A method and system for identifying high-risk areas in the air and reconstructing hoisting paths

Abstract

This application relates to the field of tower crane control technology, and provides a method and system for identifying high-risk areas in the air and reconstructing hoisting paths. The method includes: installing two rotating radars on the rotatable support of the tower crane to form a radar sensing module containing dual radar modules; dynamically scanning potential risk areas in the construction scene using the radar sensing module to obtain real-time location information of the potential risk areas; receiving the real-time location information through a processing module and converting it into real-time three-dimensional coordinate data in the motion coordinate system of the hoisting equipment using a three-dimensional coordinate transformation algorithm; constructing a three-dimensional model of the potential risk areas based on the real-time three-dimensional coordinate data; identifying high-risk areas by recognizing the three-dimensional model; and using an improved fast exploration random tree algorithm as a path avoidance algorithm, combined with the three-dimensional model, dynamically reconstructing the initial hoisting path set by the user to generate an optimal avoidance hoisting trajectory that avoids high-risk areas.

Description

Technical Field

[0001] This application relates to the field of tower crane control technology, and more specifically, to a method and system for identifying high-risk areas in the air and reconstructing hoisting paths. Background Technology

[0002] With the acceleration of urbanization and the increase in large-scale projects, hoisting operations, as a core construction process, are facing increasingly complex working environments, such as remote control of port container cranes. Scenarios involving multiple tower cranes operating collaboratively, overhead cables, and overlapping of crane boom slewing ranges with surrounding buildings or equipment have become commonplace. Identifying and avoiding high-risk areas in the air, such as tower boom intersection zones, cable interference zones, and crane boom slewing danger zones, has become crucial for ensuring operational safety.

[0003] In related technologies, risk control and path planning during hoisting operations often rely on single-view (e.g., only overhead or only under-the-horizontal) sensing devices (such as ordinary cameras or single radars). These devices are susceptible to blind spots due to dust and obstructions (such as building components and scaffolding) in the construction environment. For example, an overhead view cannot accurately capture the three-dimensional position of low-altitude overhead cables, while an under-the-horizontal view cannot fully cover the dynamically overlapping areas of multiple tower arms, leading to missed or misjudged high-risk areas and creating potential collision hazards. Therefore, a technical solution is urgently needed to address at least one of these technical problems. Summary of the Invention

[0004] In this context, the embodiments of this application aim to provide a method and system for identifying high-risk areas in the air and reconstructing hoisting paths, which can improve the autonomous judgment and safe scheduling capabilities of hoisting operations in complex air environments. It is applicable to high-risk operation scenarios such as multi-tower collaboration and intensive construction, and solves the technical problems of missed or misjudged high-risk areas and safety hazards in the above-mentioned solutions.

[0005] In a first aspect of this application, a method for identifying high-risk areas in the air and reconstructing hoisting paths is provided, comprising: installing two rotating radars on the rotatable support of a tower crane to form a radar sensing module containing a dual-radar module, wherein the radar sensing module is communicatively connected to a processing module; using the radar sensing module, with the first rotating radar in the dual-radar module as an upward-looking radar and the second rotating radar as a downward-looking radar, and based on a perception strategy combining the upward-looking and downward-looking radars, dynamically scanning the tower arm intersection area, overhead cables, and the swing range of the crane arm in the construction scene to obtain real-time location information of the potential risk areas; receiving the real-time location information and using a three-dimensional coordinate transformation algorithm to convert the real-time location information into real-time three-dimensional coordinate data in the motion coordinate system of the hoisting equipment; constructing a three-dimensional model of the potential risk areas based on the real-time three-dimensional coordinate data; identifying high-risk areas by recognizing the three-dimensional model, and using an improved fast exploration random tree algorithm as a path obstacle avoidance algorithm, and combining the three-dimensional model to dynamically reconstruct the initial hoisting path set by the user to generate an optimal avoidance hoisting trajectory that avoids the high-risk areas.

[0006] In a second aspect of the embodiments of this application, a high-risk area identification and hoisting path reconstruction system is provided, comprising: a construction module for mounting two rotating radars on the rotatable support of a tower crane to form a radar sensing module containing a dual-radar module, the radar sensing module being communicatively connected to a processing module; the radar sensing module for dynamically scanning the tower arm intersection area, overhead cables, and crane boom rotation range in the construction scene using a combined sensing strategy of the first rotating radar in the dual-radar module as an upward-looking radar and the second rotating radar as a downward-looking radar, to obtain real-time location information of the potential risk area; and a processing module for receiving the real-time location information, converting the real-time location information into real-time three-dimensional coordinate data consistent with the motion coordinate system of the hoisting equipment using a three-dimensional coordinate transformation algorithm; constructing a three-dimensional model of the potential risk area based on the real-time three-dimensional coordinate data; identifying high-risk areas by recognizing the three-dimensional model, and using an improved fast exploration random tree algorithm as a path obstacle avoidance algorithm, dynamically reconstructing the initial hoisting path set by the user in conjunction with the three-dimensional model to generate an optimal avoidance hoisting trajectory that avoids the high-risk area.

[0007] According to an embodiment of this application, a method and system for identifying high-risk areas in the air and reconstructing lifting paths involves installing two rotating radars on the rotatable support of a tower crane, forming a radar perception module with dual radar modules. The radar perception module is communicatively connected to a processing module built upon an AI-optimized operating system and function library, integrating computer vision processing technology to enhance the collaborative efficiency of perception and computation. Through the radar perception module, using the first rotating radar in the dual radar modules as an upward-looking radar and the second rotating radar as a downward-looking radar, a perception strategy combining upward and downward-looking radars is employed to dynamically scan the tower arm intersection area, overhead cables, and the slewing range of the crane boom in the construction scenario, acquiring real-time location information of the potential risk areas. Then, the real-time location information is received and converted into real-time three-dimensional coordinate data in the lifting equipment's motion coordinate system using a three-dimensional coordinate transformation algorithm. Next, a three-dimensional model of the potential risk area is constructed based on the real-time three-dimensional coordinate data. Finally, the high-risk areas are identified by the 3D model, and an improved fast exploration random tree algorithm is used as the path obstacle avoidance algorithm. Combined with the 3D model, the initial hoisting path set by the user is dynamically reconstructed to generate the optimal avoidance hoisting trajectory that avoids the high-risk areas. This embodiment of the application, through the deep fusion of collaborative perception of dual radar modules and intelligent algorithms, dynamically reconstructs the local path where the hoisting equipment is currently located in the optimal avoidance hoisting trajectory to obtain the optimal avoidance hoisting trajectory. This ensures the stability and safety of hoisting operations, improves the dynamic perception accuracy and real-time obstacle avoidance capability of obstacles in complex construction scenarios, enhances hoisting path planning and operational efficiency, and meets the safety and efficiency requirements of high-altitude cross-operations, multi-equipment collaboration, and harsh environments. In practical applications, this embodiment of the application is suitable for remote control scenarios of hoisting equipment such as tower cranes and port container cranes, and can be deeply adapted to the application requirements of remote control systems for port container cranes. Attached Figure Description

[0008] Figure 1 This is a flowchart illustrating a method for identifying high-risk areas in the air and reconstructing hoisting paths, as shown in this application.

[0009] Figure 2 This is a schematic diagram illustrating a high-altitude hoisting scenario as shown in this application;

[0010] Figure 3 This is a schematic diagram of a dual-radar cooperative sensing hardware system shown in this application. Detailed Implementation

[0011] The following is for reference. Figure 1 , Figure 1This is a flowchart illustrating a method for identifying high-risk areas in the air and reconstructing lifting paths, provided as an embodiment of this application. It should be noted that the implementation methods of this application can be applied to any applicable underwater operation system usage and / or maintenance scenario.

[0012] To address at least one of the aforementioned technical problems, embodiments of this application provide a method and system for identifying high-risk areas in the air and reconstructing hoisting paths.

[0013] Figure 1 The flowchart of a method for identifying high-risk areas in the air and reconstructing hoisting paths, as shown in an embodiment of this application, includes:

[0014] Step S101: Install two rotating radars on the rotatable support of the tower crane to form a radar sensing module containing dual radar modules. The radar sensing module is communicatively connected to the processing module.

[0015] In the embodiments of this application, Figure 2 Taking the high-altitude hoisting scenario shown as an example, two sets of rotatable radar sensors are installed at key positions on the tower crane. The radar angle is dynamically adjusted via a rotatable support (such as an electric pan-tilt head), constructing a dual-radar collaborative sensing hardware system with angle adjustment capabilities (such as...). Figure 3 As shown), this is the radar sensing module. The two sets of radars are defined as the first rotating radar and the second rotating radar (not fixed main and auxiliary, but dynamically switchable). Both establish bidirectional communication with the processing module (such as an industrial-grade PLC or edge computing controller, which integrates multiple types of sensors or is connected to multiple types of sensors) to realize data interaction and control command transmission.

[0016] Alternatively, in the dual-radar module, the two rotating radars can scan synchronously at a frequency of 10Hz. For example, based on the technical parameters of TF-ALS-LIDAR-01 (detection distance ≥150 meters, accuracy ≤3 cm, angular resolution ≤0.05°), a rotating bracket meeting industrial-grade protection standards (IP67) is selected. The bracket material must have corrosion resistance, and the rotating mechanism uses a high-precision harmonic reducer to ensure a turning accuracy of 0.1 degrees. The dual radars are distributed and installed on both sides of the bracket, rigidly connected to the tower crane boom through a customized structure. Alternatively, the installation angle needs to be optimized using 3D modeling based on the tower crane's operating radius to ensure overlapping coverage areas and eliminate blind spots.

[0017] For example, the communication connection adopts a dual-redundant architecture of CAN bus and Ethernet. Each radar is configured with an independent CAN channel to transmit point cloud data, while gigabit-level data interaction between the processing module and the radar is achieved through an industrial switch. The processing module configures the radar parameters (such as scanning frequency and sampling rate) through a preset protocol and establishes a data buffer queue to process the point cloud data stream. Optionally, dynamic calibration is required during installation. In a static state, a standard reflector (99% reflectivity) is used to calibrate the angle deviation. During dynamic testing, data from the tower crane tilt sensor is synchronously acquired to establish a real-time conversion model between the rotating support angle and the radar coordinate system, compensating for measurement drift caused by mechanical vibration. This ultimately forms a three-dimensional scanning system, capable of generating high-precision positioning data that meets the tower crane's requirements within a long-distance operating radius in a single scan.

[0018] Step S102: Using the radar perception module, with the first rotating radar in the dual radar module as the upward-looking radar and the second rotating radar as the downward-looking radar, based on the perception strategy of combining the upward-looking radar and the downward-looking radar, the tower arm intersection area, the overhead cable, and the swing range of the crane arm in the construction scene are dynamically scanned to obtain the real-time location information of the potential risk area.

[0019] In this embodiment, both the upward-looking radar and the downward-looking radar are rotating radars, belonging to dual radar modules, and are installed on rotatable supports at different locations on the tower crane. The upward-looking radar uses an upward view as its primary scanning direction, and based on the initially configured scanning elevation angle range, horizontal scanning range, and scanning frequency, it mainly covers the area above and to the side of the crane boom, focusing on capturing dynamic risk sources in this area. The downward-looking radar uses a downward view as its primary scanning direction, and similarly based on the initially configured scanning parameters, it mainly covers the area below the crane boom and the low-altitude operating area, focusing on monitoring low-altitude obstacles and related dynamic risks.

[0020] For example, taking the remote control scenario of lifting equipment such as port container cranes as an example, the upward-looking radar adopts a high-elevation-angle scanning optimization strategy, with its beam axis at an angle greater than 30° to the horizontal plane. This can cover the intersection area of ​​the tower crane boom and the space above the boom's rotation trajectory. It uses millimeter-wave frequency bands to penetrate the reflection interference from the tower crane's metal structure and identify potential collision risks in the intersection operation area. The downward-looking radar adopts a low-dip-angle scanning optimization strategy, with its beam axis at an angle less than 15° to the horizontal plane. It focuses on monitoring overhead cables, ground obstacles, and the area covered by the boom's rotation radius, using the Doppler effect to analyze the motion vectors of dynamic obstacles. These angles are only examples; in actual application scenarios, configurations will be tailored to specific usage requirements.

[0021] The perception strategy based on a combination of looking-up and looking-down radar relies on division of labor and priority-driven approaches. In the initial phase, initial scanning parameters are set for both radars according to the characteristics of the construction scenario, ensuring seamless coverage without blind spots, and prioritizing potential risk areas. During real-time operation, tower crane attitude data is continuously received, and combined with risk priorities, a risk-priority-driven dynamic scanning scheduling algorithm predicts the type and distance of risk areas, thereby adjusting the scanning strategies of both radars accordingly. For risk areas of different priorities, the two radars adjust their respective scanning range, frequency, and beam pattern to achieve collaborative scanning. After scanning, the processing module performs spatiotemporal verification on the location information of the same target in the target point cloud data acquired by the two radars to ensure data accuracy.

[0022] In this embodiment, the potential risk areas include at least one of the following: tower arm intersection area, overhead cables, area where the hoisting assembly is located, crane boom range of motion, excavator boom range of motion, counterweight boom range of motion, and jib range of motion. Among these, the tower arm intersection area in the construction scenario refers to the area where the booms of different tower cranes overlap in space when multiple tower cranes are operating; this area carries the highest risk due to the potential for boom collisions. Overhead cables refer to high-voltage cables, communication cables, etc., that cross the work area; they are usually suspended at high or low altitudes. If they come into contact with the boom, hook, or hoisted load, they can easily cause safety accidents, thus posing the next highest risk. The jib slewing range refers to the circular area covered by the tower crane boom as it rotates around its slewing center. This area may contain dynamic obstacles such as temporary construction equipment, personnel, or structures, and the collision risk is relatively low.

[0023] The area occupied by the hoisting assembly refers to the three-dimensional space occupied by the hoisting assembly itself, along with supporting hoisting equipment such as slings, lifting gear, and lifting rings, throughout the entire process of lifting, moving, hovering, and lowering. This includes not only the space covered by the outline of the hoisting assembly itself but also the surrounding safety buffer zone extended by the swaying of the hoisting object and the swinging of the slings. This area dynamically changes with the movement trajectory of the hoisting assembly, and its spatial extent is directly related to the weight and volume of the hoisting object, the length of the slings, and hoisting conditions such as wind speed and acceleration. The heavier the hoisting object, the larger its volume, and the longer the slings, the greater the swaying amplitude and the wider the area. The core risk lies in the possibility of the hoisting assembly falling, becoming detached, or the slings breaking, or colliding with surrounding equipment, personnel, or buildings due to swaying. If the hoisting object is flammable, explosive, or fragile, it may also cause secondary accidents. The risk level is second only to the tower arm intersection area and is on par with the overhead power line cable crossing area.

[0024] The range of motion of a crane boom refers to the complete three-dimensional spatial area covered by the boom (including the main boom and auxiliary boom) of a mobile crane such as a truck crane, crawler crane, or all-terrain crane during its combined movements of extension (length adjustment), pitch (angle adjustment), and slewing (rotation around the crane body). It is the core dynamic risk area during crane operations. Unlike tower crane booms, which primarily rely on slewing motion, crane booms offer greater freedom of movement, a more flexible and fluid coverage area, and change with the crane's position. Their spatial shape is a conical or fan-shaped three-dimensional body, with boundaries determined by the maximum boom extension, maximum pitch angle, and maximum slewing angle. The range of motion of an excavator boom refers to the maximum spatial area covered by the excavator's working devices (including the boom, stick, and bucket) during digging, slewing, and unloading operations. The range of motion of a counterweight boom specifically refers to the space reached by the counterweight boom, located at the rear of the crane to balance the lifting torque, during its rotation or swinging during operation. For example, during luffing, the counterweight boom will move accordingly to balance the torque generated by the boom and the load.

[0025] In this embodiment, the real-time location information includes multiple aspects. For example, the potential risk areas are defined by three-dimensional spatial coordinates based on the radar sensing module, clearly indicating the specific location range of each risk area. Other aspects include the dynamic attributes of the potential risk areas, such as changes in the overlapping area of ​​the tower arm intersection, the swaying amplitude of the crossing cable, and the movement direction of dynamic obstacles. Additionally, the verification attributes of the scan data include the scanning timestamps of the dual radars, the confidence level of the point cloud data, and the field-of-view overlap rate, used to verify the reliability of the data.

[0026] As an optional embodiment, in step S102, based on the initially configured scanning elevation angle range, horizontal scanning range, and scanning frequency, a looking-up radar is set to cover the dynamic risk areas above and to the side of the crane boom. Furthermore, based on the initially configured scanning elevation angle range, horizontal scanning range, and scanning frequency, a looking-down radar is set to cover the area below the crane boom and the low-altitude working area. The time synchronization deviation of the two radars is calibrated, and the priority of potential risk areas in the construction scenario is obtained, with priorities from high to low: tower boom intersection area, overhead cables, and crane boom slewing range. Tower crane attitude data is received in real time, including slewing angle, luffing length, and lifting height. Based on the tower crane attitude data and the priority of potential risk areas, combined with a risk priority-driven dynamic scanning scheduling algorithm, the type and distance of the scanning area are predicted, and based on the prediction results, the looking-up radar and the looking-down radar are controlled to execute scanning optimization strategies corresponding to the scanning area type to obtain target point cloud data. After collecting the target point cloud data, the spatiotemporal dimension verification of the same target location information in the target point cloud data is performed to obtain the real-time location information.

[0027] It's important to understand that during the initial preparation phase, the upward-looking radar is installed on a rotatable bracket in the upper part of the tower crane's jib, ensuring the scanning direction covers the area above and to the side of the jib. The downward-looking radar is installed on a rotatable bracket below the tower crane's slewing platform, ensuring the scanning direction covers the area below the jib and the low-altitude region. Both radars establish communication connections with the processing module. Simultaneously, initial scanning parameters are set for both radars, and the priority ranking of potential risk areas is recorded. During real-time operation, the processing module automatically calibrates the time synchronization deviation between the two radars upon startup, avoiding positional information discrepancies caused by time asynchrony. The processing module receives attitude data such as slewing angle, luffing length, and lifting height in real time from the tower crane's built-in sensors, updating the tower crane's real-time position. After predicting the distance to the risk area based on the attitude data, it triggers a corresponding scanning optimization strategy, controlling the two radars to adjust their scanning parameters for synchronous scanning, generating target point cloud data. After receiving the point cloud data, the processing module compares the scanning timestamps of the two radars on the same target, matches the spatial coordinates, eliminates data with significant deviations, and finally outputs the real-time position information of the potential risk areas.

[0028] By employing coordinated scanning with both upward-looking and downward-looking radars, comprehensive coverage of the area above, to the sides, below, and at low altitudes of the crane boom is achieved. This solves the problem of blind spots at high and low altitudes inherent in traditional single-radar scanning, ensuring that all types of potential risk areas can be effectively identified. A priority-driven dynamic scheduling algorithm adjusts scanning parameters for different potential risk areas and incorporates spatiotemporal verification to eliminate erroneous data, improving the accuracy of potential risk area location information and meeting the precision requirements of hoisting operations. Real-time updates of high-risk area location information provide accurate data support for subsequent path reconstruction, enabling early avoidance of various collision risks and reducing the incidence of hoisting accidents. Furthermore, this strategy can automatically update the priority and scanning parameters of potential risk areas according to changes in the construction scenario, reducing manual intervention. It is applicable to various scenarios such as multi-tower crane collaborative operations and complex building construction, demonstrating strong adaptability.

[0029] Further optionally, in the optional embodiment of step S102, in the position information verification and anomaly handling stage, the processing module performs a two-dimensional verification on the same target position information collected by the dual radars. In the spatial dimension verification, the processing module calculates the three-dimensional coordinate deviation of the target's center point collected by the upward-looking radar and the downward-looking radar. If the deviation is within a reasonable range, the average of the two coordinates is taken as the target's real-time position information. If the deviation exceeds the reasonable range, the radar module is triggered to rescan. If the deviation still exceeds the reasonable range after rescanning, a backup feature verification method is activated, which corrects the coordinates by analyzing the trend of the target's movement trajectory, thereby reducing the coordinate deviation. In the time dimension verification, the processing module checks the time difference between the position information collected by the dual radars. If the time difference exceeds a reasonable range, time compensation processing is performed on the lagging data to ensure that the two sets of position information remain synchronized in time. When radar signal obstruction occurs and lasts for more than the set duration, the processing module will issue a blind spot warning to the operator and automatically adjust the angle of the radar bracket to try to eliminate the signal obstruction problem. If the obstruction cannot be eliminated after adjusting the angle, the system will switch to single radar emergency identification mode to ensure that basic position information can be output normally until the signal obstruction problem is resolved.

[0030] The above steps, through a dual-dimensional verification mechanism of space and time, effectively improve the accuracy and synchronization of the location information acquired by the dual radars, avoiding misjudgments of risk areas due to coordinate deviations or time asynchrony. In the event of radar signal obstruction, the system employs a progressive approach of early warning, automatic adjustment of the support angle, and switching to emergency modes. This not only promptly alerts operators to anomalies but also maintains the output of basic location information under special circumstances, preventing risk identification from stalling due to signal interruption. This further ensures the stability and continuity of the potential risk area identification process, providing more reliable data support for subsequent hoisting path reconstruction and indirectly improving the safety level of hoisting operations.

[0031] Further optionally, in an optional embodiment of step S102, based on the tower crane attitude data and the priority of potential risk areas, and combined with a risk priority-driven dynamic scanning scheduling algorithm, the type and distance of the scanning area are predicted, including: spatiotemporally fusing the real-time received tower crane attitude data with the user-set initial hoisting path data to construct a positional association model between the crane boom movement trajectory and potential risk areas within the working area; synchronously collecting the dynamic characteristics of each potential risk area, wherein the dynamic characteristics of each potential risk area include the presence of dynamic obstacles and the obstacle movement rate; and defining multiple weighting dimensions based on a multi-weight dynamic scanning scheduling algorithm, including priority weight, distance weight, and dynamic weight, wherein the priority weight is highest in the tower boom intersection area, followed by the cable crossing area, and then the crane boom. The minimum boom slewing range is set, and the distance weight is set according to the closer the current position of the boom is to the potential risk area. The dynamic weight is set according to the area with dynamic obstacles being higher than the static area and the faster the dynamic obstacle moves, and preset weight coefficients are configured for each weight dimension. The fused position association model and the dynamic characteristics of each potential risk area are combined and the comprehensive weight of each potential risk area is calculated by weighted summation according to the multi-weight dynamic scanning scheduling algorithm. The priority of the areas to be scanned is determined by sorting them from high to low comprehensive weight. The dynamic geometric topology analysis algorithm is used to determine the type of the area to be scanned with the highest comprehensive weight, which is determined to be a tower boom intersection area, a cable crossing area, or the boom slewing range, and the straight-line distance from the current position of the boom to the area to be scanned is calculated.

[0032] The above steps, by fusing tower crane attitude and path data and combining them with multi-dimensional weighted analysis, enable the system to accurately pinpoint the most critical potential risk areas, ensuring that scanning resources are prioritized for the highest-risk and most urgent areas. Simultaneously, the clear determination of area type and distance provides an accurate basis for targeted adjustments to subsequent radar scanning strategies, avoiding blind scanning and resource waste. This multi-factor dynamic assessment approach improves the timeliness and accuracy of potential risk area prediction, enabling the entire sensing system to more efficiently cope with complex and ever-changing construction environments, and providing a more reliable guarantee for the safe operation of hoisting operations.

[0033] For example, when the tower crane begins operation, the processing module continuously receives real-time attitude data of the tower crane, including the current slewing angle of the boom, the extended luffing length, and the lifting height of the hook, while simultaneously retrieving the user-preset initial lifting path data. The processing module integrates these two sets of data to construct a motion trajectory model of the boom within the operating area and identifies potential risk areas near the trajectory. During this process, the system synchronously collects dynamic information about these potential risk areas, such as determining whether other tower crane booms are moving and their speed in the boom intersection area, whether overhead cables are swaying due to wind, and whether there are dynamic obstacles such as construction vehicles within the boom's slewing range and their speed.

[0034] Subsequently, the system processes the data using a multi-weighted dynamic scanning scheduling algorithm. In terms of weight settings, the priority weight of the tower arm intersection area is explicitly set higher than that of the overhead cable, which in turn is prioritized over the boom's slewing range. For distance weights, if the current position of the boom is close to a potential risk area, the distance weight for that area is increased accordingly. Regarding dynamic weights, areas with dynamic obstacles have higher weights than static areas, and the faster the obstacle moves, the higher the weight. Corresponding weight coefficients are assigned to each of these three weight dimensions.

[0035] To further explain, the multi-weighted dynamic scan scheduling algorithm uses multiple weights to comprehensively evaluate the scanning priority of potential risk areas, ensuring that radar resources are accurately deployed to the areas requiring the most attention. Priority weights are set based on the inherent risk level of different areas. Tower arm intersections, which may cause serious accidents such as tower crane collisions, are assigned the highest priority weight. Cable crossings, which could lead to electric shock or cable damage if contact occurs, have a lower risk level and therefore a lower priority weight. Dynamic obstacles within the crane boom's swing range, while posing a collision risk, have relatively minor consequences and thus receive the lowest priority weight. Distance weights reflect the urgency of the risk; the closer the crane boom's current position is to a potential risk area, the more imminent the possibility of a collision, and the higher the corresponding distance weight, ensuring priority scanning of nearby risk areas. Dynamic weights focus on the movement characteristics of obstacles within the area; areas with dynamic obstacles are more difficult to predict than static areas, and therefore have higher weights. Meanwhile, the faster the movement speed of dynamic obstacles, the more complex their trajectory changes, and the greater the uncertainty of collision risk. Therefore, the dynamic weight increases with the movement speed. To ensure that each weight dimension plays a reasonable role in the comprehensive evaluation, the algorithm configures preset weight coefficients for priority weight, distance weight, and dynamic weight. These coefficients can be adjusted according to the specific characteristics of the construction scenario. The weights of the three dimensions are integrated into a comprehensive weight through weighted calculation. Finally, the priority ranking of the areas to be scanned is determined based on the comprehensive weight, realizing the dynamic optimization scheduling of radar scanning resources.

[0036] Then, the processing module inputs the constructed location association model and the collected dynamic features into the algorithm, calculates the comprehensive weight of each potential risk area through weighted calculation, and determines the scanning order according to the comprehensive weight.

[0037] Then, the dynamic geometric topology analysis algorithm is used to identify the area to be scanned with the highest comprehensive weight, determine whether it belongs to the tower arm intersection area, the cable crossing area, or the swing range of the crane arm, and calculate the straight-line distance from the current position of the crane arm to the area.

[0038] It should be further explained that, for example, in the above dynamic geometric topology analysis algorithm, a dynamic coordinate system that can be updated in real time is first constructed with the tower crane's slewing center as the origin. This coordinate system synchronously corrects its coordinate parameters as the crane boom's posture changes, ensuring that the coordinate system always matches the real-time position of the crane boom. Next, the spatial features of the area to be scanned are transformed into the association relationship between topological nodes and edges. For example, features such as the spatial intersection of two crane booms in the tower boom intersection area, the linear suspension shape of the overhead cable, and the fan-shaped boundary formed when the crane boom rotates are all transformed into nodes and edges in the topological structure, establishing a correspondence between the area to be scanned and the topological model. Then, the area type is determined by calculating the topological matching degree. The system pre-sets corresponding topological matching factors for the features of different potential risk areas. During the calculation process, the topological structure of the area to be scanned after transformation is compared with the preset topological models of each area. Based on whether the matching degree reaches the preset matching factor for the corresponding area, it is determined whether the area to be scanned belongs to the tower boom intersection area, the overhead cable area, or the crane boom's slewing range. Meanwhile, by combining the real-time motion vector of the crane boom derived from the tower crane's attitude data, including its speed and direction, the topological distance iteration calculation method is adopted. In the process of calculating the straight-line distance from the current position of the crane boom to the area to be scanned, the relative motion trend of the two is incorporated, thereby avoiding the lag that may occur if only static distance calculation is used, and realizing dynamic correction of the distance.

[0039] The aforementioned algorithm ensures the accuracy of spatial location description by constructing a dynamic coordinate system that updates in real time with the crane boom's posture, providing a reliable coordinate foundation for subsequent analysis. Transforming the spatial features of the area to be scanned into topological relationships and combining them with preset matching factors improves the accuracy of area type identification and reduces misjudgments caused by environmental interference. Furthermore, incorporating the real-time motion vector of the crane boom into distance calculation enables dynamic distance correction, effectively avoiding the lag problem of static distance calculation. This allows the obtained distance information to better reflect actual operational conditions, providing more accurate and timely data support for subsequent radar scanning strategy adjustments and hoisting path planning, further ensuring the safety and efficiency of hoisting operations.

[0040] As an optional embodiment, in step S102, based on the prediction result, the upward-looking radar and the downward-looking radar are controlled to execute scanning optimization strategies corresponding to the scanning area type to obtain target point cloud data, including:

[0041] When a tower arm intersection is identified and the straight-line distance is less than a preset safety distance, the first scanning optimization strategy is triggered. This involves controlling the upward-looking radar to reduce its horizontal scanning range and increase its scanning frequency to focus on the dynamic contour of the intersection; and controlling the downward-looking radar to expand its horizontal scanning range and increase its scanning frequency to simultaneously capture the associated area below the intersection, ensuring that the overlap coverage of the dual radar fields of view reaches a set ratio and periodically updating the original point cloud data corresponding to the intersection. When a cable is identified and the straight-line distance is less than a preset safety distance, the second scanning optimization strategy is triggered. This involves controlling the upward-looking radar to adjust its scanning elevation angle range and enable narrow-beam scanning to enhance the identification of cable reflection signals; and controlling the downward-looking radar to adjust its scanning elevation angle range and enable wide-beam scanning to capture the cable's sag trajectory. Through beam coordination scheduling, the dual radars synchronously acquire the upper and lower cable trajectories to obtain the corresponding original point cloud data. When the crane boom's slewing range is determined and a dynamic obstacle is detected, a third scanning optimization strategy is triggered. This strategy controls both the upward-looking and downward-looking radars to maintain their initial horizontal scanning range and simultaneously increases their scanning frequency. Regional segmentation scanning covers the entire slewing range of the crane boom to avoid missing obstacles, collecting raw point cloud data corresponding to the slewing range and dynamic obstacles. Finally, the raw point cloud data collected under each scanning optimization strategy is timestamped to ensure time synchronization of the dual radar data. Noise interference points are filtered out using a filtering algorithm, forming target point cloud data that can be used for subsequent spatiotemporal dimension verification.

[0042] The above steps dynamically switch scanning optimization strategies and adjust the scanning parameters of the dual radars based on different types of potential risk areas, actual distances, and obstacle conditions. This achieves targeted scanning of various risk areas, ensuring the accuracy of potential risk area scanning while avoiding waste of radar resources. Timestamp calibration and noise filtering of the raw point cloud data effectively improve the accuracy and reliability of the data, providing a high-quality data foundation for subsequent acquisition of real-time location information of potential risk areas, further ensuring the accuracy of potential risk area identification.

[0043] In the example tower crane operation scenario, when the system determines that the area to be scanned is the tower arm intersection area, and calculates that the straight-line distance from the current position of the crane arm to this area is less than the preset safety distance, the first scanning optimization strategy will be triggered immediately. At this time, the processing module will control the upward-looking radar to reduce its original horizontal scanning range and increase the scanning frequency, allowing the radar to focus more accurately on the tower arm intersection area and capture the dynamic outline of the crane arm in this area in real time. At the same time, the downward-looking radar will expand its horizontal scanning range and similarly increase the scanning frequency to synchronously cover and capture the related areas below the tower arm intersection area, ensuring that the overlapping part of the fields of view of the upward-looking and downward-looking radars reaches the set coverage ratio, and regularly updating the original point cloud data related to the tower arm intersection area.

[0044] If the system determines that the area to be scanned is a passing cable and the straight-line distance is less than the preset safety distance, it triggers the second scanning optimization strategy. The processing module instructs the looking-down radar to adjust its scanning elevation angle range and enable a narrow beam scanning mode to enhance the identification capability of the reflected signals from the passing cable and more clearly capture the cable's position information. At the same time, it controls the looking-down radar to also adjust its scanning elevation angle range and enable a wide beam scanning mode to completely capture the suspended trajectory of the passing cable. Through the coordinated scheduling of the dual radar beams, the synchronous acquisition of the upper and lower end trajectories of the cable is achieved, thereby obtaining the original point cloud data corresponding to the passing cable.

[0045] When the system determines that the area to be scanned is the swing range of the crane boom and detects dynamic obstacles such as construction vehicles and temporary equipment within that area, it triggers the third scan optimization strategy. At this time, both the upward-looking and downward-looking radars maintain their initially set horizontal scanning ranges, and the processing module simultaneously increases the scanning frequency of both. By using a region-segmented scanning method, it ensures coverage of the entire swing range of the crane boom, avoiding missed detection of dynamic obstacles due to incomplete scanning, and thus collecting the original point cloud data corresponding to the swing range of the crane boom and the dynamic obstacles within that area.

[0046] After completing the acquisition of raw point cloud data under each scanning optimization strategy, the system will perform timestamp calibration on these data to ensure that the data acquired by the upward-looking radar and the downward-looking radar are synchronized in time. Then, the system will filter out noise interference points in the data through a filtering algorithm, and finally form target point cloud data that can be used for subsequent spatiotemporal dimension verification.

[0047] As an optional embodiment, in step S102, based on the tower crane attitude data and the priority of potential risk areas, combined with a risk priority-driven dynamic scanning scheduling algorithm, the type and distance of the scanning area are predicted. Based on the prediction result, the upward-looking radar and the downward-looking radar are controlled to execute scanning optimization strategies corresponding to the scanning area type to obtain target point cloud data. The process further includes: performing multimodal feature association processing on the target point cloud data, extracting and associating feature parameters acquired by the dual radars for different types of potential risk areas. Specifically, for tower arm intersection areas, the continuity of the tower arm metal reflectivity acquired by the upward-looking radar and the projection contour of the intersection area acquired by the downward-looking radar are correlated, and the dynamic motion vector between the tower arm metal reflectivity and the projection contour continuity of the intersection area is combined to determine the intersection area boundary. For overhead cables, the slender linear characteristics of the cable acquired by the upward-looking radar and the sag curvature of the cable acquired by the downward-looking radar are correlated, and the dielectric constant feedback of both is combined to eliminate interference from metal pipelines and determine the three-dimensional trajectory of the cable. For the boom's slewing range, the boom's structural outline acquired by the looking-down radar and the distance to obstacles in the slewing area acquired by the looking-up radar are correlated, and the real-time slewing speed of the boom is combined to define the slewing safety boundary. Furthermore, after multimodal feature association processing, optimized location information for each potential risk area, verified by features, is obtained and used as the real-time location information for subsequent three-dimensional coordinate transformation.

[0048] The above steps, by associating different modal features acquired by dual radars, fully leverage the advantages of both looking-up and looking-down radars, overcoming the limitations of single-radar feature recognition, and significantly improving the accuracy of determining the location information of different types of potential risk areas. Employing differentiated feature association strategies for different potential risk areas can accurately address the identification difficulties of various areas, such as eliminating interference from metal pipelines on cable identification and accurately determining the boundaries of intersection areas. The resulting optimized location information provides a high-quality data foundation for subsequent 3D coordinate transformation, indirectly improving the reliability of subsequent hoisting path reconstruction and further ensuring the safety of tower crane hoisting operations.

[0049] For example, after the dual radars complete scanning and acquire target point cloud data, the processing module initiates a multimodal feature association processing flow. If the potential risk area to be processed is a tower arm intersection area, the metal reflectivity features of the tower arms are extracted from the point cloud data of the upward-looking radar, and the projection contour continuity features of the intersection area are extracted from the point cloud data of the downward-looking radar. These two types of features are then associated, combined with the dynamic motion vector between them, such as the vector formed by the relative movement direction and speed of the two tower arms, to accurately determine the actual boundary of the tower arm intersection area, avoiding boundary ambiguity caused by single feature judgment. If the potential risk area is a cable crossing, the processing module first extracts the cable's unique thin linear features from the upward-looking radar data, and extracts the cable's sag curvature features from the downward-looking radar data. These two features are then associated, and combined with the dielectric constant feedback information collected by the dual radars, the difference in dielectric constant is used to distinguish the cable crossing from surrounding metal pipelines, eliminating the interference of metal pipelines on cable identification, and thus accurately determining the three-dimensional trajectory of the cable crossing, ensuring that the trajectory information is not confused with other objects. When the potential risk area falls within the crane boom's slewing range, the system extracts the crane boom's structural contour features from the upward-looking radar data and the distance features between obstacles and the tower crane within the slewing area from the downward-looking radar data. After associating these two types of features and combining them with the crane boom's real-time slewing speed, the system comprehensively calculates and delineates the crane boom's slewing safety boundary, clearly defining the range within which the crane boom will not collide with obstacles during slewing. After completing the multimodal feature association processing for different types of potential risk areas, the processing module generates optimized location information for each potential risk area after feature verification. This optimized location information is then used as the real-time location information required for subsequent 3D coordinate transformation.

[0050] Step S103: Receive the real-time position information and use a three-dimensional coordinate transformation algorithm to convert the real-time position information into real-time three-dimensional coordinate data in the motion coordinate system of the hoisting equipment.

[0051] For example, firstly, a motion coordinate system for the hoisting equipment is established based on the tower crane's slewing center. The orientation of each axis in the coordinate system is clearly defined: the X-axis corresponds to the initial direction of the horizontal projection of the jib, the Y-axis is the direction perpendicular to the X-axis on the horizontal plane, and the Z-axis is the vertically upward direction, providing a unified spatial reference for subsequent coordinate transformation. Next, the processing module calls pre-stored dual-radar module installation parameters. These parameters include the installation coordinates of the radar relative to the slewing center and the possible angular deviation values ​​during installation. Based on these parameters, a spatial mapping relationship is established between the radar sensing coordinate system and the hoisting equipment motion coordinate system, and an initial matrix for coordinate transformation is generated. Then, the real-time location information of the potential risk area, obtained after multimodal feature association processing, is input into the 3D coordinate transformation algorithm. This information is initially presented in the form of coordinates in the radar sensing coordinate system. Rotation and translation operations are performed using the initial coordinate transformation matrix to initially convert it into 3D coordinates in the hoisting equipment motion coordinate system. Based on this, combined with real-time received tower crane attitude data, including the slewing angle, luffing length, and lifting height of the jib, the coordinate results obtained from the initial transformation are dynamically calibrated. The real-time pointing of the X-axis is adjusted by using the slewing angle to correct directional deviations caused by the boom rotation. The horizontal distance parameter is corrected by adjusting the luffing length to ensure horizontal accuracy. The Z-axis coordinate is calibrated using the lifting height to match the actual vertical height of the hook and the load. Finally, a Kalman filter algorithm is used to perform time-series optimization on the calibrated 3D coordinate data, filtering out noise generated during radar measurement and coordinate fluctuations caused by equipment vibration. The final output is smooth 3D coordinate data of the potential risk area that meets preset real-time requirements. This data will serve as the foundation for subsequently constructing a 3D model of the potential risk area.

[0052] The above steps establish a dedicated motion coordinate system with the tower crane's slewing center as the origin, ensuring a high degree of consistency between the spatial reference for coordinate transformation and the actual motion state of the lifting equipment. Utilizing dual radar installation parameters to construct a spatial mapping relationship and generate a transformation matrix provides crucial support for high-precision transformation between different coordinate systems. Dynamic calibration using real-time tower crane attitude data effectively offsets the impact of boom attitude changes on coordinate accuracy, while the application of the Kalman filter algorithm further enhances the stability and reliability of the coordinate data. The overall process achieves precise transformation of potential risk area location information from the radar perception dimension to the lifting equipment's motion dimension, providing high-quality, highly adaptable data support for subsequent 3D model construction and path obstacle avoidance algorithm applications, ensuring the accuracy and safety of lifting path planning.

[0053] Step S104: Construct a three-dimensional model of the potential risk area based on the real-time three-dimensional coordinate data.

[0054] As an optional embodiment, in step S104, the real-time three-dimensional coordinate data is preprocessed and completed using an interpolation algorithm to form a potential risk area coordinate dataset; a differentiated modeling strategy is adopted for different types of potential risk areas, and a sub-model of each potential risk area is constructed by combining the potential risk area coordinate dataset; the sub-models of each potential risk area are integrated according to the spatial position of the hoisting equipment motion coordinate system, and the consistency of the spatial position of each sub-model is ensured by a coordinate alignment algorithm to form a three-dimensional model containing all potential risk areas.

[0055] Further optionally, in step S104, preprocessing the real-time three-dimensional coordinate data and using an interpolation algorithm to complete it to form a potential risk area coordinate dataset can be implemented as follows: preprocessing the real-time three-dimensional coordinate data, removing outliers caused by equipment noise or environmental interference, removing data points whose deviation from the mean of the area coordinates exceeds a preset deviation threshold, and using an interpolation algorithm to complete the missing local coordinate data to form a complete potential risk area coordinate dataset.

[0056] Furthermore, in step S104, a differentiated modeling strategy is adopted for different types of potential risk areas. Sub-models for each potential risk area are constructed by combining the potential risk area coordinate dataset. This can be achieved as follows: For tower arm intersection areas, the spatial intersection range of the two tower arms is calculated based on coordinate data. A closed surface model of the intersection area is generated through Boolean operations on spatial polygons. Simultaneously, the material properties corresponding to the metal reflectivity of the tower arms are associated with the model surface of the sub-model, and the collision risk level of the intersection area is labeled. For overhead cables, a tubular three-dimensional model is constructed with the cable's three-dimensional trajectory coordinates as the central axis and the actual diameter parameters of the cable. The dielectric constant feedback result assigns a non-metallic material identifier to the model, distinguishing it from metal pipelines. For the crane boom slewing range, a fan-shaped three-dimensional region model is constructed with the tower crane's slewing center as the vertex, the crane boom length as the radius, and the slewing angle range as the fan angle. The obstacle distance data in the slewing region is mapped to the obstacle sub-model within the model, and the obstacle type and dynamic motion parameters are labeled.

[0057] Finally, in step S104, the sub-models of each potential risk area are integrated according to their spatial positions in the coordinate system of the hoisting equipment. A coordinate alignment algorithm is used to ensure the consistency of the spatial positions of each sub-model, forming a 3D model that includes all potential risk areas. Optionally, the 3D model can be dynamically iteratively updated using real-time scanning information of the hoisting area dynamically updated by dual radar modules. When the change in coordinate data of a potential risk area exceeds a preset update threshold, a local model reconstruction is automatically triggered to correct the coordinate parameters and model shape of the corresponding area.

[0058] Optionally, a model accuracy verification algorithm is employed to compare the spatial deviation between the reconstructed 3D model and historical measured data. This ensures that the overall coordinate parameter deviation in the reconstructed 3D model is less than a preset accuracy threshold, resulting in the final output 3D model of the potential risk area. The model accuracy verification algorithm further ensures the accuracy of the model, guaranteeing that the constructed 3D model meets the preset accuracy requirements in coordinate parameters. This provides a high-quality spatial reference for subsequent operations such as hoisting path adjustment and risk avoidance, significantly reducing safety hazards caused by model errors and improving the efficiency and reliability of the entire hoisting operation planning.

[0059] Combining the preceding steps, the 3D model of the potential risk area is a digital model constructed based on real-time 3D coordinate data, accurately reflecting the spatial morphology and characteristics of various potential risk areas in the construction scenario. The 3D model covers different types of potential risk areas, such as tower arm intersections, overhead cables, and the slewing range of the crane arm. It not only presents the spatial location and size of each area but also associates corresponding attribute information, such as risk level, material characteristics, and obstacle parameters. Furthermore, it possesses dynamic update capabilities, adjusting the model's shape and parameters based on real-time scanning data, providing an intuitive and accurate spatial reference for subsequent hoisting path adjustments. Based on this, step S104, by constructing the 3D model of the potential risk area, transforms abstract real-time 3D coordinate data into a concrete spatial model, allowing operators to more clearly grasp the distribution and characteristics of potential risk areas and avoid cognitive biases caused by data abstraction. The dynamic update characteristic of the 3D model ensures that the model always remains consistent with the state of potential risk areas in the actual construction scenario, providing a reliable basis for the accurate operation of subsequent path obstacle avoidance algorithms, improving the accuracy and efficiency of risk assessment before hoisting operations, and laying the foundation for safe hoisting operations.

[0060] Step S105: Identify the high-risk area by the three-dimensional model, and use an improved fast exploration random tree algorithm as the path avoidance algorithm. Combine the three-dimensional model to dynamically reconstruct the initial hoisting path set by the user, and generate the optimal hoisting trajectory that avoids the high-risk area.

[0061] The optimal obstacle avoidance hoisting trajectory is a hoisting path scheme that avoids all high-risk areas. It is formed by dynamically adjusting the initial hoisting path set by the user, combined with a 3D model of potential risk areas, and using a path obstacle avoidance algorithm. This trajectory prioritizes the safety of hoisting operations while also considering operational efficiency and equipment stability. During the planning process, the spatial location, size, and dynamic characteristics of high-risk areas are fully considered, such as the real-time overlap range of tower arm intersections, the sag trajectory of overhead cables, and the movement trend of obstacles within the crane boom's swing range. This ensures that the trajectory maintains a safe distance from all high-risk areas throughout the entire process, avoiding the risk of collision.

[0062] Furthermore, the optimal hazard-avoidance hoisting trajectory also incorporates the actual motion characteristics of the hoisting equipment, such as the slewing speed limit of the tower crane boom, the luffing length adjustment range, and the lifting height constraints, to avoid operational difficulties or equipment damage caused by the trajectory design exceeding the equipment's performance limits. During the path adjustment process, the algorithm evaluates multiple potential hazard-avoidance paths, comprehensively weighing factors such as path length, number of turns, and operation time, to select a trajectory scheme that meets safety requirements while minimizing operational delays and reducing operational complexity. The resulting optimal hazard-avoidance hoisting trajectory not only has a clear spatial coordinate sequence but also includes motion parameter suggestions for the hoisting equipment in different trajectory segments, providing a clear and feasible execution basis for subsequent hoisting operations.

[0063] As an optional embodiment, in step S105, the initial hoisting path set by the user is parsed into a set of discrete path points in the motion coordinate system of the hoisting equipment, and the motion vectors of adjacent path points are calculated to form initial path points; the motion vectors of adjacent path points include direction and velocity parameters; based on the three-dimensional model, a set of spatial obstacles is constructed, and the three-dimensional models of tower arm intersection areas, overhead cables, crane arm slewing range, and dynamic obstacles are transformed into spatial geometric constraints prohibiting passage, and multi-level safety distance thresholds are marked for different types of potential risk areas; a dangerous area semantic segmentation model constructed using U-Net is used, combined with the spatial geometric constraints prohibiting passage, to classify the risk of different types of potential risk areas, and the classification results are encoded into different levels of risk areas; an improved A fast random tree algorithm is explored as a path avoidance algorithm. Using an initial set of path points as an index, it searches for paths within a 3D model constraint space. The 3D model is then used as the obstacle constraint input algorithm. A path cost function is set, where the weighting factors include path length, turning angle, and distance margin in high-risk areas. Multiple candidate avoidance paths are generated through iterative optimization. These candidate paths undergo multi-objective evaluation, and a weighted score is calculated based on hoisting operation efficiency parameters. The path with the highest comprehensive score is selected as the optimal candidate trajectory. A B-spline curve fitting algorithm is used to smooth the optimal candidate trajectory, eliminating acute inflection points and ensuring the trajectory meets the kinematic constraints of the hoisting equipment, thus generating a continuously executable optimal avoidance hoisting trajectory. Alternatively, after generating a continuously executable optimal hazard avoidance hoisting trajectory, parametric Bezier curve fitting is used. Using the key control points of the trajectory in the optimal hazard avoidance hoisting trajectory as anchor points, the curve parameters in the optimal hazard avoidance hoisting trajectory are adjusted to eliminate kinematic abrupt change points, so that the angular velocity and acceleration change rate of the trajectory are within the preset range allowed by the hoisting equipment, and the optimal hazard avoidance hoisting trajectory is obtained again through re-optimization.

[0064] The above steps, by transforming the initial path into a mathematical model and combining it with a 3D model to construct precise constraints, ensure the targeted nature and safety of obstacle avoidance, effectively avoiding various high-risk areas. The improved algorithm and multi-objective evaluation mechanism ensure both safety and lifting operation efficiency, preventing delays caused by excessive risk avoidance. Curve fitting optimization ensures the trajectory conforms to the equipment's kinematic constraints, eliminating the risk of sudden motion changes. This protects the lifting equipment from damage and ensures the stability of the lifting process, providing strong support for the safety, efficiency, and stability of lifting operations.

[0065] Specifically, in the above steps, the processing module first receives the initial hoisting path set by the user through the operating terminal. This path typically exists in a visual form with a start point, intermediate points, and an end point. The system discretizes the initial path into a series of uniformly distributed three-dimensional path points based on the parameters of the hoisting equipment's motion coordinate system (such as the origin of the rotation center and the direction of each axis). Each path point contains X, Y, and Z coordinate values ​​in the motion coordinate system. Subsequently, the system calculates the motion vector between two adjacent path points: the direction parameter of the vector is calculated using the coordinate difference to determine the direction of travel from the current point to the next point; the speed parameter of the vector is determined by combining the user-preset hoisting speed requirements and the equipment's rated speed range to ensure that the speed meets the equipment's safe operating standards. Finally, the discrete path point set is integrated with the motion vectors of adjacent points to form a mathematical expression model of the initial path with a point-vector-point structure, providing a standardized data format for subsequent algorithm calculations.

[0066] Based on the constructed 3D model, the processing module performs geometric constraint transformations on different types of potential risk areas. For the closed surface model of the tower arm intersection area, its surface and internal space are defined as an absolutely prohibited area, and no path point is allowed to enter this space. For the tubular model of the cable crossing, a safety distance of the corresponding level is extended outward from the outer wall of the tube to form a safety buffer area, and paths must maintain a distance from this area. For the fan-shaped solid model of the crane boom's swing range and the sub-model of internal obstacles, both the fan-shaped boundary and the obstacle surface are set as constraint boundaries, while a dynamic safety margin is reserved according to the obstacle's dynamic motion parameters (such as direction of movement and speed). In addition, multiple levels of safety distance thresholds are configured according to the priority of potential risk areas. The safety distance threshold is the highest for the tower arm intersection area, followed by the cable crossing, and the lowest for obstacles within the crane boom's swing range, ensuring that high-risk areas receive stricter spatial protection.

[0067] Furthermore, a semantic segmentation model for hazardous areas constructed using U-Net is employed. This model, combined with spatial geometric constraints prohibiting passage, classifies different types of potential risk areas and encodes the classification results into risk areas of varying levels. Specifically, in high-risk area identification based on 3D models, the semantic segmentation model for hazardous areas constructed using the U-Net architecture, through an encoder-decoder structure combined with skip connections, can efficiently achieve pixel-level risk area classification. The semantic segmentation model first converts 3D point cloud or mesh data into voxelized 3D tensors as input. The encoder path progressively extracts multi-level features through 3D convolutional and pooling layers, thereby capturing abstract representations from local geometric features (such as the fine structure of cables) to global spatial context (such as tower arm layout). The decoder path then progressively recovers spatial details through upsampling operations such as deconvolution, and fuses feature maps of the same resolution from the encoder path using skip connections. This approach balances the richness of semantic information with the accuracy of spatial localization, which is crucial for accurately delineating risk boundaries.

[0068] During risk classification, the hazardous area semantic segmentation model incorporates predefined geometric constraints of prohibited spaces (such as tower arm intersections and high-voltage cable buffer zones). These constraints are typically transformed into distance fields or spatial relationship features, which are input into the classification head along with deep learning features. By introducing attention mechanisms (such as spatial attention modules), the hazardous area semantic segmentation model can dynamically adjust the attention weights for different areas, thereby improving the segmentation accuracy of key risk areas (such as the edge of the crane boom's slewing range). Furthermore, the hazardous area semantic segmentation model can output a risk level label for each voxel; for example, marking areas of direct conflict as "red" prohibited zones, areas requiring high vigilance as "orange" high-risk zones, and areas requiring careful operation as "yellow" medium-risk zones. This classification result is then encoded into different levels of risk areas in a 3D mesh map, providing structured spatial risk distribution information for subsequent path avoidance algorithms.

[0069] In the path search phase, an improved fast exploration random tree algorithm is combined with the initial path mathematical model and spatial obstacle constraints. Using the discrete point set of the initial path as guiding nodes, random expansion branches are generated first, prioritizing directions towards adjacent initial path points, avoiding aimless blind searches and improving search efficiency. The geometric constraints corresponding to the 3D model are used as collision detection conditions. Each newly generated path node is immediately checked for spatial overlap with the constraint region; if overlap exists, the node is discarded and a new branch is generated. Simultaneously, a path cost function is set, including path length, turning angle, and distance margin. Shorter path lengths, smaller turning angles, and larger distance margins to high-risk areas result in lower cost function values. Through iterative optimization, low-cost path branches are continuously filtered, ultimately generating multiple candidate avoidance paths that meet the constraints. Each path is stored as a discrete point set.

[0070] In the multi-objective evaluation phase, the efficiency dimension calculates the total length of each candidate path and, combined with the rated speed of the hoisting equipment, estimates the operation time for each path; shorter operation time results in a higher score. The smoothness dimension calculates the rate of curvature change at each node of the path, statistically analyzing the frequency and amplitude of curvature fluctuations; paths with smaller curvature fluctuations and no significant abrupt changes receive higher scores. The safety dimension measures the minimum distance between each path and each high-risk area. If the minimum distance meets the corresponding area's safety distance threshold, points are added based on the magnitude of the excess; otherwise, the path is eliminated. Subsequently, weights are assigned to the three dimensions based on hoisting operation efficiency parameters (such as schedule requirements and equipment energy consumption indicators). Efficiency weights increase when the schedule is tight, smoothness weights increase when equipment is sensitive, and safety weights are highest in high-risk scenarios. A weighted summation is used to calculate the comprehensive score for each candidate path, and the path with the highest score is selected as the optimal candidate trajectory.

[0071] In B-spline curve fitting and smoothing, although the optimal candidate trajectory meets the requirements of safety and efficiency, the connection of discrete points may contain acute inflection points, leading to unstable equipment movement. The system employs a B-spline curve fitting algorithm, using the discrete points of the candidate trajectory as control vertices to generate a continuous, smooth curve. By adjusting the curve interpolation parameters, acute inflection points are eliminated, ensuring continuous curvature changes in the trajectory while maintaining a safe distance from high-risk areas and avoiding breaches of constraint boundaries. The fitted trajectory initially satisfies the kinematic constraints of the hoisting equipment, enabling continuous operation.

[0072] To further eliminate kinematic abrupt changes, the system constructs a parametric Bézier curve model using key control points of the B-spline curve (such as turning nodes and velocity change nodes) as anchor points. By adjusting the coordinates and weight parameters of the control points on the curve, the system optimizes the angular velocity and acceleration rate of change of the trajectory. For abrupt changes in angular velocity, intermediate control points are added to mitigate the change in turning velocity. For abrupt changes in acceleration, the curve curvature gradient is adjusted to ensure a smooth acceleration transition. During the optimization process, the system compares the allowed upper limits of angular velocity and acceleration with the equipment in real time to ensure that the kinematic parameters of the final trajectory are within the preset range, forming an optimal hoisting trajectory that combines safety, efficiency, and stability.

[0073] Optionally, in step S105, when conducting a multi-objective evaluation of multiple candidate avoidance paths, the total length of each candidate avoidance path is calculated, and the shorter path is prioritized. For example, a candidate path from the raw material storage area to the floor work surface, if its total length is shorter than another, will be prioritized if other evaluation dimensions are met, because a shorter path can reduce the slewing and luffing stroke of the tower crane boom, reduce equipment energy consumption, and shorten the single operation time of rebar hoisting. On the other hand, the path smoothness of each candidate avoidance path is calculated, and the path with smaller curvature fluctuations is prioritized through the curvature change rate evaluation. Taking a section of the hoisting path that needs to bypass the overhead cable as an example, the curvature values ​​at each discrete point of the section are calculated, and the amplitude and frequency of curvature changes are statistically analyzed. If a candidate path, when bypassing the cable, has a curvature that slowly transitions from the initial value to the peak value and then smoothly falls back, with a small fluctuation range and no sudden jumps, it indicates that its smoothness is higher. On the other hand, if another path has a sudden increase or decrease in curvature at the same location, its smoothness is poor and it will be prioritized for elimination. This evaluation method avoids the swaying of hoisted loads due to path bumps, reducing safety hazards. Furthermore, it calculates the safety redundancy of each candidate avoidance path and the minimum distance to high-risk areas, ensuring that the minimum distance meets the corresponding area's safety distance threshold. For example, for tower arm intersections, the system locates the point on the path closest to the intersection, calculates the distance between that point and the surface of the closed surface model of the intersection, and compares it with the preset safety distance threshold for tower arm intersections. For cables, it calculates the minimum distance between the path and the safety buffer zone of the cable tubular model to determine if it meets the safety threshold requirements for cables. If the minimum distance between a candidate path and the tower arm intersection just meets the threshold, while another path's minimum distance significantly exceeds the threshold, the latter has higher safety redundancy and is more advantageous in the evaluation. If the minimum distance between a path and any high-risk area does not meet the threshold, it is directly eliminated, ensuring the absolute safety of the final selected path.

[0074] The above steps, through quantitative evaluation of path length, smoothness, and safety redundancy, achieve a comprehensive screening of candidate avoidance paths. Prioritizing shorter paths improves lifting operation efficiency and reduces equipment operating costs and operation time; prioritizing paths with minimal curvature fluctuations reduces impact and vibration during tower crane operation, protecting mechanical components and preventing the hoisted load from swaying or falling due to path bumps; strictly controlling safety redundancy ensures a sufficient safe distance between the path and high-risk areas, fundamentally preventing collision accidents. The multi-objective evaluation system formed by these three factors allows the final selected optimal candidate trajectory to achieve the best balance between safety, efficiency, and equipment adaptability.

[0075] Following step S105, optionally, by combining the real-time operation status obtained from the real-time acquisition of the hoisting operation with the real-time updated real-time scanning information of the hoisting area, the local path where the hoisting equipment is currently located in the optimal avoidance hoisting trajectory is dynamically reconstructed. The optimal avoidance hoisting trajectory is corrected according to the dynamically reconstructed path, and corresponding operation correction information is prompted to ensure the stability and safety of the hoisting operation.

[0076] As an optional embodiment, in the above steps, real-time operation status data of the hoisting operation is collected; the real-time scanning information of the hoisting area is dynamically updated by dual radar modules according to a preset high-frequency scanning cycle, wherein the real-time scanning information includes the location of newly appearing dynamic obstacles, changes in the boundary of high-risk areas, and environmental interference parameters; combining the real-time operation status data and the real-time scanning information of the hoisting area, the local path of the hoisting equipment in the optimal avoidance hoisting trajectory is dynamically reconstructed, and the optimal avoidance hoisting trajectory is corrected according to the dynamically reconstructed path, and corresponding operation correction information is displayed. Optionally, the real-time operation status data includes the current position coordinates of the hoisting equipment, the real-time slewing angle of the boom, the luffing length, the lifting height, the swing angle of the load, and the equipment operating speed.

[0077] Specifically, in the above steps, real-time operation status data is spatiotemporally fused with updated real-time scanning information of the hoisting area. A Kalman filter algorithm is used to filter noise from the fused data. Based on the current position of the hoisting equipment, a path segment starting from the current position and ending at a preset future travel time is taken as the local path, and environmental change features of the local path are extracted. Here, the Kalman filter algorithm is mainly used for the fusion processing of real-time operation status data and real-time scanning information of the hoisting area. The processing module first collects operation status data including the current position coordinates of the hoisting equipment and the real-time rotation angle of the boom. Simultaneously, it acquires scanning information including the positions of new dynamic obstacles and changes in the boundaries of high-risk areas through dual radar modules at a preset high-frequency cycle. Then, both types of data are input into the Kalman filter algorithm. The Kalman filter algorithm combines the statistical characteristics of the data to filter out noise caused by equipment vibration and environmental interference, such as eliminating small fluctuations in the measurement of the hoisting swing angle and false interference points in the radar scanning information, resulting in more accurate fused data and providing a reliable basis for subsequent local path analysis.

[0078] When the system detects that the positional change of a high-risk area within a local path exceeds a preset dynamic threshold or that a new dynamic obstacle has appeared, it replans the local path using a rolling temporal optimization algorithm based on the local sub-model of the updated 3D model of the high-risk area. During the planning process, the smoothness of the connection between the new path and the original optimal avoidance hoisting trajectory is constrained. For example, in actual hoisting operations, when the system detects a change in the position of a high-risk area within the local path exceeding a preset dynamic threshold through real-time scanning by dual radars and monitoring of the operational status—such as a previously stable overhead cable shifting due to wind, or the sudden appearance of new dynamic obstacles like construction workers or temporary equipment in the work area—it immediately calls up the updated local sub-model of the 3D model of the high-risk area. This sub-model focuses only on the area covered by the current local path and includes key information such as the changed spatial morphology of the high-risk area and the position and movement trend of the new obstacle, avoiding efficiency losses caused by calling the entire model. Then, a rolling time-domain optimization algorithm is used to replan the local path. This algorithm takes the current position of the hoisting equipment as the starting point and defines the operation journey within a preset time period as the optimization time-domain window. Path reconstruction is only performed within this window, rather than a comprehensive adjustment of the entire trajectory, ensuring that the planning process is fast and efficient and adapts to the timeliness requirements of dynamic changes on site. During the planning process, the smoothness of the connection between the new path and the original optimal avoidance hoisting trajectory is used as a core constraint. By analyzing the motion parameters of the original trajectory at the connection point, such as the slewing angular velocity of the crane boom, the luffing speed, and the lifting acceleration of the hook, the motion parameters of the newly planned local path at the connection point are made continuous with the original trajectory, avoiding sudden changes in angular velocity or abrupt increases or decreases in speed. For example, if the slewing velocity of the original trajectory at the connection point is a certain stable value, the new path will gradually adjust its speed as it approaches the connection point to ensure a seamless connection with the original speed, preventing crane boom swaying or load swinging due to abrupt path transitions, and ensuring the stability of the hoisting process. At the same time, the updated local sub-model will be used to treat high-risk areas and newly added obstacles as spatial constraints to ensure that the newly planned local paths maintain a safe distance from these risk sources. This not only meets the obstacle avoidance requirements, but also maintains the continuity of the overall trajectory through smooth connection constraints, so that the equipment operation does not need to be significantly adjusted, reducing the difficulty of operation for operators and achieving the dual goals of dynamic obstacle avoidance and stable operation.

[0079] Finally, a multi-objective decision-making model is used to evaluate the reconstructed local path. Evaluation indicators include the safe distance from the updated high-risk area, the magnitude of path adjustment, and equipment feasibility. The optimal local reconstructed path is selected based on these indicators. For example, the reconstructed path is substituted into the model, and evaluation is conducted based on three core indicators: first, the safe distance from the updated high-risk area to determine if it meets the corresponding area's safety threshold; second, the magnitude of path adjustment to analyze the deviation between the reconstructed path and the original trajectory, avoiding excessive adjustments that increase operational difficulty; and third, equipment feasibility to verify whether the required parameters such as the slewing angle and luffing speed are within the equipment's rated capacity. The model assigns corresponding weights to the three indicators based on the priority of the current hoisting operation scenario. For example, the safety distance has a higher weight in high-risk scenarios, and the feasibility of execution has a higher weight when the equipment load is heavy. By comprehensively calculating the scores of each reconstructed path, the path with the highest score is selected as the final local reconstructed path.

[0080] The above steps improve data accuracy through the Kalman filter algorithm, laying a reliable data foundation for path reconstruction. The rolling time-domain optimization algorithm enables dynamic adjustment of local paths, ensuring timely response to environmental changes while guaranteeing smooth trajectory transitions. The multi-objective decision model selects the optimal path from multiple dimensions, including safety, operation, and equipment compatibility, avoiding the limitations of single-index evaluation. The overall process can respond to dynamic changes in the hoisting scenario in real time, quickly correct local paths, and ensure the safety and feasibility of the corrected trajectory. This effectively avoids collision risks caused by sudden environmental changes, while minimizing interference with the overall hoisting operation process, ensuring the stability and safety of the entire hoisting operation.

[0081] As an optional embodiment, the above steps, including correcting the optimal risk-avoidance hoisting trajectory based on the dynamically reconstructed path and providing corresponding operation correction information, include: splicing the local reconstructed path with the original optimal risk-avoidance hoisting trajectory, eliminating kinematic abrupt changes at the splicing points using a Bézier curve smoothing transition algorithm, generating a corrected optimal risk-avoidance hoisting trajectory, and ensuring the overall continuity of the trajectory; based on the corrected trajectory, analyzing the adjustment amounts of the operation parameters that the hoisting equipment needs to perform, including the slewing angle correction value, the luffing speed adjustment coefficient, the lifting height compensation value, and the load stability control parameters, such as suppressing load sway by fine-tuning the speed. Then, multi-dimensional operation correction information is output through the human-machine interface. For example, the corrected local path, the changed location of high-risk areas, and the operation parameter adjustment values ​​are highlighted in the 3D visualization interface. For example, key operation instructions are broadcast in natural language, such as "Please note: Cable swaying has occurred 3 meters ahead; it is recommended to reduce the slewing speed by 20%." Furthermore, when an emergency risk is detected (such as a rapidly approaching dynamic obstacle), an audible and visual alarm can be triggered, and the priority of hazard avoidance operations can be displayed. Real-time feedback on operational corrections is recorded, and the actual obstacle avoidance effect of the corrected trajectory is verified using dual radar modules. If a risk still exists, the aforementioned local reconstruction and correction steps are repeated until safe operating conditions are met. These steps ensure the continuity and stability of the lifting equipment's movement through smooth trajectory splicing, avoiding safety hazards caused by sudden path changes. Detailed operational parameter analysis provides operators with precise adjustment guidelines, reducing operational difficulty. Multi-dimensional information prompts cater to both visual and auditory senses, ensuring operators can quickly obtain and understand correction requirements. Emergency alarms and repeated verification mechanisms enhance safety assurance capabilities in extreme situations. These steps achieve accurate transmission and effective execution of operational correction information, helping operators respond promptly to changes in the lifting scenario and ensuring that lifting operations remain safe and stable during dynamic adjustments.

[0082] As an optional embodiment, after re-optimizing the optimal avoidance hoisting trajectory in step S105, a digital twin scene can be built that includes a digital twin of the hoisting equipment, a 3D model of the high-risk area, and operational environment elements. The physical performance parameters of the hoisting equipment, including the maximum lifting capacity, slewing angular velocity range, and luffing speed limit, as well as the dynamic characteristic parameters of the high-risk area, including the dynamic obstacle movement rate and cable swing amplitude, and the operational environment parameters, including the hoisting weight swing coefficient corresponding to the wind speed level, are integrated into the digital twin scene to construct a multi-physics simulation environment consistent with the actual operational scenario. The optimal avoidance hoisting trajectory is then input into the digital twin scene, the virtual operation program is started, and the trajectory simulation verification algorithm is called simultaneously to perform dual verification of spatiotemporal collision detection and equipment operation constraint verification. Understandably, in the dual verification process, the first verification, spatiotemporal collision detection, uses an algorithm to discretize the trajectory into high-density sampling points. It calculates the minimum spatial distance between each sampling point and the 3D model of the high-risk area, ensuring that the distances between all sampling points and the tower arm intersection area, overhead cables, the slewing range of the crane arm, and dynamic obstacles are all greater than the corresponding safety thresholds. The safety threshold for dynamic obstacles is dynamically adjusted according to the real-time movement speed. Simultaneously, it detects whether the trajectory exceeds the operating radius of the hoisting equipment. The second verification is equipment operation constraint verification. The algorithm extracts equipment parameters (slewing angle change rate, luffing speed, lifting acceleration, etc.) and load status parameters (load swing angle, wire rope stress, etc.) in real time during trajectory operation, comparing them with preset physical performance parameter thresholds to ensure no parameters exceed limits, such as the slewing angle change rate not exceeding the equipment's maximum allowable value and the wire rope stress not exceeding the safe load. During virtual operation, the digital twin scene can be used to visualize the trajectory's running status, the relative positions of high-risk areas, and key equipment parameter curves in real time. When potential collision risks or parameter exceedances are detected, the coordinates of abnormal points and corresponding parameter values ​​are automatically recorded, triggering a local trajectory correction prompt. If there is no collision risk and all parameters are within the constraints, the dual verification is deemed successful, and the optimal avoidance hoisting trajectory is output as the final execution trajectory. If there are verification failures, the abnormal data is fed back to the path obstacle avoidance algorithm, triggering secondary trajectory optimization. The improved fast exploration random tree algorithm is used as the path obstacle avoidance algorithm, using the initial path point set as an index to perform path search within the 3D model constraint space until an optimal avoidance hoisting trajectory that satisfies all verification conditions is generated.

[0083] In the above optional embodiments, after the optimal hoisting trajectory for risk avoidance is obtained through re-optimization, a digital twin scenario is first built. Taking the operation of a tower crane hoisting large equipment in a chemical industrial park as an example, the system will construct a virtual scenario that includes a digital twin of the tower crane, a 3D model of the high-risk area of ​​the construction area, and operational environment elements. The physical performance parameters of the tower crane, such as maximum lifting capacity, slewing angular velocity range, and luffing speed limit, the dynamic characteristic parameters of the high-risk area, such as the movement speed of dynamic obstacles such as transport vehicles, the swing amplitude of cables crossing the work area, and operational environment parameters, such as the swing coefficient of the hoisting load corresponding to the current wind speed level, are all integrated into the digital twin scenario to create a multi-physics simulation environment consistent with the actual on-site working conditions, ensuring that the virtual scenario can truly reflect the working state of the physical world.

[0084] Next, the optimal hoisting trajectory for risk avoidance is input into the digital twin scenario, and the virtual operation program is started, simultaneously calling the trajectory simulation verification algorithm for dual verification. In the first layer of spatiotemporal collision detection, the algorithm discretizes the trajectory into a large number of high-density sampling points, and calculates the minimum spatial distance between each sampling point and the 3D model of high-risk areas such as tower arm intersection areas, overhead cables, jib slewing range, and dynamic obstacles. For example, it checks whether the distance between a sampling point on the trajectory and the cable model is greater than the safety threshold. The safety threshold for dynamic obstacles is flexibly adjusted according to its real-time movement speed. At the same time, it also checks whether the trajectory exceeds the tower crane's operating radius to avoid situations where the equipment cannot reach or operates beyond its range. In the second layer of equipment operation constraint verification, the algorithm extracts various parameters of the tower crane in real time during trajectory operation, such as the rate of change of slewing angle, luffing speed, lifting acceleration, as well as the swing angle of the suspended load and the wire rope stress value. These parameters are compared with preset physical performance parameter thresholds to ensure that the rate of change of slewing angle does not exceed the maximum allowable value of the tower crane and that the wire rope stress does not exceed the safe load, preventing equipment overload or operation exceeding parameter limits.

[0085] During virtual operation, the digital twin scenario displays the trajectory's operational status in real-time in a visual format. This includes the movement trajectory of the tower crane's boom within the virtual scene, the relative position of high-risk areas to the tower crane, and the generation of key equipment parameter curves such as slewing speed variation curves and load swing angle curves. If a sampling point is detected to be close to a safety threshold in distance from a dynamic obstacle, or if the lifting acceleration exceeds a preset range, the coordinates and corresponding parameter values ​​of the abnormal point are automatically recorded, triggering a local trajectory correction prompt. If no issues are found during dual verification (i.e., no collision risk and all parameters are within constraints), the optimal avoidance hoisting trajectory is deemed valid, and this trajectory is output as the final execution trajectory. If any verification fails, such as a trajectory segment exceeding the operating radius or the wire rope stress exceeding limits, the abnormal data is fed back to the path obstacle avoidance algorithm, triggering a second trajectory optimization. This involves re-using an improved fast-exploration random tree algorithm, using the initial path point set as an index, to re-search for a path within the 3D model's constraint space, iterating repeatedly until an optimal avoidance hoisting trajectory that satisfies all verification conditions is generated.

[0086] The above steps, by building a high-fidelity digital twin scenario, enable the pre-simulation of the hoisting trajectory in a virtual environment, allowing for the early detection of potential risks in physical operations and avoiding the trial-and-error costs of actual operation. A dual verification mechanism comprehensively verifies the trajectory from two core dimensions: spatial safety and equipment operation. This eliminates the risk of collisions and prevents mechanical damage or malfunctions caused by parameters exceeding limits. Visual display and anomaly alerts allow operators to intuitively understand the trajectory issues, while a secondary optimization mechanism ensures that the final output trajectory fully complies with safety and equipment operation requirements, improving the reliability and safety of the hoisting trajectory.

[0087] In this embodiment, the deep integration of collaborative perception of dual radar modules and intelligent algorithms can improve the dynamic perception accuracy and real-time obstacle avoidance capability of obstacles in complex construction scenarios, enhance hoisting path planning and operation efficiency, and meet the safety and efficiency requirements of high-altitude cross-operations, multi-equipment collaboration and harsh environments.

[0088] After introducing the method of the exemplary embodiments of this application, the following describes an aerial high-risk area identification and hoisting path reconstruction system according to an exemplary embodiment of this application. The system includes: a construction module for mounting two rotating radars on the rotatable support of a tower crane, forming a radar sensing module containing a dual-radar module, the radar sensing module being communicatively connected to a processing module; the radar sensing module, using the first rotating radar in the dual-radar module as an upward-looking radar and the second rotating radar as a downward-looking radar, dynamically scanning the tower arm intersection area, overhead cables, and the slewing range of the crane boom in the construction scene based on a combined upward-looking and downward-looking radar sensing strategy, to obtain real-time location information of the potential risk areas; and a processing module for receiving the real-time location information and performing three-dimensional coordinate transformation. The algorithm converts the real-time location information into real-time three-dimensional coordinate data consistent with the motion coordinate system of the hoisting equipment; constructs a three-dimensional model of the potential risk area based on the real-time three-dimensional coordinate data; identifies high-risk areas by recognizing the three-dimensional model, and uses an improved fast exploration random tree algorithm as a path avoidance algorithm, dynamically reconstructing the user-set initial hoisting path in conjunction with the three-dimensional model to generate an optimal avoidance hoisting trajectory that avoids the high-risk areas; dynamically reconstructs the local path of the hoisting equipment in the optimal avoidance hoisting trajectory by combining the real-time operation status obtained from real-time acquisition of hoisting operations and the real-time updated real-time scanning information of the hoisting area; corrects the optimal avoidance hoisting trajectory according to the dynamically reconstructed path and prompts corresponding operation correction information to ensure the stability and safety of the hoisting operation. The above system can implement the steps described in the above method implementation, and the specific implementation methods of each step will not be repeated here.

[0089] After introducing the methods and systems of the exemplary embodiments of this application, a terminal device of the exemplary embodiments of this application will be described next. The terminal device can implement the steps described in the above method embodiments, and the specific implementation of each step will not be repeated here.

[0090] After introducing the methods, systems, and terminal devices of exemplary embodiments of this application, the computer-readable storage medium of exemplary embodiments of this application will now be described. The computer-readable storage medium may be an optical disc or other form of physical medium storing a computer program (i.e., a program product). When the computer program is run by a processor, it implements the steps described in the above method embodiments. The specific implementation methods of each step will not be repeated here.

[0091] It should be noted that the above-described embodiments are merely specific implementations of this application, used to illustrate the technical solutions of this application, and not to limit them. The protection scope of this application is not limited thereto. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that any person skilled in the art can still modify or easily conceive of changes to the technical solutions described in the foregoing embodiments, or make equivalent substitutions for some of the technical features, within the scope of the technology disclosed in this application. Such modifications, changes, or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be covered within the protection scope of this application. Therefore, the protection scope of this application should be determined by the protection scope of the claims.

Claims

1. A method for identifying high-risk areas in the air and reconstructing hoisting paths, characterized in that, The method includes: Two rotating radars are installed on the rotatable support of the tower crane to form a radar sensing module containing dual radar modules. The radar sensing module is communicatively connected to the processing module. Using a radar sensing module, with the first rotating radar in the dual radar module acting as an upward-looking radar and the second rotating radar acting as a downward-looking radar, a sensing strategy combining upward-looking and downward-looking radars is employed to dynamically scan potential risk areas in the construction scenario and obtain real-time location information of these potential risk areas. These potential risk areas include at least one of the following: tower arm intersection area, cable crossing area, area where the hoisting assembly is located, crane arm movement range, excavator arm movement range, counterweight arm movement range, and loader arm movement range. The processing module receives the real-time position information and uses a three-dimensional coordinate transformation algorithm to convert the real-time position information into real-time three-dimensional coordinate data in the motion coordinate system of the hoisting equipment. A three-dimensional model of the potential risk area is constructed based on the real-time three-dimensional coordinate data. High-risk areas in the 3D model are identified, and an improved fast exploration random tree algorithm is used as the path avoidance algorithm. Combined with the 3D model, the initial hoisting path set by the user is dynamically reconstructed to generate an optimal hoisting trajectory that avoids the high-risk areas, including: The user-defined initial hoisting path is parsed into a set of discrete path points in the hoisting equipment's motion coordinate system. Motion vectors of adjacent path points are calculated to form initial path points; these vectors include direction and velocity parameters. Based on the 3D model, a spatial obstacle set is constructed. The 3D models of tower arm intersection areas, cable crossing areas, crane boom slewing range, and dynamic obstacles are transformed into prohibited spatial geometric constraints. Multi-level safety distance thresholds are labeled for different types of potential risk areas. A hazardous area semantic segmentation model constructed using U-Net is employed, combined with prohibited spatial geometric constraints, to classify different types of potential risk areas into risk levels, and the classification results are encoded as different risk levels. An improved fast exploration random tree algorithm is used as the path avoidance algorithm, with the initial path points as the index. The algorithm performs path search within the constraint space of a 3D model. Using the 3D model as the obstacle constraint input algorithm, a path cost function is set, and multiple candidate avoidance paths are generated through iterative optimization. The geometric constraints corresponding to the 3D model are used as collision detection conditions. Each newly generated path node is immediately checked for spatial overlap with the constraint region; if overlap exists, the newly generated path node is discarded, and a new branch is generated. Multiple candidate avoidance paths are evaluated against multiple objectives, and a weighted score is calculated based on the hoisting operation efficiency parameter. The path with the highest comprehensive score is selected as the optimal candidate trajectory. A B-spline curve fitting algorithm is used to smooth the optimal candidate trajectory, eliminating acute inflection points in the path, ensuring the trajectory meets the kinematic constraints of the hoisting equipment, and generating a continuously executable optimal avoidance hoisting trajectory.

2. The method for identifying high-risk areas in the air and reconstructing hoisting paths according to claim 1, characterized in that, The radar sensing module uses a dual-radar module, with the first rotating radar in the dual-radar module acting as an upward-looking radar and the second rotating radar acting as a downward-looking radar. Based on a sensing strategy combining upward-looking and downward-looking radars, it dynamically scans potential risk areas in the construction scene to obtain real-time location information of the potential risk areas, including: Based on the initial configuration of the scanning pitch angle range, horizontal scanning range and scanning frequency, an upward-looking radar is set up to cover the dynamic risk area above and to the side of the crane boom. Based on the initial configuration of the scanning elevation angle range, horizontal scanning range and scanning frequency, a top-view radar is set up to cover the area below the crane boom and the low-altitude operating area. The time synchronization deviation of the two radars is calibrated, and the priority of potential risk areas in the construction scenario is obtained. The priority of potential risk areas is dynamically configured based on the historical activity information of tower cranes in the construction scenario. The system receives tower crane attitude data in real time, including slewing angle, luffing length, and lifting height. Based on the tower crane attitude data and the priority of potential risk areas, a risk priority-driven dynamic scanning scheduling algorithm is used to predict the type and distance of the scanning area. Based on the prediction results, the system controls the upward-looking radar and downward-looking radar to execute scanning optimization strategies corresponding to the scanning area type to obtain target point cloud data. The scanning strategy includes the scanning angles corresponding to the upward-looking radar and downward-looking radar, respectively. After collecting target point cloud data, spatiotemporal dimension verification is performed on the same target location information in the target point cloud data to obtain the real-time location information.

3. The method for identifying high-risk areas in the air and reconstructing hoisting paths according to claim 2, characterized in that, The process of predicting the type and distance of the scanning area based on the tower crane attitude data and the priority of potential risk areas, combined with a risk priority-driven dynamic scanning scheduling algorithm, includes: The real-time received tower crane attitude data is spatiotemporally fused with the user-set initial hoisting path data to construct a position association model between the crane boom movement trajectory and potential risk areas within the operating area. The dynamic characteristics of each potential risk area are collected synchronously, including the presence of dynamic obstacles and the speed of obstacle movement. The dynamic scanning scheduling algorithm based on multiple weights defines multiple weights as constituent dimensions, including priority weights, distance weights, and dynamic weights, and configures preset weight coefficients for each weight dimension. The fused location association model and the dynamic characteristics of each potential risk area are used to calculate the comprehensive weight of each potential risk area by weighted summation according to the multi-weight dynamic scanning scheduling algorithm. The areas to be scanned are then sorted from high to low according to the comprehensive weight to determine the priority of the areas to be scanned. A dynamic geometric topology analysis algorithm is used to determine the type of the area to be scanned with the highest comprehensive weight, and to calculate the straight-line distance from the current position of the crane boom to the area to be scanned.

4. The method for identifying high-risk areas in the air and reconstructing hoisting paths according to claim 3, characterized in that, The method of controlling the upward-looking radar and the downward-looking radar to execute scanning optimization strategies corresponding to the scanning area type based on the prediction results to obtain target point cloud data includes: When a tower arm intersection is identified and the straight-line distance is less than the preset safety distance, the first scanning optimization strategy is triggered. The upward-looking radar is controlled to reduce the horizontal scanning range and increase the scanning frequency to focus on the dynamic outline of the intersection. The downward-looking radar is controlled to expand the horizontal scanning range and increase the scanning frequency to simultaneously capture the associated area below the intersection, ensuring that the overlap coverage of the dual radar fields of view reaches the set ratio and the original point cloud data corresponding to the intersection is updated regularly. When the area is determined to be affected by the cable and the straight-line distance is less than the preset safety distance, the second scanning optimization strategy is triggered. The upward-looking radar is controlled to adjust the scanning elevation angle range and enable narrow beam scanning to enhance the identification of cable reflection signals. The downward-looking radar is controlled to adjust the scanning elevation angle range and enable wide beam scanning to capture the cable sag trajectory. Through beam cooperative scheduling, the dual radars can synchronously collect the upper and lower end trajectories of the cable to obtain the original point cloud data corresponding to the cable. When the crane boom's slewing range is determined and a dynamic obstacle is detected, the third scanning optimization strategy is triggered. The upward-looking radar and the downward-looking radar maintain their initial horizontal scanning range and simultaneously increase the scanning frequency. The crane boom's slewing range is covered by regional segmented scanning to avoid missing obstacles. The original point cloud data corresponding to the slewing range and dynamic obstacles are collected. The original point cloud data collected under each scanning optimization strategy is timestamped to ensure time synchronization of the dual radar data, and noise interference points are filtered out by a filtering algorithm to form target point cloud data that can be used for subsequent spatiotemporal dimension verification.

5. The method for identifying high-risk areas in the air and reconstructing hoisting paths according to claim 3, characterized in that, After obtaining target point cloud data by controlling the upward-looking radar and the downward-looking radar to execute scanning optimization strategies corresponding to the scanning area type based on the prediction results, the method further includes: Multimodal feature association processing is performed on the target point cloud data to extract and associate feature parameters acquired by dual radars for different types of potential risk areas; Specifically, for the tower arm intersection area, the continuity of the tower arm metal reflectivity collected by the upward-looking radar and the intersection area projection profile collected by the downward-looking radar is correlated, and the dynamic motion vector between the tower arm metal reflectivity and the intersection area projection profile continuity is combined to determine the intersection area boundary; for the cable influence area, the cable slender linear characteristics collected by the upward-looking radar and the cable sag curvature collected by the downward-looking radar are correlated, and the dielectric constant feedback is combined to eliminate metal pipeline interference and determine the cable three-dimensional trajectory; for the boom slewing range, the boom structural profile collected by the upward-looking radar and the distance to obstacles in the slewing area collected by the downward-looking radar are correlated, and the real-time boom slewing speed is combined to delineate the slewing safety boundary. After multimodal feature association processing, optimized location information of each high-risk area is obtained after feature verification, which is used as the real-time location information for subsequent three-dimensional coordinate transformation.

6. The method for identifying high-risk areas in the air and reconstructing hoisting paths according to claim 1, characterized in that, The construction of a 3D model of the potential risk area based on the real-time 3D coordinate data includes: The real-time three-dimensional coordinate data is preprocessed and completed using an interpolation algorithm to form a potential risk area coordinate dataset; Differentiated modeling strategies are adopted for different types of potential risk areas, and sub-models for each potential risk area are constructed by combining the potential risk area coordinate dataset; The sub-models of each potential risk area are integrated according to the spatial position of the hoisting equipment's motion coordinate system. The consistency of the spatial position of each sub-model is ensured by the coordinate alignment algorithm, forming a three-dimensional model that includes all potential risk areas.

7. The method for identifying high-risk areas in the air and reconstructing hoisting paths according to claim 1, characterized in that, After generating the continuously executable optimal hazard-avoidance hoisting trajectory, the process also includes: A digital twin scene is constructed, which includes a digital twin of the hoisting equipment, three-dimensional models of risk areas of different levels, and operational environment elements. The physical performance parameters of the hoisting equipment, including the maximum lifting capacity, slewing angular velocity range, and luffing speed limit, as well as the dynamic characteristic parameters of high-risk areas, including the dynamic obstacle movement rate and cable swing amplitude, and the operational environment parameters, including the hoisting weight swing coefficient corresponding to the wind speed level, are integrated into the digital twin scene to construct a multiphysics simulation environment consistent with the actual operational scenario. The optimal hoisting trajectory is input into the digital twin scenario, the virtual operation program is started, and the trajectory simulation verification algorithm is called simultaneously to perform dual verification of spatiotemporal collision detection and equipment operation constraint verification. During virtual operation, the digital twin scene displays the trajectory operation status, the relative position of high-risk areas, and the key parameter curves of the equipment in real time. When a potential collision risk or parameter exceedance is detected, the coordinates of the abnormal point and the corresponding parameter value are automatically recorded, and a local trajectory correction prompt is triggered. If there is no collision risk and all parameters are within the constraints, the double verification is deemed to be valid and the optimal avoidance hoisting trajectory is output as the final execution trajectory.

8. The method for identifying high-risk areas in the air and reconstructing hoisting paths according to claim 1, characterized in that, After generating the optimal hoisting trajectory to avoid the high-risk area, the process also includes: Real-time acquisition of hoisting operation status data; The real-time scanning information of the hoisting area is dynamically updated by dual radar modules according to a preset high-frequency scanning cycle. The real-time scanning information includes the location of newly emerging dynamic obstacles, changes in the boundary of high-risk areas, and environmental interference parameters. By combining real-time operation status data and real-time scanning information of the hoisting area, the local path of the hoisting equipment in the optimal avoidance hoisting trajectory is dynamically reconstructed. The optimal avoidance hoisting trajectory is corrected according to the dynamically reconstructed path, and corresponding operation correction information is displayed.

9. A system for identifying high-risk areas in the air and reconstructing hoisting paths, characterized in that, The system includes: The module is used to install two rotating radars on the rotatable support of the tower crane to form a radar sensing module containing dual radar modules. The radar sensing module is communicatively connected to the processing module. The radar perception module is used to dynamically scan potential risk areas in the construction scene using a perception strategy that combines the first rotating radar in the dual radar module (looking up) and the second rotating radar (looking down), and to obtain the real-time location information of the potential risk areas. The potential risk areas include at least one of the following: tower arm intersection area, cable crossing area, area where the hoisting group is located, crane arm movement range, excavator arm movement range, counterweight arm movement range, and crane arm movement range. The processing module is used to receive the real-time location information, convert the real-time location information into real-time three-dimensional coordinate data in the motion coordinate system of the hoisting equipment using a three-dimensional coordinate transformation algorithm; construct a three-dimensional model of the potential risk area based on the real-time three-dimensional coordinate data; identify high-risk areas in the three-dimensional model, and use an improved fast exploration random tree algorithm as a path obstacle avoidance algorithm, combined with the three-dimensional model, to dynamically reconstruct the initial hoisting path set by the user, and generate the optimal avoidance hoisting trajectory to avoid the high-risk areas; Specifically, the processing module, when generating the optimal hoisting trajectory to avoid the high-risk area, is used to: parse the user-set initial hoisting path into a set of discrete path points in the hoisting equipment's motion coordinate system; calculate the motion vectors of adjacent path points to form initial path points; the motion vectors of adjacent path points include direction and velocity parameters; based on the 3D model, construct a set of spatial obstacles, transforming the 3D model of the tower arm intersection area, the influence area of ​​the overhead cable, the swing range of the crane arm, and dynamic obstacles into spatial geometric constraints prohibiting passage, and labeling different types of potential risk areas with multi-level safety distance thresholds; using a dangerous area semantic segmentation model constructed with U-Net, combined with the spatial geometric constraints prohibiting passage, to classify different types of potential risk areas into risk levels, and encoding the classification results into different levels of risk areas; and employing an improved fast exploration random tree algorithm. As a path avoidance algorithm, a path search is performed within the constraint space of a 3D model using the initial path point as an index. The 3D model is used as the obstacle constraint input algorithm, and a path cost function is set to generate multiple candidate avoidance paths through iterative optimization. The geometric constraints corresponding to the 3D model are used as collision detection conditions. Each time a new path node is generated, it is immediately checked for spatial overlap with the constraint region. If overlap exists, the newly generated path node is discarded, and a new branch is generated. Multiple candidate avoidance paths are evaluated against multiple objectives, and a weighted score is calculated based on the hoisting operation efficiency parameter. The path with the highest comprehensive score is selected as the optimal candidate trajectory. A B-spline curve fitting algorithm is used to smooth the optimal candidate trajectory, eliminating acute inflection points in the path, ensuring the trajectory meets the kinematic constraints of the hoisting equipment, and generating a continuously executable optimal avoidance hoisting trajectory.

Images