A near electric warning method based on cylinder and intersection surface model

Abstract

The application relates to the field of safety monitoring of near-electricity construction machinery, and discloses a near-electricity warning method based on a cylindrical model and an intersection surface model, which adopts a ground support lightning and visual integrated monitoring and shooting device to collect image and point cloud data in real time at a near-electricity operation site, uses a coordinate system flatness and multi-feature method to extract power line point clouds under a complex background of a power transmission channel, and further establishes a power line directed boundary cylindrical model and an intersection surface model, removes the point cloud interference of the conductor itself through the directed boundary cylindrical model, and converts three-dimensional space ranging into two-dimensional plane projection calculation through the intersection surface model; a double-checking mechanism containing space distance preliminary screening and image visual rechecking is adopted to determine whether construction machinery intrusion warning occurs. The near-electricity intrusion monitoring of the construction machinery is realized, the ground slope and unstructured environment interference at the operation site are overcome, and the power line multi-model and the double-checking mechanism not only guarantee the ranging accuracy, but also improve the rapidity of the warning on the edge device with limited computing power.

Description

Technical Field

[0001] This invention relates to the field of safety monitoring of construction machinery near electric shock, specifically a method for near electric shock alarm based on a cylindrical and intersecting surface model. Background Technology

[0002] Large construction machinery such as cranes and pump trucks frequently pass through or operate within power transmission lines. Early warning systems for safe distances from live conductors are crucial for the safety of personnel and equipment. Construction machinery moves rapidly in complex construction sites, and if it inadvertently enters the safe clearance zone near power lines, accidents can easily occur. Therefore, analyzing the spatial relationship between construction machinery and power lines and issuing warnings for near-current intrusions is a critical requirement for power construction safety.

[0003] Existing near-charge safety distance early warning technologies can be categorized into two main types from a sensor perspective: One approach is based on proximity sensors: sensor probes are deployed at the mechanical or personnel end, and audible and visual alarms are triggered by electric field or capacitive coupling measurements. The advantage is the simplicity of the technical principle. However, it is susceptible to electric field distortion under different voltage levels, postures, and weather conditions, and requires mounting on the monitored object, making it unsuitable for multi-machine collaboration and temporary on-site operations. The second method is based on lidar, commonly involving installing lidar ranging devices on poles or conductors. For example, Zheng Binyan et al. applied for invention patent CN116014890A, titled "Electrical Safety Distance Alarm Device for Overhead Line Conductors Based on LiDAR Technology." This device uses fixed components and an adjustable pan-tilt unit to install radar on poles to monitor and alarm for electrical safety distances. However, ranging devices installed on poles or conductors cannot be flexibly deployed across different construction sites by construction teams, nor can they identify conductors and intrusions in three-dimensional space and analyze their spatial relationship.

[0004] In contrast, ground-mounted camera-radar integrated monitoring devices are convenient to deploy at near-electricity work sites, can monitor multiple construction machines, and can also handle various scenarios involving power transmission and substations, offering advantages in flexibility and adaptability. For images acquired from a third-view perspective and LiDAR point clouds, a common approach is to identify live wires and construction machinery separately through 3D target detection, thereby determining in three-dimensional space whether a near-electricity alarm is needed. However, because construction scenarios include wires and various construction machines, the scene structure is complex and unstructured, making it difficult for edge devices to handle the complex computations such as 3D target detection of construction machinery within the point cloud. Therefore, the 3D target detection solution that integrates vision and LiDAR is unsuitable for the high reliability and low latency requirements of near-electricity safety distance early warning at construction sites.

[0005] To quickly measure the near-electrical distance of intruders while simultaneously overcoming interference, separate power line models need to be established for different purposes. Although similar multi-model-based power line reconstruction methods exist, they are only used to improve the 3D reconstruction of power lines, not to solve the problem of rapid intruder distance measurement. For example, deep learning-based 3D cable reconstruction methods (see Zhang et al., "3D Reconstruction Method of Flexible Cable Model Based on Point Cloud," Automation Technology and Application, 2024), or the multi-model power line segmentation method in invention patent CN110533075A, proposes two segmentation processes based on XOY projection overlap and non-overlap, recovering missing segments through local statistics, voltage levels, and elevation sorting, solving the segmentation problem caused by multiple missing conductors and projection overlap in large-scale point clouds. These similar achievements only solve the problem of conductor fitting and reconstruction in point clouds and are not suitable for quickly measuring the near-electrical distance of intruders while simultaneously overcoming interference. Some existing technologies, such as the patent disclosed in publication number CN119380283A, are based entirely on 2D images for rotating target detection and image distance calculation. The data source is completely different, with no point cloud, relying entirely on two-dimensional scene images. Therefore, compared with using LiDAR point clouds, this method cannot accurately measure safe distances in three-dimensional space. Another example is the patent disclosed in publication number CN121171008A, which is a multi-sensor probability prediction method. It not only uses radar and binoculars but also relies on electric field sensors. Its method calculates the intrusion probability by cross-sensor target association and trajectory risk prediction, rather than based on measuring the net distance between the construction machinery and the specific conductor segment for early warning. Finally, the patent disclosed in publication number CN121353167A is a cylindrical cross-section detection method. The cylinder is used to retain internal points, and the projection surface is perpendicular to the axis. The cross-section of the conductor is observed to detect the deformation of its own attachments, rather than deleting the conductor's own points inside the cylinder to avoid interference from the conductor itself and to expose external intrusion points. At the same time, it does not consider combining multiple models to judge the near-electric intrusion of construction machinery.

[0006] Therefore, there is an urgent need for technical solutions to overcome ground slope interference, reduce algorithm complexity, and achieve rapid and accurate identification of intruders and differentiation of non-dangerous obstacles. Summary of the Invention

[0007] To address the shortcomings of existing technologies, this invention provides a proximity alarm method based on a cylinder and intersecting surface model to solve the aforementioned problems.

[0008] To achieve the above objectives, the present invention is implemented through the following technical solutions.

[0009] A proximity alarm method based on a cylinder and intersecting surface model includes: Step S1. Deploy a mobile ground-mounted camera-radar integrated monitoring device at the near-electricity work site to acquire lidar point cloud and video images containing power lines and construction machinery from a third-person perspective. After removing the ground point cloud, align the coordinate system and use a multi-feature method to segment and extract candidate clusters of power line point cloud. Step S2. For each candidate cluster of power line point cloud, by extracting the power line span and curvature features corresponding to the cluster, determine whether to use a single line segment model or a segmented straightening model to fit the power line. The single line segment model is used to provide axial constraints for the directed boundary cylinder, and the segmented straightening model is used to provide intersection line constraints for the intersection surface model. Step S3. Using the line segment model as the axis, calculate the maximum vertical distance from the electric field line point cloud to the axis to determine the minimum envelope radius, thereby establishing a directed boundary cylinder model. Further filter out the electric field line point cloud within the directed boundary cylinder model, so that a clear and complete intruder point cloud can be obtained during ranging. Step S4. For each line segment in the piecewise linearization model Construct a surface-intersecting model, which contains two orthogonal surfaces that intersect at... plane perpendicular to the ground plane And its normal vector and Direction vector Orthogonal, plane Perpendicular to the plane And its normal vector and Direction vector Orthogonal, using this intersecting surface model, any point outside the electric field lines can be directed to the plane. Projection to calculate horizontal distance To the plane Projection to calculate vertical distance ; Step S5. For each power line, a dual verification mechanism is used to determine whether construction machinery has intruded into the power line's safe zone, reducing false positives and false negatives. On one hand, the point closest to the surface of the directed boundary cylinder model of the power line is extracted, and on the other hand, the horizontal distance output by the intersection model is determined in real time. Vertical distance If the value is less than the threshold, it is considered a suspicious intruder. On the other hand, the location of the suspicious intruder in the image is checked by projecting the point cloud onto the image. An alarm is triggered only when the projected pixels of the intruder's point cloud are within the bounding box of the construction machinery image.

[0010] Furthermore, the specific method for power line point cloud segmentation and extraction in step S1 includes: By preprocessing the original point cloud with coordinate system alignment and ground separation, and then comprehensively utilizing the spatial connectivity and suspended physical characteristics of the power line point cloud, power lines are identified and extracted from complex non-ground point clouds. The coordinate system alignment preprocessing helps to overcome the interference caused by the sloped and not completely level ground at the work site on subsequent processing, and provides a unified discrimination benchmark for accurately screening power line point clouds based on suspended physical characteristics.

[0011] Furthermore, the specific method for determining which model to use for fitting based on the power line span and curvature characteristics in step S2 includes: After aligning the coordinate systems, determine whether the projection length of the point cloud on the horizontal plane (XOY plane) is greater than a preset length threshold. Furthermore, its fitted quadratic curve in the vertical plane (XOZ plane) The absolute value of the coefficient of the quadratic term Is it greater than the preset curvature threshold? If all the above conditions are met, then the candidate cluster is modeled using segmented linearization with straightening of bends; otherwise, a single line segment model is established. Finally, the generated segmented line segments are refined and collinearized based on gaps and angles to form the final set of each line segment and its corresponding point cloud cluster. For each line segment Assign a unique ID and store the direction vector endpoints This provides axis constraints for subsequent OBC models and intersection line constraints for surface-type models.

[0012] Furthermore, the specific methods for power line OBC model reconstruction and power line point cloud filtering in step S3 include: by Using the axis as the reference line, calculate the perpendicular distance from each point within the cluster to the axis. Take the radius To construct an OBC (Online Component Controller) for real-time point cloud ranging, and The relative size is used to determine the value: if If the points are identified as power line points, they are filtered out. Only non-power line points are retained to reduce subsequent computation and interference. The candidate intrusion point cloud after power line removal is output as the input prior for cross-surface ranging and dangerous area determination.

[0013] Furthermore, the specific method for reconstructing the power line intersection model in step S4 includes: Step S4-1. Using the first part Line segment direction vector With ground normal Using the base, construct pairwise orthogonal structures that both pass through the base. auxiliary plane :make Legal direction (with ground plane) (orthogonal) Legal direction (and plane) (orthogonal) Step S4-2. For any point First project onto the corresponding auxiliary plane. have to , in The reference point can be a line segment. For any point on the surface, the horizontal distance is... : , Then along Constructing a perpendicular line and intersection vertical distance : Step S4-3. For any point outside the directed boundary cylinder model of the electric field lines, calculate the horizontal and vertical distances between that point and each electric field line. ...If there are two power lines in this layer, calculate their spacing W to determine the region of the intruder relative to the conductor and calculate the closest distance.

[0014] Furthermore, the specific method for proximity intrusion alarm based on dual verification in step S5 includes: Step S5-1. Set the vertical and horizontal safety thresholds according to the voltage level. The system uses a set of hierarchical thresholds and a faceted model for preliminary judgment to perform real-time screening. For each candidate intrusion point cluster, the distance from its left edge is calculated. and the distance to the right sideline The system determines the spatial region based on distance: if only one horizontal distance exceeds the limit, it is classified as line gap region B; if both the left horizontal and vertical distances exceed the limits simultaneously, it is classified as left neighbor region A; if both the right horizontal and vertical distances exceed the limits simultaneously, it is classified as right neighbor region C, and the "candidate intrusion" flag is triggered according to the following rules: like or This triggers the "candidate intrusion" flag in region B; like and This triggers the "candidate intrusion" flag in region A; like and Triggering the "candidate intrusion" flag in region C. Step S5-2. Project the candidate intruder point cloud clusters onto the image and check whether they fall within the bounding box of the construction machinery target recognition. If not, remove them to reduce false alarms. Step S5-3. For each cluster of construction machinery points that has passed 2D target detection and spatial relation verification, calculate its distance to the set of line segments. minimum Euclidean distance ,Record Timestamp; and minimum Euclidean distance The distance is compared with the preset graded protection thresholds: when the distance is less than the first-level safety warning threshold, an audible and visual alarm is triggered to remind the workers to take precautions; when the distance continues to decrease and falls below the second-level emergency braking threshold, regardless of which area it is in, it is determined to be a high-risk intrusion, and a control signal is directly output to stop the construction machinery, so as to achieve accurate graded protection based on distance perception.

[0015] Compared to existing technologies, this invention discloses a proximity alarm method based on a cylinder and a surface-intersection model. ① First, at the near-electricity work site, a ground-mounted monitoring device collects lidar point clouds and video images containing power lines and construction machinery. A level coordinate system overcomes the potential for uneven ground due to slope, and a multi-feature method is used to extract the power line point cloud. Then, based on the power line morphology, an adaptive line segment model or a segmented linear model is constructed to provide axial and intersection constraints for the subsequently established directed boundary cylinder (OBC) and surface model. The OBC model filters out interference from the power line's own point cloud, and the surface model converts three-dimensional spatial ranging into two-dimensional planar projection calculation. Finally, a dual verification mechanism is used for each power line. The surface model determines spatial boundary violations, and the image target bounding box is used for visual verification. This dual verification ensures that alarms are triggered only for construction machinery intrusion. This invention is applicable to mobile, mounted, camera-radar integrated monitoring devices, achieving non-intrusive near-electricity intrusion monitoring of construction machinery. It overcomes interference from ground slope and unstructured environment at the work site, significantly improving the real-time performance and accuracy of alarms while ensuring ranging accuracy. ② By performing coordinate system alignment preprocessing and multi-feature segmentation on the point cloud, the problem of traditional algorithms based on the assumption of a horizontal plane failing due to uneven and sloping ground at construction sites is overcome. Simultaneously, based on the power line span and curvature characteristics, this invention adaptively selects a single line segment or a segmented straightening modeling strategy, which not only solves the nonlinear fitting problem caused by gravity sag in long-distance transmission lines, but also provides geometric constraints closely following the actual spatial orientation of the conductor for subsequent accurate distance measurement, ensuring the consistency of safety warning areas at different locations on the conductor. ③ By constructing a power line intersection model, this invention innovatively simplifies the complex calculation of the distance from a point cloud to a 3D curve into the calculation of the projected distance from a point to two orthogonal planes. Compared to methods for precisely locating the 3D position of construction machinery, this invention avoids the computationally burdensome approach of 3D target detection of point clouds, making it more suitable for edge devices with limited computing power. Combined with the efficient filtering of the power line's own point cloud using a directed boundary cylinder model, this invention avoids complex deep learning point cloud segmentation networks, greatly reducing algorithm complexity and ensuring real-time near-electric distance measurement of construction machinery on edge devices with limited computing power. ④ By introducing a dual verification mechanism of point cloud spatial distance and visual image projection, the shortcomings of relying solely on lidar ranging cannot distinguish between construction machinery and other non-dangerous obstacles (such as birds and trees), and relying solely on two-dimensional visual detection cannot obtain accurate distances. By back-projecting the "candidate intruders" identified in the point cloud space to the image space and verifying whether they fall within the two-dimensional bounding box of the construction machinery, the advantages of lidar-visual fusion are achieved, effectively eliminating false alarms from non-mechanical targets and significantly reducing the false alarm rate and missed alarm rate. ⑤ By combining geometric modeling optimization, lightweight computing, and multimodal perception, the contradiction between the complex environment of the power transmission channel construction site, the limited edge computing resources, and the high requirements for real-time safety early warning was systematically resolved. Ultimately, a flexible, non-intrusive, and highly accurate and real-time near-electric intrusion rapid alarm solution was realized, which effectively improved the safety protection level of near-electric work. Attached Figure Description

[0016] Figure 1 This is a schematic diagram of the overall process of the system and method; Figure 2 The graph shows the results of ground filtering and RANSAC plane fitting. Figure 3 Image showing the results of electric field line extraction for candidate segmentation and dangling behavior screening; Figure 4 These are schematic diagrams of the intersecting surface model and the multi-model output. Figure 5 This is a schematic diagram of the area near the danger zone for live-line work. Detailed Implementation

[0017] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0018] Example 1 The present invention provides a rapid alarm method for near-electric intrusion of construction machinery based on a directed boundary cylinder of power lines and an intersection model, comprising: A mobile, bracket-mounted camera-radar integrated monitoring device is deployed at the near-electrical work site to acquire lidar point cloud and video images, including those of the conductors and construction machinery, from a third-person perspective. Figure 2 After removing the ground point cloud shown, the problem of the ground being sloping and not completely horizontal is overcome by aligning the coordinate system, and a multi-feature method is used to segment and extract the power line point cloud.

[0019] Extracting candidate clusters of electric field lines based on octree connectivity and dangling conditions, such as... Figure 3 As shown, line segment models are generated first for each candidate cluster. For conductors with long spans or significant curvature, segmented straightening modeling is adopted to straighten the bends. Axis constraints and intersection constraints are provided for the subsequent directed boundary cylinder (OBC) and intersection surface models, respectively.

[0020] An OBC model is constructed based on the line segment axis. A radius threshold constraint is used to retain neighboring points of the conductor and filter out power line point clouds, achieving minimal transmission line filtering. This allows for the acquisition of a clear and complete intrusion point cloud during ranging. The direction vector of each line segment is used as the basis for this process. With ground normal For each line segment, construct a plane-intersection model, which is formed by planes containing the line segments and being pairwise orthogonal. The three-dimensional closest distance from construction machinery to the guide wire is reduced to a simple analytical geometric calculation. The minimum clear distance and threshold value output in the intersection form are then compared. The system compares and provides tiered warnings and alarms; simultaneously, it uses YOLO object detection on the video side for image domain constraints, only detecting targets at the "minimum Euclidean distance". An alarm is triggered when the condition that "the intruder's point cloud projection is located within the target bounding box" is met simultaneously. The system records the minimum clearance and corresponding line segment ID at the time of the alarm, and archives them with the timestamp for traceability and quality inspection.

[0021] Example 2 This embodiment is a further design based on Embodiment 1. The electric field point cloud segmentation and extraction method described in this example utilizes the spatial connectivity and suspended physical characteristics of electric field point clouds and the alignment of the coordinate system. The specific method is as follows.

[0022] First, set the maximum retention distance. As a configurable value, it is preferably placed within the range that meets the sensor's measurement range and task coverage. It can be adaptively adjusted according to the scene scale to reduce false detections.

[0023] Ground point cloud (CSF) is used to separate ground point cloud and non-ground point cloud. Key parameters are set: cloth resolution. Time step Number of iterations Stiffness coefficient Based on the average spacing of point clouds Residual scale fitted to ground Adaptive parameter selection. The ground points are fitted using Random Sample Consistency (RANSAC). And refine using least squares to output the normal direction. With plane constant .

[0024] Using the minimum rotation from the ground normal to the standard vertical direction , center point of the ground Translate to Plane, forming rigid body transformation Transform the non-ground point cloud to a new coordinate system in which the ground strictly satisfies... And not a ground feature This facilitates subsequent determination of suspension.

[0025] Secondly, construct an octree index, with leaf nodes at voxel scale. A connected component search is performed using a 3D 26-neighborhood, with a minimum threshold number of points set. To eliminate noisy clusters; perform shape anisotropy screening on slender clusters, and prioritize the retention of slender targets.

[0026] Calculate each connected component to minimum vertical distance quantile distance to the local surface With cluster thickness Define the suspending threshold. in Estimate the height of the power line. It is a monotonically increasing positive function; when and Retain as candidate clusters for power lines. Perform bridging and merging on candidate clusters that are close to each other and collinear and in the same direction to reduce breakage.

[0027] Example 3 This embodiment is a further design based on Embodiment 1, which involves the segmentation of the power line and the reconstruction of the model of each segment. The specific method is as follows.

[0028] First, geometric feature analysis is performed on each candidate power line point cloud cluster to determine the modeling method. After aligning the coordinate systems, the point clouds are projected onto the horizontal plane (XOY plane) and the vertical plane (XOZ plane), respectively. The projection length on the horizontal plane is calculated and compared with a preset length threshold. A comparison was made; simultaneously, quadratic curve fitting was performed on the projected point cloud of the vertical plane. Obtain the coefficient A of the quadratic term and its absolute value. Compare with a preset curvature threshold.

[0029] Based on the above analysis results, the modeling method is selected as follows: if the projected length is greater than the length threshold... and Greater than the curvature threshold If the line segment is long-span curved, it is identified as a long-span curved conductor and a segmented straightening model is adopted to straighten the curve. Otherwise, a single segment model is established for the entire candidate cluster. Orthogonal least squares (OLS) is used to model the line segment. Refine the parameters of the line segment, denoted by the direction vector. The reference point on the straight line is For points within the cluster Calculate projection parameters Pick , The endpoints of the line segment are , For a single line segment model, the principal direction vector is directly obtained through principal component analysis (PCA), and the model parameters are obtained by combining it with least squares linear fitting.

[0030] Finally, for adjacent and collinear sub-segments, if the endpoint gap is ≤ And the included angle ≤ Perform bridging and refining to obtain a set of line segments. For each Assign a unique ID and store the direction vector endpoints .Will It serves as the axis constraint input for subsequent OBCs and as the intersection line constraint input for intersecting surface models.

[0031] Example 4 This embodiment, based on Embodiment 1, further incorporates the following methods for power line OBC model reconstruction, power line point cloud filtering, intersection model reconstruction, and foreign object point cloud detection: First of all, with Construct an OBC for the axis and calculate the perpendicular distance from points within the cluster to the axis. Take the radius Constructing OBC. In actual implementation, ranging is performed on real-time frame point clouds, using... and The relative size is used to determine the value: if Points identified as power line points are filtered out, retaining only non-power line points to reduce subsequent computation and interference. The output set of candidate intrusion points after power line removal serves as the prior input for interfacial ranging and hazardous area determination.

[0032] Secondly, with the first part Line segment direction vector With ground normal Using the base, construct pairwise orthogonal structures that both pass through the base. auxiliary plane ,like Figure 4 As shown: Let Legal direction (with ground plane) (orthogonal) Legal direction (and plane) (orthogonal) For any point First project onto the corresponding auxiliary plane. have to , in The reference point can be a line segment. For any point on the surface, the horizontal distance is... : , Then along Constructing a perpendicular line and intersection vertical distance : Finally, the distance between the intruder and each conductor of the power transmission channel was calculated separately, and the results were obtained. ...If there are two power lines on this layer, calculate their spacing W (W=0 if there is only one edge line per layer) to determine the region of the intruder relative to the conductor and calculate the closest distance. Take the conductor X that is closest to the intruder in both horizontal and vertical distances, and the opposite edge conductor Y on the same layer; output the directional distance used for determination. and European style closest distance This provides a basis for threshold comparison and alarm triggering.

[0033] Example 5 This embodiment is a further design based on Embodiment 1, wherein the near-electrical intrusion alarm based on dual verification described in this example is as follows.

[0034] First, set the vertical and horizontal safety thresholds according to the voltage level. and a set of hierarchical thresholds. Based on a facet-based model for preliminary judgment, real-time initial screening is performed, such as... Figure 5 As shown: Calculate the distance to the left edge for each candidate intrusion point cloud cluster. and the distance to the right sideline The system determines the spatial region based on distance: if only one horizontal distance exceeds the limit, it is classified as line gap region B; if both the left horizontal and vertical distances exceed the limit simultaneously, it is classified as left neighbor region A; if both the right horizontal and vertical distances exceed the limit simultaneously, it is classified as right neighbor region C. The "candidate intrusion" flag is triggered according to the following rules: Region B: If or ; Region A: If and ; Region C: If and Secondly, the candidate intrusion point cloud clusters are projected onto the image, and it is checked whether they fall within the target detection bounding box (YOLO BBox) of the construction machinery. If the projected points do not fall within the bounding box, they are determined to be non-mechanical interference (such as trees, birds, etc.) and are removed to reduce false alarms. Each candidate intrusion point cloud cluster that passes the mutual verification of 2D target detection and spatial relationship is identified as a construction machinery point cloud cluster, and its distance to the line segment set is calculated. minimum Euclidean distance .

[0035] Finally, the system records the alarm. Timestamp; and minimum Euclidean distance The distance is compared with the preset graded protection thresholds: when the distance is less than the first-level safety warning threshold, an audible and visual alarm is triggered to remind the workers to take precautions; when the distance continues to decrease and falls below the second-level emergency braking threshold, regardless of which area it is in, it is determined to be a high-risk intrusion, and a control signal is directly output to stop the construction machinery, so as to achieve accurate graded protection based on distance perception.

[0036] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the exemplary embodiments according to this application. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0037] It should be noted that the terms "first," "second," etc., used in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in sequences other than those illustrated or described herein.

[0038] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A proximity alarm method based on a cylinder and an intersecting surface model, characterized in that: include Step S1. Deploy a mobile ground-mounted camera-radar integrated monitoring device at the near-electricity work site to acquire lidar point cloud and video images containing power lines and construction machinery from a third-person perspective. After removing the ground point cloud, align the coordinate system and use a multi-feature method to segment and extract candidate clusters of power line point cloud. Step S2. For each candidate cluster of power line point cloud, by extracting the power line span and curvature features corresponding to the cluster, determine whether to use a single line segment model or a segmented straightening model to fit the power line. The single line segment model is used to provide axial constraints for the directed boundary cylinder, and the segmented straightening model is used to provide intersection line constraints for the intersection surface model. Step S3. Using the line segment model as the axis, calculate the maximum vertical distance from the electric field line point cloud to the axis to determine the minimum envelope radius, thereby establishing a directed boundary cylinder model. Further filter out the electric field line point cloud within the directed boundary cylinder model, so that a clear and complete intruder point cloud can be obtained during ranging. Step S4. For each line segment in the piecewise linearization model Construct a surface-intersecting model, which contains two orthogonal surfaces that intersect at... plane perpendicular to the ground plane And its normal vector is Direction vector Orthogonal, plane Perpendicular to the plane And its normal vector is Direction vector Orthogonal, using this intersecting surface model, any point outside the electric field lines can be directed to the plane. Projection to calculate horizontal distance To the plane Projection to calculate vertical distance ; Step S5. For each power line, a dual verification mechanism is used to determine whether construction machinery has intruded into the power line's safe zone, reducing false positives and false negatives. On one hand, the point closest to the surface of the directed boundary cylinder model of the power line is extracted, and on the other hand, the horizontal distance output by the intersection model is determined in real time. Vertical distance If the value is less than the threshold, it is considered a suspicious intruder. On the other hand, the location of the suspicious intruder in the image is checked by projecting the point cloud onto the image. An alarm is triggered only when the projected pixels of the intruder's point cloud are within the bounding box of the construction machinery image.

2. The proximity alarm method based on a cylinder and intersecting surface model according to claim 1, characterized in that: The specific method for power line point cloud segmentation and extraction in step S1 includes: By preprocessing the original point cloud with coordinate system alignment and ground separation, and then comprehensively utilizing the spatial connectivity and suspended physical characteristics of the power line point cloud, power lines are identified and extracted from complex non-ground point clouds. The coordinate system alignment preprocessing helps to overcome the interference caused by the sloped and not completely level ground at the work site on subsequent processing, and provides a unified discrimination benchmark for accurately screening power line point clouds based on suspended physical characteristics.

3. The proximity alarm method based on a cylinder and intersecting surface model according to claim 1, characterized in that: The specific methods for determining which model to use for fitting based on the power line span and curvature characteristics in step S2 include: After aligning the coordinate systems, determine whether the projection length of the point cloud on the horizontal plane (XOY plane) is greater than a preset length threshold. Furthermore, its fitted quadratic curve in the vertical plane (XOZ plane) The absolute value of the coefficient of the quadratic term Is it greater than the preset curvature threshold? If all the above conditions are met, then the candidate cluster is modeled using segmented linearization with straightening of bends; otherwise, a single line segment model is established. Finally, the generated segmented line segments are refined and collinearized based on gaps and angles to form the final set of each line segment and its corresponding point cloud cluster. For each line segment Assign a unique ID and store the direction vector endpoints This provides axis constraints for subsequent OBC models and intersection line constraints for surface-type models.

4. The proximity alarm method based on a cylinder and intersecting surface model according to claim 1, characterized in that: The specific methods for power line OBC model reconstruction and power line point cloud filtering in step S3 include: by Using the axis as the reference line, calculate the perpendicular distance from each point within the cluster to the axis. Take the radius To construct an OBC (Online Component Controller) for real-time point cloud ranging, and The relative size is used to determine the value: if If the points are identified as power line points, they are filtered out. Only non-power line points are retained to reduce subsequent computation and interference. The candidate intrusion point cloud after power line removal is output as the input prior for cross-surface ranging and dangerous area determination.

5. The proximity alarm method based on a cylinder and intersecting surface model according to claim 1, characterized in that: The specific method for reconstructing the power line intersection model in step S4 includes: Step S4-1. Using the first part Line segment direction vector With ground normal Using the base, construct pairwise orthogonal structures that both pass through the base. auxiliary plane :make Legal direction (with ground plane) (orthogonal) Legal direction (and plane) (orthogonal) Step S4-2. For any point First project onto the corresponding auxiliary plane. have to , in The reference point can be a line segment. For any point on the surface, the horizontal distance is... : ， Then along Constructing a perpendicular line and intersection vertical distance : Step S4-3. For any point outside the directed boundary cylinder model of the electric field lines, calculate the horizontal and vertical distances between that point and each electric field line. ...If there are two power lines in this layer, calculate their spacing W to determine the region of the intruder relative to the conductor and calculate the closest distance.

6. The proximity alarm method based on a cylinder and intersecting surface model according to claim 1, characterized in that: The specific method for near-electrical intrusion alarm based on dual verification in step S5 includes: Step S5-1. Set the vertical and horizontal safety thresholds according to the voltage level. The system uses a set of hierarchical thresholds and a faceted model for preliminary judgment to perform real-time screening. For each candidate intrusion point cluster, the distance from its left edge is calculated. and the distance to the right sideline The system determines the spatial region based on distance: if only one horizontal distance exceeds the limit, it is classified as line gap region B; if both the left horizontal and vertical distances exceed the limits simultaneously, it is classified as left neighbor region A; if both the right horizontal and vertical distances exceed the limits simultaneously, it is classified as right neighbor region C, and the "candidate intrusion" flag is triggered according to the following rules: like or This triggers the "candidate intrusion" flag in region B; like and This triggers the "candidate intrusion" flag in region A; like and Triggering the "candidate intrusion" flag in region C. Step S5-2. Project the candidate intruder point cloud clusters onto the image and check whether they fall within the bounding box of the construction machinery target recognition. If not, remove them to reduce false alarms. Step S5-3. For each cluster of construction machinery points that has passed 2D target detection and spatial relation verification, calculate its distance to the set of line segments. minimum Euclidean distance ,Record Timestamp; and minimum Euclidean distance The distance is compared with the preset graded protection thresholds: when the distance is less than the first-level safety warning threshold, an audible and visual alarm is triggered to remind the workers to take precautions; when the distance continues to decrease and falls below the second-level emergency braking threshold, regardless of which area it is in, it is determined to be a high-risk intrusion, and a control signal is directly output to stop the construction machinery, so as to achieve accurate graded protection based on distance perception.

Images