or region growing clustering is performed on the extracted local point cloud, after eliminating outlier noise points, the maximum connected cluster is taken as the point cloud cluster corresponding to the j-th monitored object ; The advantages are: by performing spatial neighborhood search with the high-precision positioning coordinate as the sphere center, narrowing the clustering range to the target local area, and performing lightweight clustering and noise filtering on the local point cloud, the effect of reducing the computational complexity from O(NlogN) of the full scene to O(MlogM) (M<<N) of the fixed neighborhood is achieved, which fundamentally avoids the redundant calculation of global clustering, and ensures that the system still maintains high real-time performance and low-latency response in the complex substation environment; S6. According to the live body point cloud selected by the operator in the scene point cloud , calculate and the shortest Euclidean distance, that is ; Wherein, is the j-th monitored object point cloud , is the point in the live body point cloud , and judge whether it is less than the safety distance , when , the point closest to the live body in the current k-th frame is automatically recorded as And trigger the danger crossing alarm for the j-th monitored object; The advantages are: by quickly searching for nearest neighbor pairs based on point cloud spatial index, comparing the minimum distance with a preset safety threshold, and recording the location of the nearest point in real time for subsequent motion trend analysis, the effect of instantly binding and alarming targets that have crossed the boundary is achieved, which solves the pain point of traditional methods that "only report that an object has crossed the boundary, but do not know who crossed the boundary"; S7. When within a safe distance, the point closest to the charged body in the previous frame is used as the reference. The point closest to the charged body in the current frame For the j-th monitored object, a prediction is made to determine whether there is a risk of electric shock. Specifically: S7.1, Predict the velocity of the j-th monitored object in frame k+1 as follows: in, The time interval between two frames is given by the velocity vector, which is obtained by dividing the position difference by the time interval. The direction represents the direction of motion, and the magnitude represents the speed. S7.2, Predict the position of the j-th monitored object in frame k+1 as follows: S7.3, Calculation Point cloud of charged body closest distance And determine whether it is less than the safe distance. ,when When the prediction of the j-th monitored object is triggered, the prediction assumes that the target will maintain a uniform linear motion for a very short time and is suitable for short-term trend estimation under normal operating speed. The advantages are: by estimating the instantaneous velocity vector based on the nearest point position in historical adjacent frames, extrapolating the predicted nearest point between the target and the charged body in the next frame, and making an advance judgment by comparing the predicted distance with the safety threshold, the early warning mechanism is upgraded from "real-time alarm" to "proactive early warning". It provides an early intervention window of several seconds to tens of seconds before the target actually enters the danger zone, giving on-site safety supervisors valuable time to take emergency avoidance measures. At the same time, when the predicted distance is greater than or equal to the safety threshold, the system does not trigger an early warning and continues to iterate and update the historical position sequence, continuously tracking the target's movement state to ensure the continuity and stability of the early warning decision. In summary, this invention establishes a joint calibration benchmark for "target-tag" in step S1, completes base station coordinate self-calibration in step S2, achieves high-precision positioning refinement in steps S3-S4, guides local clustering with the positioning results in step S5, completes real-time boundary judgment and alarm in step S6, and achieves forward-looking early warning based on motion trends in step S7, forming a complete closed loop from identity binding → precise positioning → guided clustering → real-time alarm → forward-looking prevention. Among them, the nonlinear iterative optimization in step S4 and the local guided clustering in step S5 constitute the dual core of this method in terms of computational accuracy and efficiency. Steps S6 and S7 together constitute an intelligent security protection system from passive response to active prevention.

[0022] The method of the present invention is applied to the following embodiments: Example 1 (Near-Electricity Risk Management in a Multi-Worker Collaborative Inspection Scenario within a Substation) In the equipment area of ​​a 220kV substation, routine infrared thermography and partial discharge detection are planned for the energized circuit breaker bays. A total of 4 personnel are on site, responsible for equipment operation, data recording, safety monitoring and auxiliary lighting respectively. There are multiple energized objects such as 220kV busbars, circuit breakers and disconnect switches around the work area. The safety distance threshold is set at 3.5m.

[0023] On-site deployment: Following step S1, install an integrated ranging device (including a 3D lidar and a built-in ranging module) 15m directly in front of the work area. Fix the first external base station on a high-stability bracket in the upper left corner of the site, fix the second external base station on an insulated support rod in the lower right corner, and temporarily set up the third external base station on a mobile tripod directly in front. High-reflectivity targets are affixed to the geometric center of the outer shell of each of the three base stations. Four workers wear lightweight identification tags numbered #01, #02, #03, and #04 respectively. The ranging module inside the tag is fixed to the top of the safety helmet, ensuring that the antenna faces upwards without obstruction.

[0024] Operation process and early warning performance: After the system is started, the 3D lidar first-frame scan acquires the global point cloud, and the external base station coordinates are calibrated by extracting the centroids of three high-reflectivity targets. ), combined with the coordinates of the fixed module inside the device A four-anchor-point positioning network was then established. During two hours of continuous operation, when worker #03 (auxiliary lighting worker) approached the B-phase bushing of the circuit breaker bay, the tag positioning achieved centimeter-level accuracy (positioning error ≤ 4cm) after nonlinear iterative optimization in steps S3-S4. Step S5 uses its coordinates as the center... =1.2m as the neighborhood for guided local clustering, accurately segmenting the head and shoulder point cloud cluster corresponding to person #03; step S6 calculates in real time the shortest Euclidean distance between this point cloud cluster and the pre-selected live body point cloud (circuit breaker B-phase bushing and adjacent busbar), when When the distance drops to 2.8m (less than the 3.5m threshold), the system immediately triggers the identity binding alarm corresponding to tag #03 and records the nearest point. Subsequently, based on the historical frame differential prediction in step S7, the system predicted the next movement trend (velocity direction pointing towards the bushing at 0.8 m / s), anticipating that the distance would further decrease to 2.1 m in the next frame. It then issued a predicted boundary crossing alarm. The safety supervisor warned personnel #03 to stop approaching 2.3 seconds in advance via voice broadcast, successfully avoiding a potential electric shock risk. During the operation, the other three personnel did not trigger any alarms, demonstrating the system's precise control of "whoever crosses the boundary, gets an alarm."

[0025] Example 2 (Near-Electrical Safety Monitoring of Vehicles and Crane Booms During Large Crane Lifting Operations) At the construction site of a 500kV substation expansion project, a 50t truck crane was required to hoist and position GIS equipment under a 500kV energized utility tunnel busbar. The work area was located above a 500kV energized utility tunnel busbar (500kV to ground, with a safety distance threshold of 8.5m). The crane's operating radius was 18m, and the boom angle ranged from 65° to 85°. This posed a significant risk that the boom and the hoisted equipment would not be within the safe distance below the busbar during rotation.

[0026] On-site deployment: Following step S1, the integrated ranging device is erected at a high observation point (4m high) 20m to the side of the work area to obtain an unobstructed scanning field of view of the top of the boom. The first external base station is deployed on a concrete foundation pier in the northwest corner of the site, the second external base station is deployed on a precast concrete block in the southeast corner, and the third external base station is temporarily erected on an insulated platform facing due north. High-reflectivity targets are installed at the center of the outer shell of each of the three base stations. Two identification tags (numbered #) are installed on the top of the crane cab and at the auxiliary hook position at the top of the boom. and# Each tag has a built-in distance measurement module.

[0027] Operation Process and Early Warning Performance: After completing step S2 (base station coordinate calibration), the ranging and positioning network is put into operation. During the crane lifting process, the 3D LiDAR scans the scene point cloud in real time. Steps S3-S4 refine the high-precision positioning of the #CRANE_BOOM tag (horizontal positioning error ≤ 5cm, elevation error ≤ 8cm). Step S5 uses this high-precision coordinate as an anchor point for guided local clustering, successfully separating the point cloud cluster at the top of the boom from surrounding scaffolding, wire ropes, and other interfering objects. Step S6 calculates the shortest distance between the top of the boom and the point cloud of the pre-selected energized body (500kV pipe gallery busbar). When the boom rotates back to the busbar side to 9.2m, steps S7.1-S7.3 extrapolate its trajectory based on three consecutive frames of historical positions, predicting that the top of the boom will enter a position 7.8m below the busbar in the next frame (less than the 8.5m threshold). The system issues a predicted line crossing alarm and simultaneously pushes the alarm information to the crane operator's cab display and the on-site safety director's handheld terminal. Upon receiving the alarm, the crane operator immediately stopped the slewing and fine-tuned the boom angle. The top of the boom eventually stabilized at 9.5m, without entering the actual danger zone. The entire lifting operation lasted 40 minutes. The system triggered three predicted boundary crossing alarms, all of which were effective warnings. There were no false alarms or missed alarms, ensuring the safety and controllability of the entire lifting operation.

[0028] This embodiment fully verifies the effectiveness and high reliability of the method of the present invention on complex monitoring objects such as large mobile machinery and long boom structures.

[0029] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for identity recognition and risk prevention in point cloud clustering guided by a ranging module, characterized in that, Includes the following steps: S1. Deploy a ranging device containing a three-dimensional lidar and a single built-in ranging module at the substation work site, and deploy three external ranging modules around the work area. Place a high reflectivity target at the geometric center of the outer shell of the ranging module. At the same time, equip the workers or special vehicles with identification tags, and the tags have built-in ranging modules. S2. Activate the 3D lidar within the ranging device to scan the work scene and acquire global 3D point cloud data containing the high-reflectivity target; based on a preset reflection intensity threshold, segment the point cloud clusters of the ranging module from the global 3D point cloud, and obtain the coordinates of the external ranging module in the lidar coordinate system. Based on the hardware structure size parameters, the offset of the antenna phase center of the ranging module in the lidar coordinate system is converted into coordinates. ; S3. A ranging module that acquires known coordinates in real time, and a ranging module based on a tag worn on the target, use these four distance measurements to perform initial calculations using the principle of polygonal positioning, obtaining a rough three-dimensional spatial coordinate of the target tag as its initial position estimate. ; S4. Due to the inherent error in distance measurement, the accuracy of the initial calculated coordinates is limited. Therefore, it is necessary to adjust the initial coordinates... Iterative optimization is performed to obtain high-precision positioning; S5. High-precision coordinates using a tag ranging module. Using spatial prior anchor points, the real-time point cloud of the lidar is clustered to obtain the point cloud of the monitored object. ; S6. Based on the charged body point cloud selected by the operator in the scene point cloud. ,calculate and The shortest Euclidean distance; S7. When within a safe distance, the point closest to the charged body in the previous frame is used as the reference. The point closest to the charged body in the current frame For the j-th monitored object, make a prediction to determine whether there is a risk of near-electric shock.

2. The method for identity recognition and risk prevention in point cloud clustering guided by a ranging module according to claim 1, characterized in that, The three external ranging modules in S1 include a first external base station, a second external base station, and a third external base station. The first external base station is fixedly installed using a high-stability bracket deployed in the upper left corner of the site; the second external base station is fixedly installed using an insulating support rod deployed in the lower right corner of the site; and the third external base station is temporarily set up and quickly calibrated using a mobile tripod deployed in front of the site.

3. The method for identity recognition and risk prevention in point cloud clustering guided by a ranging module according to claim 1, characterized in that, S1 includes: embedding the ranging module in a lightweight tag shell and fixing it to the top of the safety helmet or a conspicuous position on the vehicle body, as a tag ranging module.

4. The method for identity recognition and risk prevention in point cloud clustering guided by a ranging module according to claim 1, characterized in that, The ranging device in S2 is an integrated safety early warning host, which has a three-dimensional laser radar and a built-in ranging module internally. The external ranging module is an independently deployed auxiliary base station, which maintains time consistency with the ranging device through a wired or wireless clock synchronization mechanism, together forming a high-precision ranging and positioning network.

5. The method for identity recognition and risk prevention in point cloud clustering guided by a ranging module according to claim 1, characterized in that, S2 obtains the coordinates of the external ranging modules in the lidar coordinate system: extract the centroid coordinates of the point cloud of the high-reflectivity target at the center of the shell of each of the three external ranging modules, and use them as the three-dimensional coordinates of the first, second, and third external base stations in the lidar coordinate system. .

6. The method for identity recognition and risk prevention in point cloud clustering guided by a ranging module according to claim 1, characterized in that, The initial coordinates in S4 The iterative optimization process is as follows: S4.1, Let the first... The known coordinates of the external ranging modules are: The actual measured distance between the tag ranging module and the external ranging module is Establish the following system of nonlinear equations: S4.

2. Linearize the nonlinear equation system using the first-order Taylor series of multivariable functions, i.e.: in, The theoretical true distance from the actual coordinates to the base station, i.e., the measured distance. ; This represents the coordinates of the current iteration. And substitute it into the equation to calculate the distance; S4.3, the initial estimated value Set as the initial value for iteration Taylor series expansion method is introduced for nonlinear optimization and refinement; let the observation point be... The Jacobian matrix of the distance observation equation is The position increment is updated according to the following iterative criteria. : in, The residual vector represents the difference between the measured distance and the theoretical distance calculated based on the current iterative coordinates.

7. The method for identity recognition and risk prevention in point cloud clustering guided by a ranging module according to claim 6, characterized in that, In S4.3, when the position increment Less than the convergence threshold The iteration terminates when the time is reached, and the high-precision spatial coordinates of the j-th tag ranging module in the global three-dimensional coordinate system are output. ; The expression for the Jacobian matrix A is as follows: in, , representing the coordinates based on the current iteration Calculate the theoretical spatial distance between the obtained tag and the i-th external ranging module; When position increment Greater than or equal to the convergence threshold Automatically update the current coordinate estimate: And order The updated coordinates are then substituted into the recalculated Jacobian matrix. With residual vector Then, proceed to the next iteration.

8. The method for identity recognition and risk prevention in point cloud clustering guided by a ranging module according to claim 1, characterized in that, In S5, with Centered on, with a preset radius Within a neighborhood, all laser points within the spatial sphere are extracted from the real-time 3D point cloud. Then, Euclidean distance clustering or region growing clustering is performed on the extracted local point cloud. After removing outlier noise points, the largest connected cluster is taken as the point cloud cluster corresponding to the j-th monitoring object. .

9. The method for identity recognition and risk prevention in point cloud clustering guided by a ranging module according to claim 1, characterized in that, In S6, the calculation and The shortest Euclidean distance is calculated using the following formula: in, Point cloud of the j-th monitoring object , Point cloud of charged body The point in the middle, and determine whether it is less than the safe distance. ,when At that time, the point closest to the charged body in the k-th frame of the current frame is automatically recorded as... And trigger the danger crossing alarm for the j-th monitored object.

10. The method for identity recognition and risk prevention in point cloud clustering guided by a ranging module according to claim 1, characterized in that, The specific process for S7 to predict whether the monitored object has a near-electrical hazard is as follows: S7.1, Predict the velocity of the j-th monitored object in frame k+1 as follows: in, The time interval between two frames; S7.2, Predict the position of the j-th monitored object in frame k+1 as follows: S7.3, Calculation Point cloud of charged body closest distance And determine whether it is less than the safe distance. ,when When this happens, the predicted over-limit alarm for the j-th monitored object is triggered.

Images