Large machinery near electric operation regulation method and device, storage medium and electronic equipment

Abstract

The application discloses a large-scale machine near-electric operation regulation method and device, a storage medium and electronic equipment, comprising: collecting three-dimensional space point cloud data and image data of an operation area and performing feature matching and fusion to obtain target point cloud data; performing point cloud segmentation processing on the preprocessed target point cloud data, constructing a three-dimensional model of a charged body and a machine operation component based on the point cloud segmentation result, and predicting a motion trajectory of the machine operation component based on the three-dimensional model by using a filtering algorithm; adaptively adjusting a dynamic safety threshold by a fuzzy logic reasoning algorithm based on voltage level parameters and environmental interference parameters of the operation area; generating a corresponding level of early warning regulation instruction based on the dynamic safety threshold, an actual distance between the machine operation component and the charged body, a motion trajectory prediction result of the machine operation component and a distance change rate; and triggering adaptive regulation actions of the large-scale machine according to the early warning regulation instruction, so that operation risks are actively avoided.

Description

Technical Field

[0001] This application relates to the field of power system safety protection technology, and in particular to a method, device, storage medium and electronic equipment for controlling near-electric operation of large machinery. Background Technology

[0002] In engineering fields such as power infrastructure and rail transit construction, near-electricity operations involving large machinery such as cranes, tower cranes, and excavators are becoming increasingly frequent within the protection zones of high-voltage transmission lines and around substations. In these operations, the long reach of the machinery's booms, the dynamically changing working range, and the often complex terrain and overlapping voltage levels create significant risks. If the distance between the machinery's working parts and live conductors falls below a safe threshold, it can easily lead to safety accidents such as arcing, short circuits, and electric shocks. These accidents can not only result in loss of life for construction workers but also cause widespread power outages, severely disrupting the stable operation of the power grid and significantly impacting social production and daily life. Therefore, developing scientific and efficient methods for controlling near-electricity operations involving large machinery, enabling accurate prediction and timely intervention of operational risks, has become a critical requirement for ensuring the safety of near-electricity operations.

[0003] Currently, the control of large machinery operating near live electrical equipment largely relies on traditional technical solutions, primarily using a combination of manual monitoring and single-sensor monitoring for risk control. Specifically, during operation, a dedicated person must observe the distance between the machinery and the live conductor in real time, while using single sensing devices such as infrared and ultrasonic sensors to collect distance data. When the detected distance approaches a preset threshold, audible and visual alarms are used to alert the operator to stop operation. Furthermore, the safety thresholds in existing control methods are mostly set with fixed values, meaning a uniform warning threshold standard is used for different voltage levels and different operating environments, without considering the dynamic changes that occur during actual operation.

[0004] The existing control methods described above have significant core technical problems: On the one hand, single sensors are easily affected by environmental factors such as strong outdoor light, rain, fog, and dust, resulting in insufficient accuracy in distance data acquisition. Furthermore, manual monitoring has inherent defects such as blind spots and reaction delays, making it difficult to accurately capture risks. On the other hand, the fixed threshold setting method cannot adapt to the safety distance requirements of different voltage levels and dynamically changing operating scenarios, making the control logic lack flexibility and pertinence. Consequently, warnings are either too frequent, affecting operational efficiency, or they lag behind the actual development of risks, failing to provide timely and accurate control support for near-electric work and making it difficult to fundamentally avoid the safety risks of near-electric work. Summary of the Invention

[0005] In view of this, this application provides a method, device, storage medium and electronic equipment for controlling near-electric work on large machinery, which can provide timely and accurate control support for near-electric work, enabling it to fundamentally avoid the safety risks of near-electric work.

[0006] According to a first aspect of this application, a method for controlling near-electric operation of large machinery is provided, comprising: Simultaneously collect 3D spatial point cloud data and image data of the work area, and use a data fusion algorithm to perform feature matching and fusion of the 3D spatial point cloud data and the image data to obtain target point cloud data with texture features; The preprocessed target point cloud data is segmented into points. Based on the point cloud segmentation results, a three-dimensional model of the charged body and the mechanical working parts is constructed. A filtering algorithm is used to predict the motion trajectory of the mechanical working parts based on the three-dimensional model. Based on the voltage level parameters and environmental interference parameters of the work area, the dynamic safety threshold is adaptively adjusted through a fuzzy logic reasoning algorithm. Based on the dynamic safety threshold, the actual distance between the mechanical working part and the charged body, the predicted motion trajectory of the mechanical working part and the distance change rate, a corresponding level of early warning and control command is generated. The early warning and control instructions trigger the adaptive control actions of large machinery to proactively avoid operational risks.

[0007] According to a second aspect of this application, a large-scale machinery near-electric operation control device is provided, comprising: The acquisition module is used to simultaneously acquire three-dimensional spatial point cloud data and image data of the work area, and to perform feature matching and fusion of the three-dimensional spatial point cloud data and the image data through a data fusion algorithm to obtain target point cloud data with texture features. The first processing module is used to perform point cloud segmentation on the preprocessed target point cloud data, construct a three-dimensional model of the charged body and mechanical working parts based on the point cloud segmentation results, and use a filtering algorithm to predict the motion trajectory of the mechanical working parts based on the three-dimensional model. The adjustment module is used to adaptively adjust the dynamic safety threshold based on the voltage level parameters and environmental interference parameters of the work area using a fuzzy logic reasoning algorithm. The generation module is used to generate early warning and control instructions of corresponding levels based on the dynamic safety threshold, the actual distance between the mechanical working part and the charged body, the predicted motion trajectory of the mechanical working part and the distance change rate. The control module is used to trigger adaptive control actions of large machinery according to the early warning control instructions, so as to achieve proactive avoidance of operational risks.

[0008] According to a third aspect of this application, a storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the above-described method for controlling near-electric operation of large machinery.

[0009] According to a fourth aspect of this application, an electronic device is provided, including a storage medium, a processor, and a computer program stored on the storage medium and executable on the processor, wherein the processor executes the program to implement the above-described method for controlling near-electric operation of large machinery.

[0010] By employing the aforementioned technical solutions, the large machinery near-electric operation control method, device, storage medium, and electronic equipment provided in this application, through simultaneous acquisition of three-dimensional point cloud and image data and feature fusion, combined with three-level point cloud preprocessing, segmentation modeling, and filtering trajectory prediction algorithms, can effectively improve the identification accuracy and ranging accuracy of charged bodies and mechanical operating parts, solving the problems of traditional single sensors being susceptible to environmental interference and visual blind spots and reaction delays in manual monitoring; based on voltage level and environmental interference parameters, the dynamic safety threshold is adaptively adjusted through fuzzy logic reasoning algorithms, which can overcome the limitation that fixed thresholds cannot adapt to complex operating scenarios; further, by combining actual distance, trajectory prediction results, and distance change rate to generate graded early warning control commands and triggering adaptive control actions of large machinery, an upgrade from passive early warning to active prevention and control can be achieved, which can not only avoid the impact of excessive early warning on operating efficiency, but also eliminate the safety hazards caused by delayed early warning, ultimately fundamentally improving the safety protection level of large machinery near-electric operation.

[0011] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, the following are specific embodiments of this application. Attached Figure Description

[0012] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings: Figure 1 A flowchart illustrating a method for controlling near-electric operation of large machinery according to an embodiment of this application is shown. Figure 2 A flowchart illustrating a method for controlling near-electric operation of large machinery according to another embodiment of this application is shown; Figure 3 This paper shows a schematic diagram of the structure of a large-scale mechanical near-electric operation control device provided in an embodiment of this application; Figure 4A schematic diagram of the structure of a large-scale machinery near-electric operation control device provided in another embodiment of this application is shown. Detailed Implementation

[0013] The present application will be described in detail below with reference to the accompanying drawings and embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in the embodiments of the present application can be combined with each other.

[0014] Currently, the control of large machinery operating near live electrical equipment largely relies on traditional technical solutions, primarily using a combination of manual monitoring and single-sensor monitoring for risk control. Specifically, during operation, a dedicated person must observe the distance between the machinery and the live conductor in real time, while using single sensing devices such as infrared and ultrasonic sensors to collect distance data. When the detected distance approaches a preset threshold, audible and visual alarms are used to alert the operator to stop operation. Furthermore, the safety thresholds in existing control methods are mostly set with fixed values, meaning a uniform warning threshold standard is used for different voltage levels and different operating environments, without considering the dynamic changes that occur during actual operation.

[0015] The existing control methods described above have significant core technical problems: On the one hand, single sensors are easily affected by environmental factors such as strong outdoor light, rain, fog, and dust, resulting in insufficient accuracy in distance data acquisition. Furthermore, manual monitoring has inherent defects such as blind spots and reaction delays, making it difficult to accurately capture risks. On the other hand, the fixed threshold setting method cannot adapt to the safety distance requirements of different voltage levels and dynamically changing operating scenarios, making the control logic lack flexibility and pertinence. Consequently, warnings are either too frequent, affecting operational efficiency, or they lag behind the actual development of risks, failing to provide timely and accurate control support for near-electric work and making it difficult to fundamentally avoid the safety risks of near-electric work.

[0016] To address the aforementioned technical problems, embodiments of the present invention provide a method for controlling near-electric operation of large machinery, such as... Figure 1 As shown, the method includes: Step 110: Synchronously collect 3D spatial point cloud data and image data of the work area, and use a data fusion algorithm to perform feature matching and fusion of the 3D spatial point cloud data and image data to obtain target point cloud data with texture features.

[0017] Among them, the three-dimensional spatial point cloud data is a discrete set of points collected by three-dimensional LiDAR, reflecting the three-dimensional spatial position information of the target object in the working area, which can accurately present the spatial outline and position parameters of the object; the image data is a two-dimensional visual image of the working area collected by a high-definition camera, which can provide intuitive visual feature information such as texture and appearance of the target object; the data fusion algorithm is an algorithm used to integrate multi-source sensing data, which matches, associates and complements the features of data collected by different devices, removes redundant information, and improves the accuracy and completeness of the data; the target point cloud data with texture features is the product obtained by fusing three-dimensional spatial point cloud data and image data, which has both the three-dimensional spatial position attributes of point cloud data and integrates the texture appearance features of image data.

[0018] In this embodiment of the present disclosure, three-dimensional spatial point cloud data and image data of the work area can be acquired simultaneously by a sensing device. The features of the two types of data are matched and associated by a data fusion algorithm. The texture information carried by the image data is assigned to the corresponding three-dimensional spatial point cloud, so that the point cloud data, which originally only has spatial location attributes, also has visual texture features. Finally, target point cloud data with texture features is generated, providing comprehensive and accurate basic data support for subsequent core links such as point cloud processing, target modeling, and trajectory prediction.

[0019] By simultaneously collecting and fusing 3D spatial point cloud data and image data, target point cloud data with texture features can be generated. This effectively integrates the spatial positioning advantages of 3D LiDAR with the visual texture advantages of high-definition cameras, breaking through the limitations of information partiality in data collected by a single sensor. It significantly improves the richness and accuracy of target data, providing high-quality basic data for subsequent target segmentation modeling, dynamic trajectory prediction, and intelligent hierarchical early warning in near-electric operations. This ensures the accuracy and reliability of the entire intelligent early warning and adaptive control process from the source.

[0020] Step 120: Perform point cloud segmentation on the preprocessed target point cloud data, construct a three-dimensional model of the charged body and mechanical working parts based on the point cloud segmentation results, and use a filtering algorithm to predict the motion trajectory of the mechanical working parts based on the three-dimensional model.

[0021] Point cloud segmentation is a processing technique that uses corresponding algorithms to divide the pre-processed target point cloud data into point cloud clusters corresponding to the charged body and mechanical working parts, thereby achieving accurate separation of different targets. The 3D model of the charged body and mechanical working parts is a digital model that accurately reflects the spatial contour and positional attributes of the charged body and mechanical working parts, based on the segmented target point cloud clusters and through operations such as feature extraction, matching, and 3D coordinate calculation. The filtering algorithm is used to denoise the continuous position data of the mechanical working parts, construct a motion state model, and then realize the prediction of its future motion trajectory. Typical algorithms include the extended Kalman filter algorithm.

[0022] In this embodiment of the disclosure, point cloud segmentation processing can be performed on the preprocessed target point cloud data. Point cloud clusters corresponding to charged bodies and mechanical working parts can be divided using corresponding algorithms. Then, feature extraction, matching and three-dimensional coordinate calculation can be performed based on these point cloud clusters to construct high-precision three-dimensional models of the two types of targets. Afterwards, filtering algorithms are used to process the continuous position data of the mechanical working parts to build their motion state model, thereby realizing the prediction of the future motion trajectory of the mechanical working parts.

[0023] By accurately segmenting and 3D modeling point cloud data, the spatial relationship between charged bodies and mechanical operating parts can be clearly defined, providing reliable basic data support for subsequent real-time actual distance calculation. Combined with the motion trajectory prediction of mechanical operating parts achieved by filtering algorithms, it is possible to predict in advance whether there is a risk of the parts intruding into the safety threshold, greatly improving the foresight and accuracy of near-electric work early warning.

[0024] Step 130: Based on the voltage level parameters and environmental interference parameters of the work area, the dynamic safety threshold is adaptively adjusted using a fuzzy logic reasoning algorithm.

[0025] Among them, voltage level parameters refer to the voltage level of power transmission lines or electrical equipment in the vicinity of the work area for large machinery, and different voltage levels correspond to different safety distance benchmark requirements; environmental interference parameters refer to environmental factors that affect the accuracy of data acquisition by sensing equipment at the work site, including key indicators such as light intensity, dust concentration, rain, fog and humidity; fuzzy logic reasoning algorithm is an intelligent algorithm used to process uncertain information. It realizes reasoning and decision-making from input parameters to output results by constructing fuzzy subsets of input and output variables, configuring membership functions and establishing a fuzzy rule base; dynamic safety threshold is a safety distance critical value that can be adaptively adjusted according to the voltage level and environmental interference of the work area, which is different from the traditional fixed threshold. It is the core benchmark for determining whether mechanical working parts have entered the dangerous area.

[0026] In this embodiment of the disclosure, voltage level parameters and environmental interference parameters of the work area can be obtained. These two types of parameters are input into a pre-constructed fuzzy logic reasoning model. The model has pre-set fuzzy input subsets corresponding to the parameters, fuzzy output subsets of safety thresholds, and a fuzzy rule base based on power industry safety standards. By fuzzifying the input parameters and matching them with the corresponding rules in the rule base, and then performing defuzzification processing on the fuzzy reasoning results obtained from the matching, a dynamic safety threshold adapted to the current work scenario is finally output, providing an accurate basis for the generation of subsequent graded early warning instructions.

[0027] Based on the voltage level parameters and environmental interference parameters of the work area, a fuzzy logic reasoning algorithm is used to achieve adaptive adjustment of the dynamic safety threshold. This overcomes the limitations of traditional fixed thresholds, which cannot take into account the safety distance requirements of different voltage levels and complex and ever-changing work environments. It makes the setting of safety thresholds more targeted and flexible, effectively avoiding the problem of over-warning caused by setting the threshold too high, and also eliminating the risk of warning lag caused by setting the threshold too low. In this way, it can provide reliable threshold determination support for accurate graded warning and active control of near-electric work.

[0028] Step 140: Based on the dynamic safety threshold, the actual distance between the mechanical working parts and the live conductor, the predicted motion trajectory of the mechanical working parts and the distance change rate, generate the corresponding level of early warning and control instructions.

[0029] Among them, the actual distance between the mechanical working part and the live conductor is the real-time spatial distance between the two calculated based on a high-precision three-dimensional model of the live conductor and the mechanical working part; the motion trajectory prediction result of the mechanical working part is the motion trajectory prediction data for a future period of time obtained by processing the continuous position data of the mechanical working part through a filtering algorithm and constructing a motion state model; the distance change rate is the real-time rate of change of the distance between the mechanical working part and the live conductor calculated based on the continuous three-dimensional position coordinates of the mechanical working part within a preset time period; the early warning and control instructions are instructions that generate corresponding differentiated control strategies based on the risk level determined by multiple parameters, such as control action instructions with three levels: prompt, early warning and alarm.

[0030] In this embodiment of the disclosure, four core parameters can be integrated: dynamic safety threshold, actual distance between mechanical working parts and live conductors, prediction results of the mechanical working parts' motion trajectory, and distance change rate. These parameters are then input into a multi-parameter early warning judgment model for comprehensive calculation and judgment. Based on preset risk level classification rules, the current risk level is determined, and an early warning control instruction matching the risk level is generated.

[0031] By integrating four core parameters—dynamic safety threshold, real-time actual distance, trajectory prediction results, and distance change rate—corresponding early warning and control instructions can be generated, breaking through the limitations of traditional single-parameter risk assessment. This enables precise quantitative assessment of operational risks, effectively avoiding the impact of excessive early warnings on operational efficiency and eliminating safety hazards caused by delayed early warnings. Through differentiated control instructions, the shift from passive early warning to proactive prevention and control in near-electric work can be promoted, comprehensively improving the safety protection level of near-electric work for large machinery.

[0032] Step 150: Trigger the adaptive control action of large machinery according to the early warning and control instructions to achieve proactive avoidance of operational risks.

[0033] Among them, adaptive control action is a differentiated operation that large machinery automatically executes based on different levels of early warning control instructions, covering action types such as fine adjustment of the speed of working parts, deceleration, and emergency shutdown; active avoidance is a protective method that differs from the passive response mode of traditional manual monitoring. It uses intelligent algorithms to predict risks and trigger automatic control of machinery, and prevents risks in advance before working parts enter dangerous areas.

[0034] In this embodiment of the disclosure, a generated early warning control command can be received, the risk level corresponding to the command can be parsed, and the large machinery can be triggered to perform an adaptive control action that matches the level information. At the same time, the control status is fed back to the interaction unit in real time, forming a complete control closed loop from command reception, action execution to status feedback, and ultimately realizing the proactive avoidance of the risk of large machinery operating near electric fields.

[0035] By triggering adaptive control actions of large machinery according to early warning and control instructions, a fundamental shift from passive early warning to proactive prevention and control of near-electricity operation risks can be achieved. Differentiated control strategies can not only avoid the impact of excessive control on work efficiency, but also intervene in time at the nascent stage of risks, effectively preventing external damage accidents caused by mechanical operating parts intruding into the safe distance of live conductors, and comprehensively improving the safety protection level and work continuity of large machinery near-electricity operations.

[0036] In summary, the large machinery near-electric operation control method provided in this application, by simultaneously acquiring 3D point cloud and image data and performing feature fusion, combined with three-level point cloud preprocessing, segmentation modeling, and filtering trajectory prediction algorithms, can effectively improve the identification accuracy and ranging accuracy of live bodies and mechanical operating parts, solving the problems of traditional single sensors being susceptible to environmental interference and blind spots and reaction delays in manual monitoring; based on voltage level and environmental interference parameters, the dynamic safety threshold is adaptively adjusted through fuzzy logic reasoning algorithm, which can overcome the limitation that fixed thresholds cannot adapt to complex operating scenarios; and by combining actual distance, trajectory prediction results, and distance change rate to generate graded early warning control commands and trigger adaptive control actions of large machinery, it can achieve an upgrade from passive early warning to active prevention and control, which can not only avoid the impact of excessive early warning on operating efficiency, but also eliminate the safety hazards caused by delayed early warning, ultimately fundamentally improving the safety protection level of large machinery near-electric operation.

[0037] Furthermore, as a refinement and extension of the specific implementation methods of the above embodiments, and to fully illustrate the implementation methods of this embodiment, this embodiment also provides another method for controlling near-electric operation of large machinery, such as... Figure 2 As shown, the method includes: Step 210: Synchronously collect 3D spatial point cloud data and image data of the work area, and use a data fusion algorithm to perform feature matching and fusion of the 3D spatial point cloud data and image data to obtain target point cloud data with texture features.

[0038] For the specific implementation process of the embodiments disclosed herein, please refer to the relevant description in step 110 of the embodiment, which will not be repeated here.

[0039] Step 220: The target point cloud data is divided into blocks based on the improved statistical filtering algorithm. The neighborhood radius of each point is dynamically determined according to the local density of the point cloud in each sub-block. The mean distance and standard deviation of the distance between each point and multiple neighboring points in the corresponding neighborhood are calculated based on the neighborhood radius. The outlier judgment threshold corresponding to the mean distance and standard deviation of the distance is determined. Outlier noise points with a mean distance greater than the outlier judgment threshold are removed from the target point cloud data to obtain the first point cloud data.

[0040] The neighborhood radius is a parameter dynamically adjusted based on the local density of the point cloud. It defines the range within which a single point searches for its neighbors; the radius decreases when the density is high and increases when the density is low. The formula for calculating the dynamic neighborhood radius is: In the formula, The dynamic neighborhood radius of the current point. The reference neighborhood radius (which can be 0.02m). The average density of the entire point cloud. The distance mean is the point cloud density of the local region where the current point is located; the distance standard deviation is the average distance from all neighboring points in the neighborhood of a single point to that point, and is a basic indicator for determining the dispersion of the point cloud; the distance standard deviation is a quantitative indicator of the dispersion of the distances from all neighboring points in the neighborhood of a single point to that point, reflecting the uniformity of the distribution of the point cloud within the neighborhood; the outlier detection threshold is a critical value calculated based on the distance mean and standard deviation, used to distinguish noise points from valid points in the point cloud data. The formula for calculating the outlier detection threshold is: In the formula, Threshold for identifying outliers The mean distance between points in the neighborhood. denoted as the standard deviation, and k is the benchmark coefficient (which can be 2.5). The noise intensity correction coefficient (which can be dynamically set from 0.8 to 1.2 based on the local area noise statistics) is the initial purified point cloud data output after the target point cloud data has been filtered by an improved statistical filtering algorithm to remove outlier noise points.

[0041] In this embodiment of the present disclosure, the target point cloud data can be segmented based on an improved statistical filtering algorithm. The neighborhood radius of each point is dynamically determined according to the local density of the point cloud in each sub-block. Then, the mean distance and standard deviation of the distance between multiple neighboring points in the neighborhood corresponding to each point are calculated based on the neighborhood radius. Based on this, an outlier judgment threshold matching the mean distance and standard deviation of the distance is determined. Outlier noise points with a mean distance greater than the outlier judgment threshold are removed from the target point cloud data, and finally the first point cloud data is obtained.

[0042] Point cloud denoising based on an improved statistical filtering algorithm achieves accurate denoising by dynamically adjusting the neighborhood radius to adapt to point cloud sub-blocks of different densities and combining the outlier judgment threshold determined by the distance mean and standard deviation. This approach can effectively remove outlier noise points from the target point cloud data while avoiding excessive removal of valid point cloud data, thus significantly improving the purity and integrity of the point cloud data.

[0043] Step 230: Perform radius filtering on the first point cloud data to obtain the second point cloud data after removing isolated noise points. The radius filtering is used to count the number of neighboring points within a fixed neighborhood radius for each point and remove isolated noise points whose corresponding number of neighboring points is less than the minimum neighborhood point threshold.

[0044] The radius filtering process employs a fixed neighborhood radius judgment mechanism to accurately identify and remove isolated noise points from the point cloud data. The fixed neighborhood radius is a fixed parameter (which can be 0.05m) used in the radius filtering to define the search range of neighboring points for a single point, providing a unified judgment standard for the number of neighboring points. The minimum neighborhood point threshold is the critical number value (which can be 10) used in the radius filtering to determine whether a point is an isolated noise point, and it is the core basis for distinguishing between isolated noise points and valid point clouds. The second point cloud data is the further purified point cloud data obtained after the first point cloud data has been processed by radius filtering to remove isolated noise points.

[0045] In this embodiment of the present disclosure, radius filtering can be performed on the first point cloud data obtained by improved statistical filtering. A fixed neighborhood radius and a minimum neighbor number threshold are preset. Each point in the first point cloud data is traversed, and the number of neighbor points within the fixed neighborhood radius is counted. Points with a number of neighbor points less than the minimum neighbor number threshold are identified as isolated noise points and removed. Finally, the second point cloud data after removing isolated noise points is obtained.

[0046] Radius filtering, through a dual judgment mechanism of fixed neighborhood radius and minimum neighborhood point threshold, can accurately identify and remove isolated noise points remaining in the first point cloud data. While preserving the integrity of effective point cloud data to the maximum extent, it significantly improves the continuity and purity of point cloud data, providing higher quality basic data support for subsequent core algorithm steps such as voxel mesh filtering, point cloud segmentation, and 3D modeling.

[0047] Step 240: Perform voxel grid filtering on the second point cloud data. Divide the three-dimensional space where the point cloud is located into a uniform voxel grid according to the preset voxel size. Calculate the mean three-dimensional coordinates of all points in each voxel grid. Use the mean three-dimensional coordinates to replace all points in the voxel grid to obtain the preprocessed target point cloud data.

[0048] Among them, voxel grid filtering is a preprocessing algorithm for point cloud data compression and feature preservation. It simplifies data through spatial partitioning and mean substitution. The voxel grid is a uniform cubic unit formed by dividing the three-dimensional space where the point cloud is located according to a preset size. It is the basic processing unit of voxel grid filtering. The three-dimensional coordinate mean is the arithmetic mean of the x, y, and z three-dimensional coordinates of all points in a single voxel grid. It is used to replace the position information of all points in the grid. The preprocessed target point cloud data is the final point cloud preprocessing result that combines purity and simplification after the second point cloud data has been processed by voxel grid filtering.

[0049] In this embodiment of the present disclosure, the second point cloud data obtained after improved statistical filtering and radius filtering can be subjected to voxel grid filtering. The voxel size is preset and the three-dimensional space where the point cloud is located is divided into a uniform voxel grid. Each voxel grid is traversed and the mean three-dimensional coordinate of all points in the grid is calculated. Then, the mean three-dimensional coordinate point is used to replace all points in the corresponding voxel grid, and finally the target point cloud data with three-level preprocessing is obtained.

[0050] Voxel mesh filtering, through uniform division of three-dimensional space and replacement of coordinate mean, significantly compresses the amount of point cloud data while preserving the core spatial contour and positional features of the point cloud to the maximum extent. This significantly reduces the computational load of subsequent core algorithms such as point cloud segmentation, 3D modeling, and trajectory prediction, while further improving the regularity and consistency of point cloud data. It can provide efficient and high-quality basic data support for the entire intelligent early warning and adaptive control process of near-electric operation of large machinery.

[0051] Step 250: Perform point cloud segmentation on the preprocessed target point cloud data.

[0052] For embodiments of this disclosure, step 250 may include the following steps: Step 250-1: Randomly select a minimum sampling set containing multiple non-collinear points from the preprocessed target point cloud data, and generate an initial planar model based on the minimum sampling set.

[0053] Among them, non-collinear points are points in three-dimensional space that are not on the same straight line, which is the basic premise for fitting an effective planar model; the minimum sampling set is the minimum number of points required to fit the planar model, which is the core input unit for realizing the point cloud segmentation algorithm; the initial planar model is a preliminary planar model calculated by the planar fitting algorithm based on the three-dimensional coordinates of points in the randomly selected minimum sampling set, which is the benchmark model for subsequent point cloud segmentation iterative optimization.

[0054] In this embodiment of the disclosure, random sampling is performed on the target point cloud data that has undergone three-level filtering preprocessing. The minimum number of points (e.g., 3) that meet the condition of not being on the same straight line are selected to form a minimum sampling set. Based on the three-dimensional spatial coordinate information of each point in the sampling set, the corresponding initial plane model is calculated by a plane fitting algorithm, which provides basic model support for the subsequent iterative optimization of point cloud segmentation.

[0055] By selecting a minimum sampling set of non-collinear points to fit the initial planar model, a basic model for planar fitting can be quickly constructed with the least amount of point cloud data, ensuring the effectiveness and rationality of the planar model. This provides an accurate initial benchmark for the iterative optimization of subsequent point cloud segmentation algorithms, improving the efficiency and accuracy of overall point cloud segmentation and laying a solid foundation for the accurate separation and 3D modeling of charged bodies and mechanical working parts.

[0056] Step 250-2: Calculate the distance from each point in the target point cloud data to the initial planar model. By comparing the distance with the adaptive distance threshold, filter the interior points in the target point cloud data. If the number of interior points in the target point cloud data is greater than the preset threshold, update the model parameters of the initial planar model. Repeat the above iterative process until the convergence condition is met to obtain the target planar model.

[0057] Among them, the adaptive distance threshold is a critical distance value that is dynamically adjusted according to the local density of the point cloud, and it is the core criterion for distinguishing between interior and exterior points; interior points are points in the target point cloud data whose distance to the initial planar model is less than the adaptive distance threshold, and they are valid point cloud data that constitute the planar model; the preset threshold is a critical value for determining whether the number of interior points needs to be updated in the planar model, and it is a key indicator for triggering model parameter iteration; the convergence condition is the criterion for stopping the iterative optimization of the planar model, which usually refers to the number of iterations reaching the dynamically calculated value or the number of interior points stabilizing; the target planar model is the optimal planar model obtained after multiple iterations and optimizations that meet the convergence condition, and it is the core basis for achieving accurate point cloud segmentation.

[0058] In this embodiment of the present disclosure, a planar model iterative optimization operation can be performed on the preprocessed target point cloud data. First, the spatial distance from each point in the target point cloud data to the initial planar model is calculated. This distance is compared with an adaptive distance threshold, and inliers with distances less than a preset threshold are selected. The number of inliers is counted and compared with a preset threshold. If the number of inliers is greater than the preset threshold, the parameters of the initial planar model are updated. The operations of calculating the distance from the point to the plane, selecting inliers, determining the number of inliers, and updating the model parameters are repeated until the convergence condition is met, and finally the target planar model for point cloud segmentation is obtained.

[0059] By calculating the distance from a point to a planar model and combining it with an adaptive distance threshold to filter interior points, and then dynamically iterating and updating the planar model parameters based on the number of interior points until convergence, the fitting accuracy of the planar model to the target point cloud can be effectively improved. At the same time, the adaptive mechanism can adapt to point cloud data of different densities, reduce the number of invalid iterations, and provide reliable model support for the subsequent accurate segmentation of point cloud clusters of charged bodies and mechanical operating parts, thereby improving the efficiency and accuracy of the overall point cloud segmentation.

[0060] Step 250-3: Based on the target plane model, divide the preprocessed target point cloud data into point cloud clusters corresponding to charged bodies and mechanical operating parts.

[0061] Among them, the point cloud clusters corresponding to the charged body and the mechanical working parts are the discrete point sets corresponding to the charged body and the mechanical working parts respectively after the target point cloud data is segmented, which can accurately reflect the spatial contour and position attributes of the two types of targets.

[0062] In this embodiment of the disclosure, the target plane model obtained after iterative optimization can be used as the criterion for point cloud segmentation. The target point cloud data that has undergone three-level filtering preprocessing can be divided into regions. Based on the spatial positional relationship between the point cloud data and the target plane model, the point clouds in different spatial regions are classified separately to form independent point cloud clusters corresponding to the charged body and the mechanical working parts, thereby completing the accurate separation of the two types of core target point clouds.

[0063] Precise segmentation of preprocessed target point cloud data based on the target plane model can quickly divide the point cloud clusters corresponding to charged bodies and mechanical operating parts, clearly defining the spatial range of the two types of core targets. This provides targeted target data support for subsequent core algorithm steps such as SIFT feature extraction, 3D model construction, and dynamic trajectory prediction, significantly improving the computational efficiency and accuracy of subsequent algorithms and laying a solid data foundation for accurate early warning and adaptive control of near-electricity operation risks.

[0064] Step 260: Construct a three-dimensional model of the charged body and mechanical working parts based on the point cloud segmentation results, and use a filtering algorithm to predict the motion trajectory of the mechanical working parts based on the three-dimensional model.

[0065] For embodiments of this disclosure, step 260 may include the following steps: Step 260-1: For the point cloud clusters corresponding to the charged body and the mechanical working parts, Gaussian difference pyramids are constructed to generate multi-scale spatial point cloud data. Based on the multi-scale spatial point cloud data, extreme points are detected and gradient directions in the neighborhood of the extreme points are calculated. Based on the gradient directions, feature descriptors corresponding to the charged body and the mechanical working parts are generated respectively.

[0066] Among them, the Difference-of-Gaussian Pyramid (DOP) is a hierarchical structure obtained by subtracting Gaussian blurred point clouds of different scales from Gaussian blurred point clouds of adjacent scales, and is the core carrier for realizing multi-scale feature detection; multi-scale spatial point cloud data is point cloud data generated by the Difference-of-Gaussian Pyramid, covering different scale levels, and can capture the core features of the target at different scale dimensions; extreme points are points in the multi-scale spatial point cloud data where the gradient value reaches a maximum or minimum state in a local range, and are the core points representing the key features of the target; gradient direction is the direction pointed to by the gradient vector of points in the neighborhood of the extreme point, and is a key parameter describing the local texture and shape features of the feature point; feature descriptor is a vector statistically generated based on parameters such as the gradient direction in the neighborhood of the extreme point, used to represent the unique features of the target, and has rotation invariance and scale invariance.

[0067] In this embodiment of the disclosure, Gaussian difference pyramids can be constructed for the point cloud clusters corresponding to the segmented charged body and mechanical working parts, respectively. Multi-scale spatial point cloud data is generated by Gaussian blurring and difference operations at different scales. Extreme points that can characterize the core features of the target are detected in the multi-scale space. Then, the gradient direction of the points in the neighborhood of each extreme point is calculated. Based on the statistical features of the gradient direction, feature descriptors corresponding to the two types of targets are generated respectively.

[0068] By constructing a Gaussian difference pyramid to generate multi-scale spatial point cloud data, target extrema can be accurately detected at different scale dimensions. The feature descriptor generated by combining the gradient direction of the neighborhood of the extrema has excellent rotation invariance and scale invariance, which can effectively characterize the core features of charged bodies and mechanical parts, greatly improve the feature matching accuracy of point clouds from different perspectives, and provide reliable feature support for subsequent 3D model construction based on the triangulation principle, thereby improving the accuracy of overall target recognition and modeling.

[0069] Step 260-2: Perform similarity matching on the feature descriptors of charged bodies and mechanical operating parts under different acquisition perspectives, establish a one-to-one correspondence between extreme points, and obtain an initial set of matching point pairs.

[0070] Among them, the feature descriptors under different acquisition perspectives are feature descriptors of charged bodies and mechanical operating parts extracted from point cloud data collected from multiple observation angles, which are used for target feature matching across perspectives; similarity matching is the process of determining whether the descriptors correspond to the same physical location extreme point by calculating the vector similarity of feature descriptors under different acquisition perspectives; the one-to-one correspondence of extreme points is the precise matching association established between feature descriptors belonging to the same physical location extreme point under different acquisition perspectives; the initial matching point pair set is the set of all extreme point pairs that have successfully established a one-to-one correspondence after similarity matching, which is the core basic data for subsequent 3D model construction.

[0071] In this embodiment of the disclosure, the feature descriptors of the segmented charged body and mechanical working parts can be classified and processed. The feature descriptors generated by the two types of targets under different acquisition perspectives are selected to carry out similarity matching operations. By calculating the vector similarity, it is determined whether the extreme points corresponding to the descriptors under different perspectives are the same physical location points. A one-to-one correspondence is established for the successfully matched extreme points. Finally, all successfully matched extreme point pairs are integrated to form an initial set of matching point pairs.

[0072] Step 260-3: Based on the RANSAC algorithm, filter the initial set of matching point pairs, remove incorrect matching pairs that do not meet the spatial geometric constraints, and retain the correct matching point pairs.

[0073] Among them, the RANSAC algorithm is a robust algorithm that iteratively fits the optimal mathematical model from a set of data containing a large amount of noise, thereby accurately distinguishing between valid and abnormal data. It is the core technical means to achieve the screening of matching point pairs. Spatial geometric constraints are the three-dimensional spatial position association rules that corresponding extreme points should satisfy under different acquisition perspectives. They are the core basis for determining whether a matching point pair is valid. Incorrect matching pairs are extreme point pairs in the initial matching point pair set that do not satisfy spatial geometric constraints, which will interfere with the accuracy of subsequent three-dimensional model construction. Correct matching pairs are extreme point pairs in the initial matching point pair set that satisfy spatial geometric constraints. They are valid data for constructing high-precision three-dimensional models of charged bodies and mechanical operating parts.

[0074] In this embodiment of the disclosure, the initial set of matching point pairs obtained by feature descriptor similarity matching can be screened based on the RANSAC algorithm. The algorithm iteratively fits a mathematical model that can characterize the spatial positional relationship of the target. Each pair of points in the initial set of matching point pairs is substituted into the model for verification. It is determined whether each pair of points meets the preset spatial geometric constraints. Incorrect matching pairs that do not meet the constraints are removed, and correct matching pairs that meet the constraints are retained, providing high-quality matching data for subsequent three-dimensional spatial coordinate calculation.

[0075] By using the RANSAC algorithm to filter the initial set of matching points, erroneous matching pairs generated during feature matching can be accurately eliminated based on spatial geometric constraints, significantly reducing the interference of noisy data on subsequent modeling stages and effectively improving the accuracy and reliability of matching point pairs.

[0076] Step 260-4: Extract the spatial position information of the correct matching point pairs under different acquisition perspectives. Based on the spatial position information, calculate the three-dimensional spatial coordinates of the extreme points using the triangulation principle. Construct a three-dimensional model of the charged body and mechanical working parts based on the three-dimensional spatial coordinates of all extreme points.

[0077] Among them, spatial location information consists of the two-dimensional pixel coordinates of the correctly matched point pairs and the pose parameters of the sensor under different acquisition perspectives, which is the core input data for calculating the three-dimensional spatial coordinates; the triangulation principle is a geometric calculation method that uses the projection relationship of the same feature point under multiple acquisition perspectives to solve the three-dimensional spatial coordinates of the feature point in reverse; the three-dimensional spatial coordinates of the extreme points are the three-dimensional position parameters of the points that characterize the key features of the charged body and the mechanical working parts, which are calculated by the triangulation principle and are the basic data for constructing the three-dimensional model; the three-dimensional model of the charged body and the mechanical working parts is a digital model that can accurately reflect the spatial contour and positional attributes of the two types of targets after integrating the three-dimensional spatial coordinates of all extreme points.

[0078] In this embodiment of the disclosure, the spatial position information of the correct matching point pairs obtained by the RANSAC algorithm under different acquisition perspectives can be extracted. Based on this information, the three-dimensional spatial coordinates corresponding to each extreme point are calculated using the triangulation principle. Then, the three-dimensional spatial coordinates of all extreme points are integrated to finally construct a three-dimensional model of the charged body and mechanical working parts that can accurately characterize the spatial features of the target.

[0079] By extracting the spatial location information of correctly matched point pairs, and combining the triangulation principle to calculate the three-dimensional spatial coordinates of extreme points and construct a three-dimensional model, the spatial contours and positional relationships of the charged body and the mechanical working parts can be accurately restored. This provides high-precision model support for subsequent calculation of the real-time actual distance between the two and prediction of the motion trajectory of the mechanical working parts, greatly improving the accuracy of risk assessment for near-electric work and the foresight of early warning and control, and ensuring the reliability of the entire safety protection process.

[0080] Step 260-5: Based on the three-dimensional model of the mechanical working component, perform continuous frame positioning on the mechanical working component to obtain its continuous three-dimensional position coordinate data within a preset time period.

[0081] Among them, continuous frame positioning is a technical operation that uses the three-dimensional model of the mechanical working part as a matching benchmark to carry out target recognition and tracking on the time-series continuous point cloud data, and determines the spatial position of the part frame by frame; the preset time period is a time interval for collecting time-series data in order to obtain the position change pattern of the mechanical working part; the continuous three-dimensional position coordinate data is a set of a series of three-dimensional position parameters of the mechanical working part obtained by continuous frame positioning within the preset time period, which can reflect the real-time position change trajectory of the part.

[0082] In this embodiment of the disclosure, a high-precision three-dimensional model of the completed mechanical working component can be used as the core matching benchmark to perform target positioning and tracking operations on the time-series continuous point cloud data, identify and match the point cloud region corresponding to the mechanical working component frame by frame, thereby obtaining a series of continuous three-dimensional position coordinate data of the component within a pre-set time period.

[0083] Step 260-6: Input the continuous three-dimensional position coordinate data into the extended Kalman filter algorithm to construct the motion state model of the mechanical working component. Based on the motion state model, generate the trajectory prediction result of the mechanical working component within a preset time in the future.

[0084] Among them, the Extended Kalman Filter (EKF) algorithm is an optimal estimation algorithm suitable for nonlinear systems. Through a closed-loop process of state prediction and update iteration, it can accurately estimate the motion state of the target and effectively reduce data noise interference. The motion state model is a mathematical model constructed by the EKF algorithm based on continuous three-dimensional position coordinate data. It can characterize the core motion parameters and change laws of mechanical operating parts, such as position and velocity. The trajectory prediction result is derived from the motion state model. It is the trajectory data of the position change of mechanical operating parts within a preset time in the future, which is a forward-looking judgment basis for the risk warning of near-electric operation.

[0085] In this embodiment of the present disclosure, continuous three-dimensional position coordinate data within a preset time period obtained from continuous frame positioning of a three-dimensional model of a mechanical working component can be input into an extended Kalman filter algorithm. The algorithm's state prediction and update iteration process filters out noise interference in the data, constructs a motion state model that can accurately characterize the motion law of the mechanical working component, and then performs inference calculations based on the model to generate trajectory prediction results of the mechanical working component within a preset time period in the future.

[0086] By processing continuous three-dimensional position coordinate data and constructing a motion state model using the extended Kalman filter algorithm, noise interference during data acquisition can be effectively filtered out, the motion patterns of mechanical operating parts can be accurately captured, and the generated trajectory prediction results have high timeliness and high accuracy.

[0087] Step 270: Based on the voltage level parameters and environmental interference parameters of the work area, the dynamic safety threshold is adaptively adjusted using a fuzzy logic reasoning algorithm.

[0088] For embodiments of this disclosure, step 270 may include the following steps: Step 270-1: Obtain the voltage level parameters and environmental interference parameters of the work area. The environmental interference parameters include light intensity, dust concentration and rain / fog humidity.

[0089] Step 270-2: Input the voltage level parameters and environmental interference parameters into the fuzzy logic reasoning model, and match the corresponding rules in the fuzzy rule base through fuzzification processing to obtain the fuzzy reasoning result.

[0090] Among them, the fuzzy logic reasoning model is an intelligent decision-making model built on fuzzy set theory, which can handle uncertain input parameters and output reasonable decision results; fuzzification is the process of converting precise input parameters such as voltage level parameters and environmental interference parameters into membership values ​​in fuzzy sets, which is a prerequisite for fuzzy reasoning; the fuzzy rule base is a pre-established set of rules containing decision logic corresponding to different combinations of voltage levels and environmental interference; the fuzzy reasoning result is the decision data adapted to the current working scenario, which is output after calculation by the fuzzy logic reasoning model and can provide a direct basis for the generation of dynamic safety thresholds.

[0091] In this embodiment of the disclosure, the voltage level parameters and environmental interference parameters of the work area can be input into a pre-constructed fuzzy logic reasoning model. These two types of precise parameters are then fuzzified and converted into membership values ​​of corresponding fuzzy sets. Based on these membership values, relevant decision rules preset in the fuzzy rule base are matched. After model calculation, a fuzzy reasoning result adapted to the current work scenario is obtained.

[0092] By inputting voltage level parameters and environmental interference parameters into a fuzzy logic inference model and completing fuzzification and rule matching, the uncertain correlation between the two types of parameters can be effectively handled. The generated fuzzy inference results can accurately support the adaptive adjustment of dynamic safety thresholds, breaking through the limitations of traditional fixed thresholds that cannot take into account different voltage levels and complex environments. This significantly improves the adaptability of safety thresholds to operating scenarios and provides a reliable decision-making basis for the accuracy of subsequent graded early warnings.

[0093] Step 270-3: Perform defuzzification processing on the fuzzy inference results and output a dynamic safety threshold that adapts to the current work scenario.

[0094] Among them, defuzzification is the process of converting the fuzzy set obtained by fuzzy inference into precise numerical values, which is a key link connecting fuzzy inference with practical application decision-making; the dynamic safety threshold is the critical value of safe distance output after defuzzification, which can adapt to the voltage level and environmental interference of the current working area, and is the core quantitative basis for judging the risk of working near power.

[0095] In this embodiment of the disclosure, the fuzzy inference result obtained by matching the voltage level parameter and the environmental interference parameter with the fuzzy rule base can be defuzzified. The decision data in the form of fuzzy set is converted into precise values ​​through a preset calculation method, and finally a dynamic safety threshold that can accurately adapt to the current working scenario is output.

[0096] Step 280: Based on the dynamic safety threshold, the actual distance between the mechanical working parts and the live conductor, the predicted trajectory of the mechanical working parts and the distance change rate, generate the corresponding level of early warning and control instructions.

[0097] For embodiments of this disclosure, step 280 may include the following steps: Step 280-1: Based on the three-dimensional model of the charged body, determine its spatial reference position, and combine it with the real-time three-dimensional position coordinates of the mechanical working parts to calculate the real-time actual distance between the mechanical working parts and the charged body.

[0098] Among them, the spatial reference position is a fixed reference position calibrated based on the three-dimensional model of the charged body, which is the core reference point for calculating the relative distance between the mechanical working part and the charged body; the real-time three-dimensional position coordinates of the mechanical working part are obtained through continuous frame positioning technology, which are the three-dimensional spatial position parameters of the mechanical working part at each moment during the operation; the real-time actual distance is the dynamic spatial straight-line distance between the two calculated based on the spatial reference position and the real-time three-dimensional position coordinates of the mechanical working part during the operation.

[0099] In this embodiment of the disclosure, the spatial reference position of the charged body can be calibrated based on the completed high-precision three-dimensional model of the charged body. This reference position is used as a fixed reference for distance calculation. Then, combined with the real-time three-dimensional position coordinates of the mechanical working parts obtained by continuous frame positioning technology, the real-time actual distance between the mechanical working parts and the charged body is obtained by spatial distance calculation method.

[0100] Step 280-2: Based on the distance deviation between the real-time actual distance and the dynamic safety threshold, determine the current basic risk level corresponding to the dynamic safety threshold range where the real-time actual distance is located.

[0101] Among them, the distance deviation value is the difference between the real-time actual distance and the dynamic safety threshold; the dynamic safety threshold range is the different risk intervals divided based on the dynamic safety threshold, and each interval corresponds to a different basic risk level; the current basic risk level is the initial risk level determined after matching the distance deviation value with the dynamic safety threshold range.

[0102] In this embodiment of the disclosure, the distance deviation between the real-time actual distance between the mechanical working part and the live conductor and the dynamic safety threshold can be calculated. The deviation value is then matched with a preset dynamic safety threshold range, and the current basic risk level corresponding to the interval where the real-time actual distance is located is determined based on the matching result.

[0103] By calculating the distance deviation between the real-time actual distance and the dynamic safety threshold and matching the corresponding threshold range, the basic risk level can be determined. This allows for the precise definition of the initial risk level during the operation process using quantitative indicators. It provides a reliable basis for subsequent multi-parameter precise graded early warning based on trajectory prediction results and distance change rate, thereby improving the scientific nature and accuracy of near-power operation risk early warning.

[0104] Step 280-3: Based on the continuous position coordinate data of the mechanical working parts within a preset time period, calculate the rate of change of distance between the mechanical working parts and the charged body.

[0105] In this embodiment of the present disclosure, based on the continuous position coordinate data of the mechanical working component within a preset time period, combined with the spatial reference position calibrated by the charged body, the real-time actual distance between the mechanical working component and the charged body at each moment within that time period can be calculated sequentially. Then, by calculating the ratio of the distance difference between adjacent moments to the time interval, the distance change rate between the mechanical working component and the charged body can be obtained. By calculating the distance change rate between the mechanical working component and the charged body, the relative motion trend between the two can be accurately quantified, and it can be determined whether the mechanical working component is moving closer to or away from the charged body.

[0106] Step 280-4: Based on the motion trajectory prediction results of the mechanical operating components, determine whether they will intrude into the dangerous critical range corresponding to the dynamic safety threshold within a preset time in the future, and determine the potential risk level based on the judgment results.

[0107] Among them, the future preset time is the trajectory projection time interval set in advance to predict risks, and is a key time parameter for achieving forward-looking early warning; the danger critical range is the safety protection range defined based on the dynamic safety threshold, and is the core range basis for determining whether there is a risk of intrusion of mechanical operating parts; the potential risk level is the risk level determined based on the judgment result of whether the future trajectory of mechanical operating parts intrudes into the danger critical range, and is an important reference dimension for graded early warning.

[0108] In this embodiment of the present disclosure, the trajectory prediction result of the mechanical working part generated by the extended Kalman filter algorithm can be compared with the dangerous critical range corresponding to the dynamic safety threshold to determine whether the position of the mechanical working part will enter the dangerous critical range within a preset time in the future, and then the corresponding potential risk level can be determined based on the judgment result.

[0109] By combining the motion trajectory prediction results of mechanical operating components to determine the potential risk level, it is possible to predict in advance whether the components will enter the dangerous critical range, giving the risk warning link a strong foresight and effectively avoiding the warning lag problem caused by traditional static distance determination. When combined with the basic risk level, it can further improve the accuracy and reliability of graded warnings, and provide key decision support for the proactive prevention and control of large machinery near electric fields.

[0110] Step 280-5: Determine the target risk level based on the current basic risk level, potential risk level, and distance change rate, and generate early warning and control instructions corresponding to the target risk level. The target risk level can be any one of prompt, early warning, and alarm. The early warning and control instructions can be any one of prompt-level instructions, early warning-level instructions, and alarm-level instructions.

[0111] The target risk level is the final risk level determined by comprehensively considering three parameters: the current basic risk level, the potential risk level, and the distance change rate. It is divided into three levels: alert, warning, and alarm. The alert level instruction is a control instruction corresponding to the alert level target risk level, used to remind operators to pay attention to the work status without stopping the work. The warning level instruction is a control instruction corresponding to the warning level target risk level, used to warn operators that the mechanical working parts are approaching the danger zone and that the operator's posture needs to be adjusted. The alarm level instruction is a control instruction corresponding to the alarm level target risk level, used to forcibly remind operators that the mechanical working parts are about to or have already entered the danger zone and that the work must be stopped immediately.

[0112] In this embodiment of the disclosure, three core parameters—the current basic risk level, the potential risk level, and the distance change rate—can be input into a multi-parameter early warning judgment model. The model performs calculations based on preset comprehensive judgment rules to determine the final target risk level. Then, based on the target risk level, a prompt-level, early warning-level, or alarm-level early warning control instruction that is precisely matched to it is generated.

[0113] By comprehensively determining the target risk level and generating corresponding instructions based on three core parameters—the current basic risk level, the potential risk level, and the distance change rate—a multi-dimensional and three-dimensional risk assessment can be achieved. This effectively avoids misjudgments or delayed early warnings caused by single-parameter assessments. Differentiated early warning and control instructions can guide operators to take precise countermeasures, promoting the transformation of large machinery near-electricity operations from passive early warning to proactive prevention and control, and comprehensively improving the safety protection level of the operation process.

[0114] Step 290: Trigger the adaptive control action of large machinery according to the early warning and control instructions to achieve proactive avoidance of operational risks.

[0115] For embodiments of this disclosure, step 290 may include the following steps: Step 290-1: If the early warning and control command is a prompt level command, then a speed adjustment command is sent to the large machinery control system to indicate the movement speed of the mechanical working parts, and the control status is fed back to the terminal interaction layer.

[0116] Among them, the large-scale machinery control system is the core control unit responsible for receiving external commands and controlling the motion state of mechanical working parts, and is the carrier of command execution; the control status refers to the execution status of the speed adjustment of mechanical working parts and the feedback information of the current motion state after the speed adjustment command is issued; the terminal interaction layer is the core unit for realizing human-machine interaction, which can display data such as early warning information and control status, and support operators to perform operations such as parameter adjustment.

[0117] In this embodiment of the disclosure, when the warning control command output by the warning layer is a prompt level command, a speed adjustment command can be sent to the large machinery control system to indicate the fine adjustment of the movement speed of the mechanical working parts. At the same time, the issuance of the speed adjustment command and the control status of the speed adjustment of the mechanical working parts are fed back to the terminal interaction layer in real time.

[0118] By issuing speed adjustment commands at the prompt level and providing feedback on the control status, the machine can gently intervene in the mechanical operation process by fine-tuning the speed. This reduces the risk of mechanical parts approaching live conductors without interrupting normal operations. At the same time, the status feedback from the terminal interaction layer allows operators to monitor the machine's control status in real time, improving the accuracy and proactivity of safety protection for near-electric work, thus balancing work efficiency and safety.

[0119] Step 290-2: If the warning control instruction is a warning level instruction, a speed reduction instruction is sent to the large machinery control system to indicate the reduction of the movement speed of the mechanical working parts. At the same time, the sound and light warning module is triggered to issue a warning signal, and the warning status and speed reduction information are displayed on the terminal interaction layer.

[0120] The sound and light warning module is a hardware unit used to emit sound and light warning signals, which can quickly remind on-site personnel to pay attention to risks; the warning status is a comprehensive information of the execution of the speed reduction command and the working status of the sound and light warning module after the warning level command is issued.

[0121] In this embodiment of the present disclosure, when the warning control command is determined to be a warning level command, a deceleration command can be sent to the large machinery control system to indicate a reduction in the movement speed of the mechanical working parts. At the same time, the sound and light warning module is triggered to issue a corresponding warning signal, and the warning status and deceleration information are displayed in real time on the terminal interaction layer.

[0122] By issuing a speed reduction command, the speed at which mechanical working parts approach live conductors can be proactively slowed down, allowing buffer time for risk handling. Simultaneously triggered audible and visual warning signals can quickly alert on-site personnel to the risks. The information display at the terminal interaction layer allows managers to monitor the equipment control status in real time. These multiple measures work together to achieve proactive intervention and visual control of risks, effectively reducing the probability of mechanical working parts intruding into dangerous areas and improving the initiative and reliability of safety protection for near-electric work.

[0123] Step 290-3: If the early warning and control instruction is an alarm level instruction, then an emergency stop instruction is sent to the large machinery control system to instruct the control machinery operating parts to stop moving immediately, so as to avoid intruding into the safety threshold range of the live parts. At the same time, the stop status and alarm information are pushed to the terminal interaction layer.

[0124] Among them, the safety threshold range of the live part is a safety protection range dynamically generated based on the fuzzy logic reasoning algorithm, which is the core critical basis for determining whether to trigger an emergency shutdown; the shutdown status is the operating status information of the mechanical working parts after the emergency shutdown command is issued, including whether the shutdown was successful and the current position; the alarm information is warning data containing the current risk level, shutdown reason and part location, used to inform the operators of high-risk situations on site.

[0125] In this embodiment of the present disclosure, when the early warning control command is determined to be an alarm level command, an emergency stop command can be issued to the large machinery control system to instruct the control machinery operating parts to stop moving immediately, thereby preventing the parts from entering the safety threshold range of the live body. At the same time, the stop status of the machinery operating parts and detailed alarm information are pushed to the terminal interaction layer in real time.

[0126] By issuing an emergency shutdown command, the movement of mechanical operating parts can be directly interrupted, fundamentally preventing external damage accidents caused by their intrusion into the safety threshold range of energized bodies. At the same time, the shutdown status and alarm information push of the terminal interaction layer can enable operators to grasp the high-risk situation on site in time, which facilitates subsequent risk handling and equipment debugging, greatly improving the safety protection level of large machinery working near electric fields, and ensuring the safety of power grid operation and personnel life.

[0127] The technical solution in this application, by simultaneously acquiring 3D point cloud and image data and performing feature fusion, combined with three-level point cloud preprocessing, segmentation modeling, and filtering trajectory prediction algorithms, can effectively improve the identification accuracy and ranging accuracy of charged bodies and mechanical operating parts, solving the problems of traditional single sensors being susceptible to environmental interference and the existence of visual blind spots and reaction delays in manual monitoring; based on voltage level and environmental interference parameters, a fuzzy logic reasoning algorithm is used to adaptively adjust the dynamic safety threshold, which can overcome the limitation that fixed thresholds cannot adapt to complex operating scenarios; and by combining the actual distance, trajectory prediction results, and distance change rate to generate graded early warning and control instructions, and triggering adaptive control actions of large machinery, it can achieve an upgrade from passive early warning to active prevention and control, which can not only avoid the impact of excessive early warning on operating efficiency, but also eliminate the safety hazards caused by delayed early warning, ultimately fundamentally improving the safety protection level of large machinery operating near electric fields.

[0128] Furthermore, as Figure 1 and Figure 2 The specific implementation of the method shown in this embodiment provides a large-scale mechanical near-electric operation control device, such as... Figure 3 As shown, the device includes: a data acquisition module 31, a first processing module 32, an adjustment module 33, a generation module 34, and a control module 35.

[0129] The acquisition module 31 can be used to simultaneously acquire three-dimensional spatial point cloud data and image data of the work area, and perform feature matching and fusion of three-dimensional spatial point cloud data and image data through data fusion algorithm to obtain target point cloud data with texture features; The first processing module 32 can be used to perform point cloud segmentation on the preprocessed target point cloud data, construct a three-dimensional model of the charged body and mechanical working parts based on the point cloud segmentation results, and use a filtering algorithm to predict the motion trajectory of the mechanical working parts based on the three-dimensional model. Adjustment module 33 can be used to adaptively adjust dynamic safety thresholds based on voltage level parameters and environmental interference parameters of the work area through fuzzy logic reasoning algorithm; The generation module 34 can be used to generate early warning and control instructions of corresponding levels based on dynamic safety thresholds, the actual distance between mechanical working parts and live conductors, the predicted motion trajectory of mechanical working parts and the distance change rate. The control module 35 can be used to trigger adaptive control actions of large machinery according to early warning control instructions, so as to achieve proactive avoidance of operational risks.

[0130] In some embodiments of this application, such as Figure 4 As shown, the device also includes: a second processing module 36, a third processing module 37, and a fourth processing module 38; The second processing module 36 can be used to perform block processing on the target point cloud data based on the improved statistical filtering algorithm. It dynamically determines the neighborhood radius of each point according to the local density of the point cloud in each sub-block. Based on the neighborhood radius, it calculates the mean distance and standard deviation of the distance between each point and multiple neighboring points in the corresponding neighborhood. It determines the outlier judgment threshold corresponding to the mean distance and standard deviation of the distance. It removes outlier noise points in the target point cloud data whose mean distance is greater than the outlier judgment threshold to obtain the first point cloud data. The third processing module 37 can be used to perform radius filtering on the first point cloud data to obtain the second point cloud data after removing isolated noise points. The radius filtering is used to count the number of neighboring points within a fixed neighborhood radius for each point and remove isolated noise points whose corresponding number of neighboring points is less than the minimum neighboring point threshold. The fourth processing module 38 can be used to perform voxel grid filtering on the second point cloud data. It divides the three-dimensional space where the point cloud is located into a uniform voxel grid according to a preset voxel size, calculates the mean three-dimensional coordinates of all points in each voxel grid, and uses the mean three-dimensional coordinates to replace all points in the voxel grid to obtain the preprocessed target point cloud data.

[0131] In some embodiments of this application, the first processing module 32 can be specifically used to randomly select a minimum sampling set containing multiple non-collinear points from the preprocessed target point cloud data, and generate an initial plane model based on the minimum sampling set; calculate the distance from each point in the target point cloud data to the initial plane model, and filter interior points in the target point cloud data by comparing the distance with an adaptive distance threshold. If the number of interior points in the target point cloud data is greater than a preset threshold, update the model parameters of the initial plane model, and repeat the above iterative process until the convergence condition is met to obtain the target plane model; based on the target plane model, divide the preprocessed target point cloud data into point cloud clusters corresponding to charged bodies and mechanical operating parts.

[0132] In some embodiments of this application, the first processing module 32 can also be used to generate multi-scale spatial point cloud data by constructing Gaussian difference pyramids for the point cloud clusters corresponding to the charged body and the mechanical working parts, respectively; detect extreme points based on the multi-scale spatial point cloud data and calculate the gradient direction in the neighborhood of the extreme points; generate feature descriptors corresponding to the charged body and the mechanical working parts respectively based on the gradient direction; perform similarity matching on the feature descriptors of the charged body and the mechanical working parts under different acquisition perspectives to establish a one-to-one correspondence between extreme points and obtain an initial set of matching point pairs; and filter the initial set of matching point pairs based on the RANSAC algorithm to remove those that do not conform to spatial geometric constraints. Incorrect matching pairs are identified, while correct matching pairs are retained. Spatial position information of correct matching pairs under different acquisition perspectives is extracted. Based on the spatial position information, the three-dimensional spatial coordinates of extreme points are calculated using the triangulation principle. A three-dimensional model of the charged body and mechanical operating parts is constructed based on the three-dimensional spatial coordinates of all extreme points. Based on the three-dimensional model of the mechanical operating parts, continuous frame positioning of the mechanical operating parts is performed to obtain its continuous three-dimensional position coordinate data within a preset time period. The continuous three-dimensional position coordinate data is input into the extended Kalman filter algorithm to construct the motion state model of the mechanical operating parts. Based on the motion state model, the trajectory prediction result of the mechanical operating parts within a preset time period is generated.

[0133] In some embodiments of this application, the adjustment module 33 can be specifically used to obtain voltage level parameters and environmental interference parameters of the work area, including light intensity, dust concentration and rain / fog humidity; input the voltage level parameters and environmental interference parameters into the fuzzy logic inference model, match the corresponding rules in the fuzzy rule base through fuzzification processing to obtain the fuzzy inference result; perform defuzzification processing on the fuzzy inference result, and output a dynamic safety threshold adapted to the current work scenario.

[0134] In some embodiments of this application, the generation module 34 can be specifically used to calibrate the spatial reference position of the charged body based on the three-dimensional model of the charged body, and calculate the real-time actual distance between the mechanical working part and the charged body by combining the real-time three-dimensional position coordinates of the mechanical working part; determine the current basic risk level corresponding to the dynamic safety threshold range where the real-time actual distance is located based on the distance deviation value between the real-time actual distance and the dynamic safety threshold; calculate the distance change rate between the mechanical working part and the charged body based on the continuous position coordinate data of the mechanical working part within a preset time period; determine whether the mechanical working part will invade the dangerous critical range corresponding to the dynamic safety threshold within a preset time period based on the motion trajectory prediction result of the mechanical working part, and determine the potential risk level based on the judgment result; determine the target risk level based on the current basic risk level, the potential risk level and the distance change rate, and generate a warning and control instruction corresponding to the target risk level. The target risk level is any one of prompt, warning and alarm, and the warning and control instruction includes any one of prompt-level instruction, warning-level instruction and alarm-level instruction.

[0135] In some embodiments of this application, the control module 35 can be specifically used to: if the warning control instruction is a prompt level instruction, send a speed adjustment instruction to the large machinery control system to instruct the fine adjustment of the movement speed of the mechanical working parts, and simultaneously feed back the control status to the terminal interaction layer; if the warning control instruction is a warning level instruction, send a speed reduction instruction to the large machinery control system to instruct the reduction of the movement speed of the mechanical working parts, simultaneously trigger the sound and light warning module to issue a warning signal, and display the warning status and speed reduction information on the terminal interaction layer; if the warning control instruction is an alarm level instruction, send an emergency stop instruction to the large machinery control system to instruct the control of the mechanical working parts to immediately stop moving, so as to avoid intruding into the safety threshold range of the live conductor, and simultaneously push the stop status and alarm information to the terminal interaction layer.

[0136] It should be noted that other corresponding descriptions of the functional units involved in the large-scale machinery near-electric operation control device provided in this embodiment can be found in [reference]. Figure 1 and Figure 2 The corresponding descriptions in [the document] will not be repeated here.

[0137] Based on the above, Figure 1 and Figure 2 Accordingly, this embodiment also provides a storage medium storing a computer program that, when executed by a processor, implements the above-described method. Figure 1 and Figure 2 The method for controlling near-electric operation of large machinery is shown.

[0138] Based on this understanding, the technical solution of this application can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (such as CD-ROM, USB flash drive, mobile hard drive, etc.) and includes several instructions to cause an electronic device (such as personal computer, server, or network device, etc.) to execute the methods of various implementation scenarios of this application.

[0139] Based on the above, Figure 1 and Figure 2 The method shown, and Figure 3 , Figure 4 To achieve the above objectives, the present application also provides an electronic device, specifically a personal computer, tablet computer, server, or other network device, as shown in the virtual device embodiment. This device includes a storage medium and a processor; the storage medium stores a computer program; the processor executes the computer program to achieve the above-described objectives. Figure 1 and Figure 2 The method for controlling near-electric operation of large machinery is shown.

[0140] Optionally, the aforementioned physical devices may also include a user interface, a network interface, a camera, radio frequency (RF) circuitry, sensors, audio circuitry, a Wi-Fi module, etc. The user interface may include a display screen, input units such as a keyboard, etc., and optional user interfaces may also include USB interfaces, card reader interfaces, etc. The network interface may optionally include standard wired interfaces, wireless interfaces (such as Wi-Fi interfaces), etc.

[0141] Those skilled in the art will understand that the physical device structure provided in this embodiment does not constitute a limitation on the physical device, and may include more or fewer components, or combine certain components, or have different component arrangements.

[0142] The storage medium may also include an operating system and a network communication module. The operating system is a program that manages the hardware and software resources of the aforementioned physical device, supporting the operation of information processing programs and other software and / or programs. The network communication module is used to enable communication between the various components within the storage medium, as well as communication with other hardware and software in the information processing physical device.

[0143] Through the above description of the embodiments, those skilled in the art can clearly understand that this application can be implemented by means of software plus necessary general-purpose hardware platform, or it can be implemented by hardware.

[0144] This invention, through simultaneous acquisition of 3D point cloud and image data and feature fusion, combined with three-level point cloud preprocessing, segmentation modeling, and filtering trajectory prediction algorithms, can effectively improve the identification accuracy and ranging accuracy of charged bodies and mechanical operating parts. It solves the problems of traditional single sensors being susceptible to environmental interference and the visual blind spots and reaction delays inherent in manual monitoring. Based on voltage levels and environmental interference parameters, a fuzzy logic reasoning algorithm adaptively adjusts the dynamic safety threshold, overcoming the limitation of fixed thresholds being unsuitable for complex operating scenarios. Furthermore, by combining actual distance, trajectory prediction results, and distance change rate to generate graded early warning and control commands, and triggering adaptive control actions on large machinery, it achieves an upgrade from passive early warning to proactive prevention and control. This avoids the impact of excessive early warning on operational efficiency and eliminates the safety hazards caused by delayed early warnings, ultimately fundamentally improving the safety protection level of large machinery operating near electric fields.

[0145] Those skilled in the art will understand that the accompanying drawings are merely schematic diagrams of a preferred embodiment, and the modules or processes shown in the drawings are not necessarily essential for implementing this application. Those skilled in the art will understand that the modules in the apparatus of the embodiment can be distributed within the apparatus of the embodiment as described, or can be modified to be located in one or more apparatuses different from this embodiment. The modules of the above-described embodiment can be combined into one module, or further divided into multiple sub-modules.

[0146] The serial numbers in this application are for descriptive purposes only and do not represent the superiority or inferiority of any particular implementation scenario. The above disclosures are merely a few specific implementation scenarios of this application; however, this application is not limited thereto, and any variations conceived by those skilled in the art should fall within the protection scope of this application.

Claims

1. A method for controlling near-electric operation of large machinery, characterized in that, include: Simultaneously collect 3D spatial point cloud data and image data of the work area, and use a data fusion algorithm to perform feature matching and fusion of the 3D spatial point cloud data and the image data to obtain target point cloud data with texture features; The preprocessed target point cloud data is segmented into points. Based on the point cloud segmentation results, a three-dimensional model of the charged body and the mechanical working parts is constructed. A filtering algorithm is used to predict the motion trajectory of the mechanical working parts based on the three-dimensional model. Based on the voltage level parameters and environmental interference parameters of the work area, the dynamic safety threshold is adaptively adjusted through a fuzzy logic reasoning algorithm. Based on the dynamic safety threshold, the actual distance between the mechanical working part and the charged body, the predicted motion trajectory of the mechanical working part and the distance change rate, a corresponding level of early warning and control command is generated. The early warning and control instructions trigger the adaptive control actions of large machinery to proactively avoid operational risks.

2. The method according to claim 1, characterized in that, Before performing point cloud segmentation and feature extraction on the preprocessed target point cloud data, the method further includes: The target point cloud data is divided into blocks based on an improved statistical filtering algorithm. The neighborhood radius of each point is dynamically determined according to the local density of the point cloud in each sub-block. The mean distance and standard deviation of the distance between each point and multiple neighboring points in the corresponding neighborhood are calculated based on the neighborhood radius. The outlier judgment threshold corresponding to the mean distance and the standard deviation of the distance are determined. Outlier noise points with a mean distance greater than the outlier judgment threshold are removed from the target point cloud data to obtain the first point cloud data. Radius filtering is performed on the first point cloud data to obtain the second point cloud data after removing isolated noise points. The radius filtering is used to count the number of neighboring points within a fixed neighborhood radius for each point and remove isolated noise points whose number of neighboring points is less than the minimum neighboring point threshold. The second point cloud data is subjected to voxel grid filtering processing. The three-dimensional space where the point cloud is located is divided into uniform voxel grids according to a preset voxel size. The mean three-dimensional coordinates of all points in each voxel grid are calculated. The mean three-dimensional coordinates are used to replace all points in the voxel grid to obtain the preprocessed target point cloud data.

3. The method according to claim 1, characterized in that, The point cloud segmentation process for the preprocessed target point cloud data includes: In the preprocessed target point cloud data, a minimum sampling set containing multiple non-collinear points is randomly selected, and an initial planar model is generated based on the minimum sampling set. Calculate the distance from each point in the target point cloud data to the initial plane model. By comparing the distance with an adaptive distance threshold, filter inliers in the target point cloud data. If the number of inliers in the target point cloud data is greater than a preset threshold, update the model parameters of the initial plane model. Repeat the above iterative process until the convergence condition is met to obtain the target plane model. Based on the target plane model, the preprocessed target point cloud data is divided into point cloud clusters corresponding to charged bodies and mechanical operating parts.

4. The method according to claim 3, characterized in that, The process of constructing a 3D model of the charged body and mechanical working parts based on point cloud segmentation results, and predicting the motion trajectory of the mechanical working parts based on the 3D model using a filtering algorithm, includes: For the point cloud clusters corresponding to the charged body and the mechanical working parts, a Gaussian difference pyramid is constructed to generate multi-scale spatial point cloud data. Based on the multi-scale spatial point cloud data, extreme points are detected and the gradient direction in the neighborhood of the extreme points is calculated. Based on the gradient direction, feature descriptors corresponding to the charged body and the mechanical working parts are generated respectively. The feature descriptors of the charged body and the mechanical working parts under different acquisition perspectives are matched for similarity, and a one-to-one correspondence between extreme points is established to obtain an initial set of matching point pairs. The initial set of matching point pairs is filtered based on the RANSAC algorithm to remove incorrect matching pairs that do not meet the spatial geometric constraints and retain the correct matching point pairs. Extract the spatial position information of the correct matching point pair under different acquisition perspectives, calculate the three-dimensional spatial coordinates of the extreme points based on the spatial position information and the triangulation principle, and construct a three-dimensional model of the charged body and the mechanical working parts based on the three-dimensional spatial coordinates of all extreme points; Based on the three-dimensional model of the mechanical working component, the mechanical working component is continuously frame-localized to obtain its continuous three-dimensional position coordinate data within a preset time period. The continuous three-dimensional position coordinate data is input into the extended Kalman filter algorithm to construct the motion state model of the mechanical working component. Based on the motion state model, the trajectory prediction result of the mechanical working component within a preset time period is generated.

5. The method according to claim 1, characterized in that, The adaptive adjustment of the dynamic safety threshold based on the voltage level parameters and environmental interference parameters of the work area using a fuzzy logic reasoning algorithm includes: Obtain voltage level parameters and environmental interference parameters of the work area, including light intensity, dust concentration, and rain / fog humidity; The voltage level parameters and the environmental interference parameters are input into the fuzzy logic reasoning model, and the corresponding rules in the fuzzy rule base are matched through fuzzification processing to obtain the fuzzy reasoning result; The fuzzy inference result is defuzzified, and a dynamic safety threshold adapted to the current working scenario is output.

6. The method according to claim 1, characterized in that, Based on the dynamic safety threshold, the actual distance between the mechanical working component and the charged body, the predicted motion trajectory of the mechanical working component, and the distance change rate, a corresponding level of early warning and control command is generated, including: Based on the three-dimensional model of the charged body, its spatial reference position is calibrated, and combined with the real-time three-dimensional position coordinates of the mechanical working component, the real-time actual distance between the mechanical working component and the charged body is calculated. Based on the distance deviation between the real-time actual distance and the dynamic safety threshold, the current basic risk level corresponding to the dynamic safety threshold range where the real-time actual distance is located is determined; Based on the continuous position coordinate data of the mechanical working component within a preset time period, the rate of change of the distance between the mechanical working component and the charged body is calculated. Based on the motion trajectory prediction results of the mechanical operating components, it is determined whether they will enter the dangerous critical range corresponding to the dynamic safety threshold within a preset time in the future, and the potential risk level is determined based on the judgment results. Based on the current basic risk level, the potential risk level, and the distance change rate, a target risk level is determined, and an early warning and control instruction corresponding to the target risk level is generated. The target risk level is any one of prompt, early warning, and alarm. The early warning and control instruction includes any one of prompt-level instruction, early warning-level instruction, and alarm-level instruction.

7. The method according to claim 1, characterized in that, The step of triggering adaptive control actions of large machinery according to the early warning control command to achieve proactive avoidance of operational risks includes: If the warning and control command is a prompt level command, a speed adjustment command is sent to the large machinery control system to indicate the fine adjustment of the movement speed of the mechanical working parts, and the control status is fed back to the terminal interaction layer at the same time. If the warning control instruction is a warning level instruction, a deceleration instruction is sent to the large machinery control system to indicate a reduction in the movement speed of the mechanical working parts, and the sound and light warning module is triggered to issue a warning signal. The warning status and deceleration information are displayed on the terminal interaction layer. If the warning and control command is an alarm level command, an emergency stop command is issued to the large machinery control system to instruct the mechanical operating parts to stop moving immediately, so as to avoid intruding into the safety threshold range of the live body. At the same time, the stop status and alarm information are pushed to the terminal interaction layer.

8. A large-scale mechanical near-electric operation control device, characterized in that, include: The acquisition module is used to simultaneously acquire three-dimensional spatial point cloud data and image data of the work area, and to perform feature matching and fusion of the three-dimensional spatial point cloud data and the image data through a data fusion algorithm to obtain target point cloud data with texture features. The first processing module is used to perform point cloud segmentation on the preprocessed target point cloud data, construct a three-dimensional model of the charged body and mechanical working parts based on the point cloud segmentation results, and use a filtering algorithm to predict the motion trajectory of the mechanical working parts based on the three-dimensional model. The adjustment module is used to adaptively adjust the dynamic safety threshold based on the voltage level parameters and environmental interference parameters of the work area using a fuzzy logic reasoning algorithm. The generation module is used to generate early warning and control instructions of corresponding levels based on the dynamic safety threshold, the actual distance between the mechanical working part and the charged body, the predicted motion trajectory of the mechanical working part and the distance change rate. The control module is used to trigger adaptive control actions of large machinery according to the early warning control instructions, so as to achieve proactive avoidance of operational risks.

9. A storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the method of any one of claims 1 to 7.

10. An electronic device, comprising a storage medium, a processor, and a computer program stored on the storage medium and executable on the processor, characterized in that, When the processor executes the computer program, it implements the method of any one of claims 1 to 7.

Images