Tower crane multi-region fusion perception method and system based on rotating radar module

Abstract

The application relates to the technical field of engineering machinery safety monitoring, and provides a tower crane multi-region fusion perception method and system based on a rotating radar module. The method controls the double radar modules to rotate in coordination according to a preset rotating speed; dynamically allocates tasks to the double radar modules according to the real-time rotating angles and monitoring ranges of the double radar modules; controls the double radar modules to collect real-time data in the corresponding regions according to the allocated monitoring tasks, and obtains point cloud data and distance data of each monitoring region; performs field-of-view fusion processing on the point cloud data and distance data of each monitoring region collected by the double radar modules, and generates fusion monitoring data; performs space mapping based on the fusion monitoring data, and constructs a global space model containing a boom region, a hook path, a rear tower body, a platform operation region and a ground landing point, so that continuous monitoring of a tower crane operation environment can be realized in the global space model. The method significantly improves the intelligent level and response speed of tower crane operation safety monitoring.

Description

Technical Field

[0001] This application relates to the field of engineering machinery safety monitoring technology, and in particular to a multi-area fusion perception method and system for tower cranes based on a rotating radar module. Background Technology

[0002] In the construction industry, tower cranes are core hoisting equipment, making the safety monitoring of their operating environment crucial. Traditional tower crane monitoring technologies typically use a single sensor (such as a camera, fixed radar, or ultrasonic sensor) to perform localized monitoring of a specific area (such as the boom tip or ground landing point), which has the following significant drawbacks:

[0003] 1. Fragmented monitoring range: A single sensor can only cover a limited area (such as the boom area or the ground landing point), and cannot simultaneously monitor multiple basic areas such as the boom area, hook path, behind the tower, and platform operation area, resulting in a large number of monitoring blind spots during tower crane operation.

[0004] 2. Fixed task allocation: In existing technologies, the monitoring tasks of sensors are usually preset (such as fixed monitoring of a certain area), and the monitoring focus cannot be dynamically adjusted according to the real-time operation status of the tower crane (such as lifting load, hook height changes). This results in insufficient data collection frequency in high-risk areas (such as hook path) and waste of resources in low-risk areas.

[0005] 3. Weak data fusion capability: For multi-sensor monitoring scenarios, traditional methods simply overlay data from different sensors without removing redundant data in the overlapping areas of the dual radar modules' fields of view, or establishing a global spatial model to systematically fuse data from multiple regions. This results in low accuracy and poor real-time performance of the monitoring data, making it difficult to support risk warning and path planning in the hoisting safety control center.

[0006] 4. Inadequate environmental modeling: Existing technologies lack detailed modeling of the tower crane operating environment and do not divide the sensing areas into multiple levels according to risk levels. They cannot achieve a combination of high-precision real-time monitoring of core high-risk areas (such as platform operation areas) and efficient scanning of low-risk areas, making it difficult to meet the safety monitoring needs of complex construction environments.

[0007] Therefore, a method is urgently needed to solve at least one of the above problems. Summary of the Invention

[0008] This application provides a multi-area fusion perception method and system for tower cranes based on rotating radar modules, aiming to solve the technical problems that have not yet emerged in the prior art, such as covering multiple basic areas through the coordinated rotation of dual radar modules, dynamically allocating tasks based on real-time angle and monitoring range, and achieving continuous monitoring of the entire domain through field fusion and multi-level modeling.

[0009] In a first aspect, embodiments of this application provide a multi-area fusion sensing method for tower cranes based on a rotating radar module; the method includes:

[0010] The dual radar modules are controlled to rotate in tandem at a preset speed, so that the monitoring range of the corresponding first radar module and second radar module respectively covers different foundation areas of the tower crane. The foundation areas include at least two areas among the boom area, hook path, behind the tower, platform working area and ground landing point; the dual radar modules include a first radar module and a second radar module.

[0011] Based on the real-time rotation angle and monitoring range of the dual radar modules, dynamic task allocation is performed on the dual radar modules, so that the first radar module and the second radar module respectively undertake the real-time data acquisition task of different monitoring areas, wherein the monitoring area is any one or more combinations of the basic areas.

[0012] The dual radar modules are controlled to collect real-time data from the corresponding areas according to the assigned monitoring tasks, and to obtain point cloud data and distance data for each monitoring area; the point cloud data and distance data collected by the dual radar modules for each monitoring area are subjected to field-of-view fusion processing to remove redundant data from repeated monitoring areas and generate fused monitoring data;

[0013] Based on the fused monitoring data, spatial mapping is performed to construct a global spatial model that includes the boom area, hook path, behind the tower, platform work area, and ground landing point. Multi-level sensing areas are then divided within the global spatial model to achieve continuous monitoring of the tower crane's operating environment.

[0014] In some embodiments, the continuous monitoring of the tower crane's operating environment in the global spatial model includes: selecting a target sensing area corresponding to the current operating state of the tower crane in the multi-level sensing area according to the current operating state of the tower crane; generating a multi-area fusion sensing result based on real-time data of the target sensing area; inputting the multi-area fusion sensing result to the hoisting safety control center for risk warning, path judgment, and target identification; and continuously looping the operations of dual radar module coordinated rotation, dynamic task allocation, data acquisition, field-of-view fusion processing, spatial mapping, and generation of fusion sensing results to achieve continuous monitoring of the tower crane's operating environment.

[0015] In some embodiments, generating multi-region fusion sensing results based on real-time data of the target sensing area includes: acquiring point cloud data and distance data of each monitoring sub-region within the target sensing area; performing coordinate unification processing on the data of each sub-region based on a preset spatial coordinate transformation algorithm to establish a unified data format including three-dimensional spatial coordinates, monitoring timestamps, and data confidence scores; matching and deduplicating duplicate data of the same monitoring target in different sub-regions through preset dynamic data association rules, retaining the valid data with the highest confidence scores; and performing feature fusion on the processed valid data according to a preset fusion strategy to generate multi-region fusion sensing results including target location information, motion trajectory information, and environmental attribute information, wherein the fusion strategy dynamically adjusts the data fusion weights according to the current tower crane operation status.

[0016] In some embodiments, inputting the multi-area fusion perception results to the hoisting safety control center for risk warning, path determination, and target identification includes: encrypting and standardizing the generated multi-area fusion perception results, and transmitting them to the hoisting safety control center in real time through a preset communication interface; after receiving the data, the hoisting safety control center calculates the risk level of obstacle distances, movement trends, and tower crane component positions within the target perception area based on a preset risk assessment model, and generates corresponding risk warning signals; it calls a preset path planning algorithm to analyze real-time environmental data of the hook path area, determines whether there is a collision risk in the current hoisting path, and outputs path adjustment suggestions; and it uses a target identification model deployed in the hoisting safety control center to identify and locate personnel and equipment in the ground landing area and platform operation area, generating target location coordinates and category information.

[0017] In some embodiments, the continuous cyclic execution of the operations of coordinated rotation of the dual radar modules, dynamic task allocation, data acquisition, field-of-view fusion processing, spatial mapping, and generation of fused perception results to achieve continuous monitoring of the tower crane's operating environment includes: setting a time interval threshold or event triggering condition for cyclic execution; starting a new round of cyclic operation when the preset time interval is reached or a change in the tower crane's operating status is detected; before each cycle begins, acquiring the current rotation angle of the dual radar modules, the equipment operating status, and the tower crane's real-time operating parameters as initial data for dynamic task allocation; during the cycle, caching and version management of the processed data to ensure that the monitoring data of two adjacent cycles have temporal continuity and spatial correlation; if abnormal data or equipment fault signals are detected during the cycle, triggering an emergency handling process, temporarily freezing the current monitoring task and maintaining the output of the last valid data until the anomaly is resolved and cyclic execution resumes.

[0018] In some embodiments, the step of dynamically allocating tasks to the dual radar modules based on their real-time rotation angle and monitoring range, so that the first and second radar modules respectively undertake real-time data acquisition tasks for different monitoring areas, includes: establishing a preset monitoring area priority mapping table, wherein the priority mapping table dynamically sorts the monitoring priorities of the boom area and the base area of ​​the hook path according to the current lifting load, hook height, and operation stage of the tower crane; calculating in real time the effective monitoring range corresponding to the current rotation angle of the first and second radar modules, generating a three-dimensional coordinate matrix of the radar field of view coverage area; and, based on the monitoring area priority mapping table and the radar field of view coverage area, using a task scheduling algorithm to allocate high-priority monitoring areas to the radar modules with more complete current field of view coverage, and allocating spare monitoring cycles to low-priority areas, wherein the task scheduling algorithm is based on a greedy strategy to prioritize satisfying the real-time data acquisition frequency requirements of high-priority areas.

[0019] In some embodiments, the point cloud data and distance data collected by the dual radar modules in each monitoring area are subjected to field-of-view fusion processing to remove redundant data in overlapping monitoring areas and generate fused monitoring data. This includes: establishing a spatial coordinate transformation relationship between the first radar module and the second radar module based on the installation position parameters and real-time rotation angle of the dual radar modules, and unifying the monitoring data of the first radar module and the second radar module into the tower crane coordinate system; performing time synchronization processing on the point cloud data after unifying the coordinate system, and removing invalid data with time differences exceeding a preset threshold according to the data acquisition timestamp; for overlapping monitoring areas, by calculating the spatial distance deviation and signal strength value of the point cloud data, setting a distance fusion threshold and strength filtering rules, retaining the valid data points with the highest signal strength and distance deviation within the threshold, and generating a non-redundant fused monitoring dataset.

[0020] In some embodiments, the step of constructing a spatial model based on the fused monitoring data, including the boom area, hook path, tower back, platform work area, and ground landing point, includes: processing the point cloud information in the fused monitoring data using a simultaneous positioning and mapping algorithm to generate a three-dimensional point cloud map of the tower crane's operating environment; marking the spatial boundaries and feature parameters of basic areas such as the boom area and hook path in the three-dimensional point cloud map according to preset area division rules, and establishing digital labels for each area; dynamically correcting the three-dimensional point cloud map using real-time updated fused monitoring data, and triggering a local map update mechanism when environmental feature changes are detected to ensure the consistency between the global spatial model and the actual operating environment.

[0021] In some embodiments, dividing the global spatial model into multi-level sensing areas includes: dividing the global spatial model into a core sensing layer, an early warning sensing layer, and a regular sensing layer according to monitoring accuracy and risk level requirements; the core sensing layer covers high-risk areas such as the hook path and platform operation area, sets the highest monitoring frequency and the minimum data acquisition interval, and captures target changes with millimeter-level accuracy in real time; the early warning sensing layer covers medium-risk areas such as the boom area and ground landing point, adopts a dynamic monitoring frequency, and automatically adjusts the data acquisition interval according to the tower crane's movement speed; the regular sensing layer covers low-risk areas such as behind the tower crane, sets a basic monitoring frequency, and periodically scans environmental features; the division boundaries and monitoring parameters of each level of sensing area are automatically adjusted according to the tower crane's operation mode to form an adaptive multi-level sensing system.

[0022] Secondly, this application provides a tower crane multi-area fusion sensing system based on a rotating radar module, the system comprising:

[0023] A rotation control unit is used to control the dual radar modules to rotate in tandem at a preset speed, so that the monitoring range of the corresponding first radar module and second radar module respectively covers different foundation areas of the tower crane. The foundation areas include at least two areas among the boom area, hook path, behind the tower, platform working area, and ground landing point; the dual radar modules include a first radar module and a second radar module.

[0024] The task allocation unit is used to dynamically allocate tasks to the dual radar modules according to the real-time rotation angle and monitoring range of the dual radar modules, so that the first radar module and the second radar module respectively undertake the real-time data acquisition tasks of different monitoring areas, wherein the monitoring area is any one or more combinations of the basic areas.

[0025] The data acquisition unit is used to control the dual radar modules to collect data in real time in the corresponding areas according to the assigned monitoring tasks, and to acquire point cloud data and distance data of each monitoring area; the point cloud data and distance data of each monitoring area collected by the dual radar modules are processed by field-of-view fusion to remove redundant data of repeated monitoring areas and generate fused monitoring data.

[0026] The continuous monitoring unit is used to perform spatial mapping based on the fused monitoring data, construct a global spatial model including the boom area, hook path, behind the tower, platform operation area and ground landing point, and divide the global spatial model into multi-level sensing areas to realize continuous monitoring of the tower crane operation environment in the global spatial model.

[0027] This application provides a tower crane multi-area fusion perception method and system based on a rotating radar module. By defining a high-weight redundancy protection zone and assigning differentiated perception weights, combined with dual-radar rotation angle optimization, it achieves overlapping coverage of the hook's vertical passage and obstacle-prone areas. Compared to traditional solutions, this improves monitoring accuracy and significantly reduces collision risk. Based on the real-time calculation of the overlap of perception areas along the hoisting path, a multi-objective optimization algorithm dynamically adjusts the radar angle, prioritizing perception resources to focus on high-risk areas. This improves monitoring efficiency and reduces redundant data under the same hardware configuration. When a single radar malfunctions, an overlapping field-of-view scheduling mechanism quickly takes over the perception tasks of critical areas, avoiding blind spots and shortening fault recovery time compared to traditional single-radar solutions, ensuring operational continuity. The optimized dual-radar perception data is deeply fused with the compensation data after fault scheduling to generate a grid map containing risk levels, providing accurate decision-making basis for tower crane control and achieving a technological upgrade from "passive detection" to "active protection."

[0028] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and do not limit this application. Attached Figure Description

[0029] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0030] Figure 1 This is a schematic flowchart illustrating the steps of a multi-area fusion sensing method for tower cranes based on a rotating radar module, provided in one embodiment of this application.

[0031] Figure 2 This is a schematic diagram of the structure of a dual radar module provided in one embodiment of this application;

[0032] Figure 3 This is a schematic block diagram of a tower crane multi-area fusion sensing system based on a rotating radar module, provided in one embodiment of this application.

[0033] Figure 4 This is a schematic block diagram of the controller provided in one embodiment of this application.

[0034] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and do not limit this application. Detailed Implementation

[0035] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0036] The flowchart shown in the attached diagram is for illustrative purposes only and does not necessarily include all content and operations / steps, nor does it necessarily have to be performed in the order described. For example, some operations / steps can be broken down, combined, or partially merged, so the actual execution order may change depending on the actual situation.

[0037] It should be understood that, in order to clearly describe the technical solutions of the embodiments of the present invention, the terms "first" and "second" are used in the embodiments of the present invention to distinguish identical or similar items with essentially the same function and effect. Those skilled in the art will understand that the terms "first" and "second" do not limit the quantity or execution order, and the terms "first" and "second" are not necessarily different.

[0038] It should be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the scope of the application. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0039] It should also be understood that the term “and / or” as used in this application specification and the appended claims means any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0040] In the construction industry, tower cranes are core hoisting equipment, making the safety monitoring of their operating environment crucial. Traditional tower crane monitoring technologies typically use a single sensor (such as a camera, fixed radar, or ultrasonic sensor) to perform localized monitoring of a specific area (such as the boom tip or ground landing point), which has the following significant drawbacks:

[0041] 1. Fragmented monitoring range: A single sensor can only cover a limited area (such as the boom area or the ground landing point), and cannot simultaneously monitor multiple basic areas such as the boom area, hook path, behind the tower, and platform operation area, resulting in a large number of monitoring blind spots during tower crane operation.

[0042] 2. Fixed task allocation: In existing technologies, the monitoring tasks of sensors are usually preset (such as fixed monitoring of a certain area), and the monitoring focus cannot be dynamically adjusted according to the real-time operation status of the tower crane (such as lifting load, hook height changes). This results in insufficient data collection frequency in high-risk areas (such as hook path) and waste of resources in low-risk areas.

[0043] 3. Weak data fusion capability: For multi-sensor monitoring scenarios, traditional methods simply overlay data from different sensors without removing redundant data in the overlapping areas of the dual radar modules' fields of view, or establishing a global spatial model to systematically fuse data from multiple regions. This results in low accuracy and poor real-time performance of the monitoring data, making it difficult to support risk warning and path planning in the hoisting safety control center.

[0044] 4. Inadequate environmental modeling: Existing technologies lack detailed modeling of the tower crane operating environment and do not divide the sensing areas into multiple levels according to risk levels. They cannot achieve a combination of high-precision real-time monitoring of core high-risk areas (such as platform operation areas) and efficient scanning of low-risk areas, making it difficult to meet the safety monitoring needs of complex construction environments.

[0045] Currently, no existing technology offers a solution for achieving continuous full-domain monitoring through dual radar module collaborative rotation covering multiple basic areas, dynamic task allocation based on real-time angle and monitoring range, and field-of-view fusion and multi-level modeling. This invention effectively addresses the aforementioned shortcomings of traditional technologies through a dual radar module collaborative mechanism, dynamic task allocation strategy, and multi-level perception system, providing a novel technical approach for tower crane safety monitoring.

[0046] Please refer to Figure 1 This application provides a multi-area fusion perception method for tower cranes based on a rotating radar module, applied to the controller. It should be noted that all information involved in the method provided in this application is extracted with the authorization of the relevant user and in accordance with relevant regulations, and will not infringe on user privacy.

[0047] The provided multi-area fusion sensing method for tower cranes based on rotating radar modules includes steps S101 to S104. Details are as follows:

[0048] Step S101. Control the dual radar modules to rotate in coordination at a preset speed, so that the monitoring range of the corresponding first radar module and second radar module respectively covers different foundation areas of the tower crane. The foundation area includes at least two areas among the boom area, hook path, behind the tower, platform working area and ground landing point; the dual radar modules include the first radar module and the second radar module.

[0049] Specifically, by controlling dual radar modules (such as...) Figure 2As shown, the first radar module and the second radar module rotate in tandem at a preset speed, so that their monitoring ranges cover different foundation areas of the tower crane (such as the boom area, hook path, behind the tower, platform work area, ground landing point, etc.), solving the problem of fragmented monitoring range of traditional single sensors and realizing multi-area collaborative monitoring.

[0050] The first radar module is installed at the end of the tower crane boom or on the rotatable base, and the second radar module is installed in the middle of the tower crane tower or on the top of the cab. The two form a spatial angle difference (such as vertical or complementary angle installation) to ensure that the monitoring field of view covers different directions.

[0051] Both radar modules use mechanical rotating radar (such as lidar installed under the trolley, radar model: TF-ALS-LIDAR-01, single radar detection range ≥150 meters, accuracy ≤3cm, supports angular resolution ≤0.05°, dual radars scan synchronously at 10Hz frequency), support 360° range scanning, preset rotation speed is set according to the tower crane's working radius and monitoring accuracy requirements (such as 5-10 revolutions / minute).

[0052] By pre-calibrating the spatial coordinates of each basic area (e.g., the boom area corresponds to a fan-shaped area within the boom length, and the platform operation area corresponds to a circular area within 3 meters around the tower), the initial scanning angle range of the dual radar modules is configured. For example, the first radar module mainly covers the boom area, hook path, and ground landing point along the boom extension direction; the second radar module covers the blind spot behind the tower, the platform operation area, and surrounding obstacles. The monitoring ranges of the two partially overlap (e.g., the middle section of the hook path) to ensure no blind spots. The controller drives the dual radar modules to rotate at the same or complementary speeds via synchronous motors (e.g., the first radar rotates clockwise, and the second radar rotates counterclockwise) to ensure complete coverage of all target base areas within the rotation cycle, avoiding missed scans.

[0053] Step S102. Based on the real-time rotation angle and monitoring range of the dual radar modules, dynamically allocate tasks to the dual radar modules, so that the first radar module and the second radar module respectively undertake the real-time data acquisition tasks of different monitoring areas, wherein the monitoring area is any one or more combinations of the basic areas.

[0054] Specifically, based on the real-time rotation angle of the two radar modules (obtained through encoders or built-in sensors) and the current monitoring range, combined with the tower crane's operating status (such as lifting load, hook height, and slewing angle), the monitoring tasks of the two radar modules are dynamically adjusted so that they focus on high-risk areas, thus solving the problem of fixed task allocation.

[0055] Status parameter acquisition is achieved by acquiring tower crane sensor data in real time: hook height (via height encoder), lifting load (via load cell), tower slewing angle (via slewing encoder), wind speed (via environmental sensor), etc., and defining high-risk scenarios (such as hook path during heavy-load lifting, when personnel approach the platform work area).

[0056] The dynamic task allocation logic includes: Triggering conditions: When the hook height is below 50 meters and the load exceeds 80% of the rated load, the hook path is determined to be a high-risk area, triggering task adjustment; when the tower crane's slewing angle points towards a densely obstructed area behind the tower, the scanning frequency of the second radar module in that area is increased. Task allocation algorithm: A priority queue is established, and monitoring resources are dynamically allocated according to the risk level. For example: High-risk areas (such as the hook path): The first radar module is allocated to monitor in real time with the highest precision (such as a 1° scanning interval); Medium-risk areas (such as the platform operation area): The second radar module increases its dwell time when rotating into this area (such as extending the scanning cycle by 20%); Low-risk areas (such as open ground landing points): Low-precision scanning (such as a 5° scanning interval) is used to reduce resource consumption.

[0057] Real-time angle matching uses the angle encoder built into the radar module to provide real-time feedback on the current scanning angle. The controller calculates the spatial overlap between the radar field of view and the target area and dynamically switches monitoring tasks (such as starting the ground landing point monitoring mode when the first radar rotates to the end of the boom).

[0058] Step S103. Control the dual radar modules to collect real-time data in the corresponding areas according to the assigned monitoring tasks, and obtain point cloud data and distance data of each monitoring area; perform field-of-view fusion processing on the point cloud data and distance data of each monitoring area collected by the dual radar modules, remove redundant data of repeated monitoring areas, and generate fused monitoring data.

[0059] Specifically, by controlling dual radar modules to collect point cloud data and distance data according to assigned tasks, and by using spatial coordinate system calibration and field-of-view fusion algorithms to remove redundant data in overlapping areas, high-precision fused monitoring data is generated, thus solving the problem of weak traditional data fusion capabilities.

[0060] Data acquisition and preprocessing involves simultaneously acquiring raw point cloud data (including 3D coordinates and reflection intensity) and distance data (straight-line distance from radar to target) for each monitoring area using dual radar modules, with timestamp synchronization ensuring data alignment. The raw data undergoes denoising (e.g., outlier removal) and coordinate transformation: each radar coordinate system is converted to the tower crane's global coordinate system (with the tower crane base center as the origin) for easier subsequent fusion.

[0061] Field-of-view overlap detection and redundancy removal are achieved by defining overlap region discrimination criteria: when the overlap of the scanning angle ranges of two radar modules exceeds 15% and the distance measurement difference is less than 5%, it is determined to be an overlap region. A clustering algorithm based on Euclidean distance (such as DBSCAN) is used to cluster the point cloud of the overlap region, retaining the most recent measurement value or weighted average coordinates of the same target while removing duplicate points.

[0062] The fused data generation process involves directly merging data from non-overlapping areas and removing redundancy from data from overlapping areas to generate fused monitoring data. The output includes a complete and non-repeating set of point clouds and a distance matrix for each monitoring area, which is used for subsequent mapping.

[0063] Step S104. Based on the fused monitoring data, spatial mapping is performed to construct a global spatial model that includes the boom area, hook path, tower back, platform work area and ground landing point. Multi-level sensing areas are divided in the global spatial model to achieve continuous monitoring of the tower crane's working environment.

[0064] Specifically, a global spatial model encompassing all basic areas is constructed based on fused monitoring data. Multi-level sensing areas are divided according to risk level (such as core high-risk areas, medium-risk buffer zones, and low-risk scanning areas) to achieve a combination of high-precision monitoring and efficient scanning, thus solving the problem of crude environmental modeling.

[0065] The full-domain spatial mapping employs SLAM (Simultaneous Localization and Mapping) algorithms or point cloud stitching technology to integrate monitoring data into a 3D spatial model of the tower crane's operating environment, including the location information of key elements such as the boom, tower body, hook, obstacles, and ground landing point. The model update frequency is synchronized with the radar scanning frequency (e.g., 10Hz) to ensure real-time performance.

[0066] The multi-level sensing area is divided into: Core high-risk area (red area): platform operation area (within 1.5 meters around the tower body) and 3 meters below the hook, requiring scanning accuracy ≤ 5cm and data update frequency ≥ 20Hz; Medium-risk buffer zone (yellow area): area within 5 meters on both sides of the boom and dense obstacle area behind the tower, scanning accuracy ≤ 10cm and update frequency 10Hz; Low-risk scanning area (green area): area 10 meters away from the landing point on open ground, using sparse scanning (accuracy ≤ 50cm) and update frequency 5Hz.

[0067] Continuous monitoring is achieved by marking the spatial extent of each level of area in the global model. The controller dynamically adjusts scanning parameters (such as scanning interval and accuracy) based on the real-time location and task allocation of the dual radar modules, ensuring high-frequency and high-precision monitoring of core areas and efficient coverage of low-risk areas. The final output is a real-time environmental map containing risk levels, supporting safety early warning and path planning.

[0068] In some embodiments, the continuous monitoring of the tower crane's operating environment in the global spatial model includes: selecting a target sensing area corresponding to the current operating state of the tower crane in the multi-level sensing area according to the current operating state of the tower crane; generating a multi-area fusion sensing result based on real-time data of the target sensing area; inputting the multi-area fusion sensing result to the hoisting safety control center for risk warning, path judgment, and target identification; and continuously looping the operations of dual radar module coordinated rotation, dynamic task allocation, data acquisition, field-of-view fusion processing, spatial mapping, and generation of fusion sensing results to achieve continuous monitoring of the tower crane's operating environment.

[0069] In the global spatial model, the target perception area is dynamically selected based on the real-time operation status of the tower crane, the fused perception results are generated and input into the safety control center, and continuous monitoring is achieved by cyclically executing the entire process operation to ensure the real-time monitoring data and decision support capabilities.

[0070] The target sensing area is dynamically selected by establishing a mapping relationship between the operation status and the sensing area: for example, when the tower crane is in the "lifting operation" state, the target sensing area is the hook path, the boom area, and the ground landing point; when it is in the "standby maintenance" state, the target area switches to the platform operation area and the area behind the tower. The tower crane's main control system obtains the operation status (lifting / slewing / luffing / stopping) in real time, and combines it with sensor data (such as hook height and load weight) to trigger the area selection logic.

[0071] The generation and transmission of fused perception results involves extracting real-time data (such as point clouds, distance, and timestamps) from the target perception area, encapsulating it in a unified format, and transmitting it to the hoisting safety control center via wired (Ethernet) or wireless (5G) communication modules. After receiving the data, the control center synchronously displays the real-time status of each area (such as obstacle locations and personnel activity trajectories) and marks high-risk areas (highlighted in red).

[0072] The cyclic execution mechanism uses a fixed cycle period (e.g., 500ms) or event-triggered conditions (e.g., tower crane slewing angle change exceeding 10°) to set a fixed cycle period. Each cycle includes the entire process of S101-S104. Key logs (e.g., radar rotation speed, task allocation results, fused data volume) are recorded during the cycle for subsequent fault tracing and performance optimization.

[0073] In some embodiments, generating multi-region fusion sensing results based on real-time data of the target sensing area includes: acquiring point cloud data and distance data of each monitoring sub-region within the target sensing area; performing coordinate unification processing on the data of each sub-region based on a preset spatial coordinate transformation algorithm to establish a unified data format including three-dimensional spatial coordinates, monitoring timestamps, and data confidence scores; matching and deduplicating duplicate data of the same monitoring target in different sub-regions through preset dynamic data association rules, retaining the valid data with the highest confidence scores; and performing feature fusion on the processed valid data according to a preset fusion strategy to generate multi-region fusion sensing results including target location information, motion trajectory information, and environmental attribute information, wherein the fusion strategy dynamically adjusts the data fusion weights according to the current tower crane operation status.

[0074] By unifying coordinates, deduplicating duplicate data, and dynamically weighting the data in the sub-regions within the target perception area, a high-precision fusion result containing location, trajectory, and environmental attributes is generated, solving the problems of inconsistent data formats and redundancy.

[0075] Coordinate unification is achieved by defining a global coordinate system for the tower crane (X-axis eastward, Y-axis northward, Z-axis vertically upward). Data from each sub-region (such as the boom area data from the first radar and the platform data from the second radar) is converted into global coordinates using rotation and translation matrices. Data format standardization: Each point cloud data set includes three-dimensional coordinates (x, y, z), a data acquisition timestamp (accurate to milliseconds), and a data confidence score (calculated based on radar signal strength, ranging from 0 to 1).

[0076] Dynamic data association and deduplication include: establishing association rules: the coordinate deviation of the same target in adjacent sub-regions should be less than the radar accuracy (e.g., ±3cm for lidar), and the timestamp difference should be less than 50ms. The Hungarian algorithm is used to match duplicate data, retaining the data with the highest confidence (e.g., points with signal strength > 80%), and eliminating low-confidence redundant points.

[0077] The dynamic weight fusion strategy includes: defining fusion weight parameters: during hoisting operations, the weight of the hook path data is set to 0.6, and the weight of the ground landing point is 0.3; during maintenance operations, the weight of the platform work area is increased to 0.7.

[0078] Feature fusion algorithm: The target position information is weighted and averaged, the motion trajectory is predicted by Kalman filter, and environmental attributes (such as obstacle material) are judged by the reflection intensity threshold (such as the reflection intensity of metal objects > 50dB).

[0079] In some embodiments, inputting the multi-area fusion perception results to the hoisting safety control center for risk warning, path determination, and target identification includes: encrypting and standardizing the generated multi-area fusion perception results, and transmitting them to the hoisting safety control center in real time through a preset communication interface; after receiving the data, the hoisting safety control center calculates the risk level of obstacle distances, movement trends, and tower crane component positions within the target perception area based on a preset risk assessment model, and generates corresponding risk warning signals; it calls a preset path planning algorithm to analyze real-time environmental data of the hook path area, determines whether there is a collision risk in the current hoisting path, and outputs path adjustment suggestions; and it uses a target identification model deployed in the hoisting safety control center to identify and locate personnel and equipment in the ground landing area and platform operation area, generating target location coordinates and category information.

[0080] By encrypting and standardizing the transmission of fused perception results, the system supports the risk warning, path planning, and target identification of the security control center, enabling the actual business application of monitoring data.

[0081] Data preprocessing and transmission include: Encryption: The fusion results containing coordinate data are encrypted using the AES-128 algorithm, and the communication interface follows the OPC UA or MQTT protocol to ensure secure data transmission. Format standardization: The fused data is converted into JSON format, including fields such as region ID, target category (personnel / equipment / obstacle), location coordinates, and risk level.

[0082] Risk assessment and early warning include: Risk assessment model: Calculating the safe distance threshold between obstacles and the boom / hook (e.g., ≤2 meters triggers a Level 1 warning, 2-5 meters triggers a Level 2 warning), combined with movement trends (speed > 0.5m / s is considered a dynamic risk). Early warning signal output: Synchronous prompts are provided via audible and visual alarms (on-site) and the control center interface (remote). A Level 1 warning automatically triggers the tower crane deceleration mechanism.

[0083] Path planning and target recognition include: Path planning algorithm: Dijkstra's algorithm is used to plan the optimal path for the hook, avoiding medium- and high-risk areas (such as yellow / red areas), and outputting path point coordinates and speed suggestions. Target recognition model: A YOLOv8 deep learning model is deployed to classify personnel (identifying safety helmets / work clothes) and ground equipment (forklifts / material boxes) in the platform work area in real time, with a positioning accuracy of ≤50cm.

[0084] In some embodiments, the continuous cyclic execution of the operations of coordinated rotation of the dual radar modules, dynamic task allocation, data acquisition, field-of-view fusion processing, spatial mapping, and generation of fused perception results to achieve continuous monitoring of the tower crane's operating environment includes: setting a time interval threshold or event triggering condition for cyclic execution; starting a new round of cyclic operation when the preset time interval is reached or a change in the tower crane's operating status is detected; before each cycle begins, acquiring the current rotation angle of the dual radar modules, the equipment operating status, and the tower crane's real-time operating parameters as initial data for dynamic task allocation; during the cycle, caching and version management of the processed data to ensure that the monitoring data of two adjacent cycles have temporal continuity and spatial correlation; if abnormal data or equipment fault signals are detected during the cycle, triggering an emergency handling process, temporarily freezing the current monitoring task and maintaining the output of the last valid data until the anomaly is resolved and cyclic execution resumes.

[0085] By using a time / event-triggered loop mechanism, combined with data caching, version management, and exception handling, the continuity and reliability of the monitoring process are ensured, avoiding monitoring failures caused by data gaps or equipment malfunctions.

[0086] The cyclic triggering mechanism includes: Time triggering: Set a fixed time interval (e.g., 100ms) and start the cycle via timer interrupt; Event triggering: When the tower crane load change is greater than 10% or the wind speed is greater than level 6, immediately start a new round of scanning. Initial data acquisition: Before each cycle, read the radar encoder angle (accuracy ±0.1°), equipment voltage / temperature (to ensure normal radar operation), and real-time tower crane parameters (e.g., current lifting weight, hook position).

[0087] Data caching and version management include: Caching strategy: A circular buffer is used to store data from the most recent 10 cycles. Each version contains a timestamp, radar angle, and fusion result for trajectory prediction and anomaly comparison. Spatiotemporal correlation processing: Adjacent cycle data are aligned using timestamps, and spatial coordinates are matched using the ICP (Iterative Closest Point) algorithm to ensure the continuity of the monitoring area.

[0088] The anomaly handling process includes: Anomaly detection: Real-time monitoring of radar signal strength (<30dB is considered abnormal), data packet loss rate (>5% triggers an alarm), and equipment temperature (>60℃ triggers an alarm). Emergency response: When an anomaly is detected, the current task is frozen and the last valid data output is preserved (to avoid misjudgment by the control center), while a fault code is sent to the operation and maintenance platform. After the anomaly is resolved, scanning continues from the current perspective.

[0089] In some embodiments, the step of dynamically allocating tasks to the dual radar modules based on their real-time rotation angle and monitoring range, so that the first and second radar modules respectively undertake real-time data acquisition tasks for different monitoring areas, includes: establishing a preset monitoring area priority mapping table, wherein the priority mapping table dynamically sorts the monitoring priorities of the boom area and the base area of ​​the hook path according to the current lifting load, hook height, and operation stage of the tower crane; calculating in real time the effective monitoring range corresponding to the current rotation angle of the first and second radar modules, generating a three-dimensional coordinate matrix of the radar field of view coverage area; and, based on the monitoring area priority mapping table and the radar field of view coverage area, using a task scheduling algorithm to allocate high-priority monitoring areas to the radar modules with more complete current field of view coverage, and allocating spare monitoring cycles to low-priority areas, wherein the task scheduling algorithm is based on a greedy strategy to prioritize satisfying the real-time data acquisition frequency requirements of high-priority areas.

[0090] By establishing a priority mapping table for monitoring areas and combining it with the real-time field of view coverage of radar, a greedy strategy is adopted to dynamically allocate tasks, ensuring that monitoring resources are given priority in high-risk areas and optimizing the utilization of sensor resources.

[0091] The priority mapping table construction includes: defining influencing factors: lifting load (priority +1 when >80% of rated load), hook height (priority +1 when <30 meters), and operation stage (priority of lifting stage > standby stage). Priority ranking: for example, the hook path priority is the highest (Level 1) during the lifting stage, the boom area is Level 2, and the ground landing point is Level 3; during the standby stage, the platform operation area priority is increased to Level 1.

[0092] Real-time radar field of view calculation includes: Field of view modeling: Based on the radar installation location (e.g., coordinates (x1, y1, z1)) and the current rotation angle (θ), calculate the monitoring range as a fan-shaped area with radius R, and output a three-dimensional coordinate matrix (including minimum / maximum x, y, z values). Coverage calculation: Calculate the percentage of overlap between the target area and the radar field of view. An overlap rate >70% is considered "complete coverage", 30%-70% is "partial coverage", and <30% is "no coverage".

[0093] The task scheduling algorithm includes: a greedy strategy: prioritizing the allocation of Level 1 areas to the radar with the most complete coverage (highest overlap rate); if both radars cover the area, allocating it to the radar with more remaining resources (e.g., the radar with a smaller scan interval). Backup period allocation: low-priority areas (Level 3) are supplemented during radar idle periods (e.g., non-high-priority scanning periods), with a supplementary scan interval not exceeding 10 seconds.

[0094] In some embodiments, the point cloud data and distance data collected by the dual radar modules in each monitoring area are subjected to field-of-view fusion processing to remove redundant data in overlapping monitoring areas and generate fused monitoring data. This includes: establishing a spatial coordinate transformation relationship between the first radar module and the second radar module based on the installation position parameters and real-time rotation angle of the dual radar modules, and unifying the monitoring data of the first radar module and the second radar module into the tower crane coordinate system; performing time synchronization processing on the point cloud data after unifying the coordinate system, and removing invalid data with time differences exceeding a preset threshold according to the data acquisition timestamp; for overlapping monitoring areas, by calculating the spatial distance deviation and signal strength value of the point cloud data, setting a distance fusion threshold and strength filtering rules, retaining the valid data points with the highest signal strength and distance deviation within the threshold, and generating a non-redundant fused monitoring dataset.

[0095] By using coordinate transformation, time synchronization, and distance-intensity filtering rules, dual radar data are fused to eliminate redundancy in overlapping areas, generating a high-precision, non-redundant fused dataset and improving the quality of monitoring data.

[0096] Coordinate transformation and time synchronization include: Installation parameter calibration: The relative positions of the two radars (translation vectors Δx, Δy, Δz, rotation matrix R) are measured using a total station to establish the transformation relationship from the radar coordinate system to the tower crane coordinate system (P_global = R*P_radar + Δ). Time synchronization uses NTP (Network Time Protocol) to calibrate the dual radar clocks, with a time difference <1ms. Data with a timestamp difference >5ms is discarded (considered as asynchronous scanning).

[0097] Redundant data filtering rules include: Distance fusion threshold: Set a distance deviation threshold for the same target (e.g., ≤5cm for lidar, ≤20cm for millimeter-wave radar); deviations exceeding the threshold are considered different targets. Intensity filtering rules: Retain the data point with the highest signal strength (e.g., lidar reflection intensity >70%, millimeter-wave radar signal-to-noise ratio >25dB); only one valid point is retained at the same location.

[0098] The fused dataset is generated by establishing a data index table: data points are grouped according to three-dimensional coordinates (rounded to 1cm precision), and a unique valid point is retained after applying filtering rules within each group. Output format: includes fused point cloud data (x, y, z, intensity, timestamp), distance data (radar to target distance), and region labels (e.g., "hook path - high-risk area").

[0099] In some embodiments, the step of constructing a spatial model based on the fused monitoring data, including the boom area, hook path, tower back, platform work area, and ground landing point, includes: processing the point cloud information in the fused monitoring data using a simultaneous positioning and mapping algorithm to generate a three-dimensional point cloud map of the tower crane's operating environment; marking the spatial boundaries and feature parameters of basic areas such as the boom area and hook path in the three-dimensional point cloud map according to preset area division rules, and establishing digital labels for each area; dynamically correcting the three-dimensional point cloud map using real-time updated fused monitoring data, and triggering a local map update mechanism when environmental feature changes are detected to ensure the consistency between the global spatial model and the actual operating environment.

[0100] A 3D point cloud map is constructed using the SLAM algorithm, and the boundaries and features of each basic area are marked. A dynamic correction mechanism is used to ensure that the map is consistent with the actual environment, providing a spatial reference for multi-level monitoring.

[0101] The 3D point cloud map construction includes: SLAM algorithm selection: For static tower crane operation scenarios, the LOAM (Laser Odometry + Map Building) algorithm is adopted to stitch radar point cloud data in real time and generate a globally consistent point cloud map. Map resolution: 0.05m / point in the core area (platform operation area), and 0.2m / point in low-to-medium risk areas, balancing accuracy and computational load.

[0102] The area boundary labeling and feature modeling include: Boundary definition: The boom area is a cylinder with a radius of 2 meters centered on the boom's central axis; the platform operation area is a cylinder with a radius of 3 meters centered on the tower's center, with a height range of ±5 meters. Feature parameters: Record the static features (such as tower coordinates and boom length) and dynamic features (such as the real-time location of the ground landing point, calculated from the tower crane's luffing parameters) of each area, and generate digital labels for the areas (ID, type, risk level).

[0103] The dynamic correction mechanism includes: Environmental change detection: By comparing the current point cloud with the map, newly added obstacles (such as those >0.5m) are detected. 3 Corrections are triggered when point cloud clusters or feature displacements (such as changes in ground landing point marker positions > 1m) occur. Local update strategy: Only the changed area (such as point cloud within a radius of 5 meters) is updated to avoid the time-consuming process of full map reconstruction. The update frequency is ≤ 200ms to ensure real-time performance.

[0104] In some embodiments, dividing the global spatial model into multi-level sensing areas includes: dividing the global spatial model into a core sensing layer, an early warning sensing layer, and a regular sensing layer according to monitoring accuracy and risk level requirements; the core sensing layer covers high-risk areas such as the hook path and platform operation area, sets the highest monitoring frequency and the minimum data acquisition interval, and captures target changes with millimeter-level accuracy in real time; the early warning sensing layer covers medium-risk areas such as the boom area and ground landing point, adopts a dynamic monitoring frequency, and automatically adjusts the data acquisition interval according to the tower crane's movement speed; the regular sensing layer covers low-risk areas such as behind the tower crane, sets a basic monitoring frequency, and periodically scans environmental features; the division boundaries and monitoring parameters of each level of sensing area are automatically adjusted according to the tower crane's operation mode to form an adaptive multi-level sensing system.

[0105] Based on risk level and monitoring accuracy requirements, the sensing areas are divided into three layers: core, early warning, and routine. The monitoring parameters of each layer are dynamically adjusted to form an adaptive system, enabling resource allocation and precise monitoring on demand.

[0106] The hierarchical classification principle includes: Core Sensing Layer (red): Covering the area below the hook (3-meter radius) and the platform work area (1.5 meters around the tower body), with a monitoring frequency of 20Hz, a data acquisition interval of 50ms, and an accuracy of ±3cm, used to detect personnel approach and component collision risks in real time. Early Warning Sensing Layer (yellow): Covering the entire length of the boom and the area around the ground landing point within 5 meters, with a monitoring frequency of 10Hz, a data acquisition interval of 100ms, and an accuracy of ±10cm, dynamically adjusted according to the tower crane's rotation speed (frequency increases to 15Hz when speed > 5° / s). Regular Sensing Layer (green): Covering the area behind the tower and open ground, with a monitoring frequency of 5Hz, a data acquisition interval of 200ms, and an accuracy of ±50cm, periodically scanning (every 2 seconds) for environmental changes.

[0107] The adaptive adjustment mechanism includes: Operation mode recognition: The operation mode is determined by the tower crane operating handle signal (e.g., "lifting mode" triggers high-precision monitoring of the core layer, "transfer mode" expands the warning layer range). Dynamic parameter configuration: A layer-level parameter database is established, and the monitoring frequency and acquisition interval are automatically adjusted according to the real-time risk level (feedback from the control center). For example, when personnel are detected entering the warning layer, the parameters for that area are temporarily increased to the core layer. Dynamic boundary correction: The boundaries of each layer are calculated in real time based on the global spatial model. For example, when the hook height changes, the vertical range of the core layer is adjusted synchronously (hook height ±5 meters); when the boom amplitude changes, the radius of the warning layer dynamically expands with the boom length (ratio 1:1).

[0108] In some embodiments, a reinforcement learning (RL) agent is constructed to address the task allocation problem of dual radar modules. Through real-time environmental state perception and reward feedback, the agent dynamically optimizes the monitoring area allocation strategy, breaking through the limitations of local optima in traditional greedy strategies and maximizing global resource efficiency.

[0109] The state space definition includes: Input states include: current tower crane operation parameters (load / height / angle), real-time dual-radar field-of-view coverage (overlap rate matrix of each area), historical task allocation efficiency (data acquisition latency / missed detection rate), and real-time risk level of the monitored area (dynamically assessed by the control center). State vector dimension: Normalization is used to concatenate continuous values ​​(e.g., load 0-100% mapped to 0-1) with discrete values ​​(one-hot encoding of the operation phase) into a 128-dimensional vector.

[0110] The action space design defines 12 atomic actions, including allocating high-priority areas to the first / second radar (6 types of areas × 2 radars), adjusting the scan interval (3 levels: 50ms / 100ms / 200ms), and triggering supplementary scan commands. The optimal allocation scheme is generated by combining these actions.

[0111] The reward function is constructed as follows: Positive rewards: high-priority area data integrity rate (reward +10 for every 1% increase), task allocation latency reduction (reward +5 for every 10ms reduction), and equipment energy saving (reward +8 for every 10% reduction in radar rotation speed). Negative penalties: high-risk area missed detection (penalty -50 for each instance), and data conflict rate exceeding the limit (penalty -20 for every 1% exceeding the limit).

[0112] Training and online applications include: Offline training: Simulating tens of millions of operational scenarios based on a digital twin tower crane environment, using the PPO (Proximity Policy Optimization) algorithm for training, and freezing the model when the cumulative reward value stably exceeds 800. Online inference: Acquiring the current state every 200ms, the agent outputs the optimal task allocation vector, controlling the radar module's rotation speed and scanning strategy in real time via API, and continuously optimizing the strategy through an experience replay buffer.

[0113] In some embodiments, to address the challenge of modeling the spatial correlation of multi-source data in complex environments, a graph neural network (GNN) is introduced to construct a relationship graph of the monitoring area. By aggregating node features and dynamically adjusting edge weights, deep feature fusion including spatial semantics is achieved, thereby improving the accuracy of target recognition and trajectory prediction.

[0114] The graph structure construction includes: Node definition: Each monitoring sub-region (such as hook path sub-region A, platform operation area sub-region B) is used as a graph node. Node features include sub-region point cloud density, number of dynamic targets, and frequency of historical risk events. Edge definition: Edge weight represents the spatial distance between sub-regions (distance < 5 meters is considered strong correlation, weight 0.8; 5-10 meters is considered weak correlation, weight 0.3; > 10 meters is considered no correlation), constructing an undirected weighted graph.

[0115] The GNN model design includes: adopting the GAT (Graph Attention Network) architecture, containing a two-layer attention mechanism: the first layer aggregates features of adjacent nodes (e.g., nodes in the boom region aggregate features of their upstream and downstream sub-regions), and the second layer generates a fused feature vector (64 dimensions). The input layer fuses point cloud coordinates (3D), distance data (1D), and timestamp encoding (converted to 16D through sinusoidal position encoding), and then inputs it into the graph network after linear transformation.

[0116] The fusion strategy includes: Dynamic weight generation: Automatically adjusting the data contribution of each sub-region through attention coefficients (e.g., the attention weight ratio of core region nodes ≥ 60%), replacing the traditional fixed weight strategy. Output application: Inputting the fused feature vector into the subsequent risk assessment model improves the target location prediction accuracy by 30% (compared to the traditional weighted average method), and reduces the motion trajectory prediction error to ±15cm.

[0117] In some embodiments, by designing a spatiotemporal Transformer model to process multi-region fused data sequences based on the spatiotemporal dynamic characteristics of the tower crane operating environment, long-distance spatiotemporal dependencies are captured, enabling forward-looking prediction of collision risks and overcoming the lag limitations of traditional threshold-based early warning.

[0118] Spatiotemporal sequence modeling includes: Input data: Organizing the multi-region fusion sensing results into a spatiotemporal tensor (batch × time step × number of regions × feature dimension) according to the time series (last 10 seconds, 10 frames per second), with features including target coordinates, velocity, acceleration, and risk level. Location encoding: Simultaneously encoding the timestamp (sine function) and spatial coordinates (relative position embedding based on the tower crane coordinate system) to generate a spatiotemporal joint location encoding vector (128 dimensions).

[0119] Transformer architecture optimizations include: Spatiotemporal attention mechanism: a dual-branch attention module is designed, with temporal axis attention capturing the target motion trend (such as acceleration changes in the hook trajectory), and spatial axis attention modeling inter-regional relationships (such as the impact of obstacles in the boom region on the hook path). Prediction head design: includes a risk level classification head (3 categories: safe / warning / dangerous) and a collision time regression head (outputting remaining safe time in seconds), trained using a cross-entropy + MSE joint loss function.

[0120] Online predictive applications include: Real-time inference: The model receives the latest 10 frames of data every 500ms and outputs a risk prediction result for the next 2 seconds. When the predicted collision time is less than 1.5 seconds, an emergency braking signal is triggered (traditional threshold methods can only detect the current distance and have no predictive capability). Incremental learning: The model is automatically synchronized with the latest fault data weekly to fine-tune it and continuously adapt to changes in complex operating environments.

[0121] In some embodiments, a federated learning (FL) framework is constructed to address the data privacy protection and collaborative modeling requirements of multi-tower crane operation scenarios. This framework enables collaborative optimization of monitoring models while ensuring that local data from each tower crane remains within its domain, thereby resolving the issues of insufficient data volume from a single tower crane and poor model generalization.

[0122] The system architecture design includes: Edge layer: Each tower crane deploys a local FL client, including a dual radar data preprocessing module, a local model (such as a target recognition CNN), and an encrypted parameter update module. Central server: Aggregates model parameters from each client (using the FedAvg algorithm), maintains the global model, and periodically distributes updated weights (e.g., weekly).

[0123] Privacy protection mechanisms include: Data preprocessing: Anonymizing the raw point cloud data locally (removing device IDs and geographic location identifiers), retaining only relative coordinates and environmental features. Encrypted communication: Homomorphic encryption (such as the Paillier algorithm) is used for parameter updates, and Laplacian noise (ε=0.5) is added during gradient upload to meet differential privacy protection.

[0124] The collaborative optimization process includes: each tower crane uses its own operational data (approximately 10GB / day) to train a target recognition model, and uploads the model increment (weight change) after 5 iterations. After integrating data from more than 100 tower cranes, the accuracy for rare scenarios (such as ground personnel identification in heavy rain) increased from 65% to 88%, and the generalization ability was significantly enhanced.

[0125] In some embodiments, by constructing a digital twin of the tower crane's operating environment and combining real-time monitoring data with physical model simulation, the virtual-real mapping and autonomous evolution of the full-domain spatial model can be realized, solving the problem of untimely updates in traditional SLAM mapping under dynamic environments.

[0126] The twin model construction includes: Geometric modeling: using Blender to build a 3D solid model of the tower crane (accuracy ±5mm), and an environmental model including standardized construction site scenes (roads / buildings / material storage areas), with 20 preset typical obstacle models. Physical modeling: defining the tower crane dynamics model (Newton-Euler equations describe the boom motion), radar signal propagation model (considering multipath effects and attenuation coefficients), and constructing a mapping relationship between virtual and real data.

[0127] The virtual-real fusion evolution includes: every 10 seconds, ICP registration is performed between the radar measured point cloud and the twin model predicted point cloud, and the environmental change offset is calculated (translation <10cm, rotation <2° is considered static, otherwise model update is triggered). When a change in the material stacking area is detected, the twin automatically simulates the possible state of the area in the next 30 minutes (based on Markov prediction of historical operation patterns), and updates the risk area boundary of the global model in advance. By simulating the effects of different task allocation strategies through the twin (such as the impact of a 15% increase in radar rotation speed on energy consumption and accuracy), the optimal parameter combination is output (such as recommending a rotation speed of 20rpm during peak hoisting periods and 10rpm during off-peak periods), which feeds back into the parameter configuration of the actual monitoring system, realizing a closed loop of "monitoring-simulation-optimization".

[0128] Please see Figure 3 As shown, Figure 3 This is a schematic diagram of the structure of a tower crane multi-area fusion sensing system 200 based on a rotating radar module, provided in an embodiment of this application. The tower crane multi-area fusion sensing system 200 based on a rotating radar module is used to execute the steps of the tower crane multi-area fusion sensing method based on a rotating radar module shown in the above embodiments. The tower crane multi-area fusion sensing system 200 based on a rotating radar module can be a single server or a server cluster, or it can be a terminal, such as a handheld terminal, a laptop computer, a wearable device, or a robot.

[0129] like Figure 3 As shown, the tower crane multi-area fusion sensing system 200 based on a rotating radar module includes:

[0130] The rotation control unit 201 is used to control the dual radar modules to rotate in coordination at a preset speed, so that the monitoring range of the corresponding first radar module and second radar module respectively covers different foundation areas of the tower crane. The foundation areas include at least two areas among the boom area, hook path, behind the tower, platform working area and ground landing point; the dual radar modules include a first radar module and a second radar module.

[0131] The task allocation unit 202 is used to dynamically allocate tasks to the dual radar modules according to the real-time rotation angle and monitoring range of the dual radar modules, so that the first radar module and the second radar module respectively undertake the real-time data acquisition tasks of different monitoring areas, wherein the monitoring area is any one or more combinations of the basic areas.

[0132] The data acquisition unit 203 is used to control the dual radar modules to perform real-time data acquisition on the corresponding areas according to the assigned monitoring tasks, and to acquire point cloud data and distance data of each monitoring area; to perform field-of-view fusion processing on the point cloud data and distance data of each monitoring area acquired by the dual radar modules, to remove redundant data of repeated monitoring areas, and to generate fused monitoring data;

[0133] The continuous monitoring unit 204 is used to perform spatial mapping based on the fused monitoring data, construct a global spatial model including the boom area, hook path, behind the tower, platform operation area and ground landing point, and divide the global spatial model into multi-level sensing areas to realize continuous monitoring of the tower crane operation environment in the global spatial model.

[0134] It should be noted that those skilled in the art will understand that, for the sake of convenience and brevity, the specific working process of the tower crane multi-area fusion perception system and its modules based on rotating radar modules described above can be referred to the corresponding content in the various embodiments of the tower crane multi-area fusion perception method based on rotating radar modules, and will not be repeated here.

[0135] The aforementioned multi-area fusion sensing method for tower cranes based on rotating radar modules can be implemented as a computer program, which can be used in various applications such as... Figure 3 It runs on the device shown.

[0136] Please see Figure 4 , Figure 4 This is a schematic block diagram of the controller provided in an embodiment of this application. The controller includes a processor, a memory, and a network interface connected via a device bus, wherein the memory may include a storage medium and internal memory.

[0137] The storage medium can store operating devices and computer programs. The computer program includes program instructions that, when executed, cause the processor to perform any multi-area fusion sensing method for tower cranes based on rotating radar modules.

[0138] The processor provides computing and control capabilities to support the operation of the entire controller.

[0139] The internal memory provides an environment for the execution of computer programs in non-volatile storage media. When the computer program is executed by the processor, it enables the processor to execute any multi-area fusion sensing method for tower cranes based on rotating radar modules.

[0140] This network interface is used for network communication, such as sending assigned tasks. Those skilled in the art will understand that... Figure 4The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the terminal to which the present application is applied. The specific controller may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0141] It should be understood that the processor can be a Central Processing Unit (CPU), but it can also be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Among these, a general-purpose processor can be a microprocessor or any conventional processor.

[0142] In one embodiment, the processor is configured to run a computer program stored in memory to perform the following steps:

[0143] The dual radar modules are controlled to rotate in tandem at a preset speed, so that the monitoring range of the corresponding first radar module and second radar module respectively covers different foundation areas of the tower crane. The foundation areas include at least two areas among the boom area, hook path, behind the tower, platform working area and ground landing point; the dual radar modules include a first radar module and a second radar module.

[0144] Based on the real-time rotation angle and monitoring range of the dual radar modules, dynamic task allocation is performed on the dual radar modules, so that the first radar module and the second radar module respectively undertake the real-time data acquisition task of different monitoring areas, wherein the monitoring area is any one or more combinations of the basic areas.

[0145] The dual radar modules are controlled to collect real-time data from the corresponding areas according to the assigned monitoring tasks, and to obtain point cloud data and distance data for each monitoring area; the point cloud data and distance data collected by the dual radar modules for each monitoring area are subjected to field-of-view fusion processing to remove redundant data from repeated monitoring areas and generate fused monitoring data;

[0146] Based on the fused monitoring data, spatial mapping is performed to construct a global spatial model that includes the boom area, hook path, behind the tower, platform work area, and ground landing point. Multi-level sensing areas are then divided within the global spatial model to achieve continuous monitoring of the tower crane's operating environment.

[0147] In some embodiments, the continuous monitoring of the tower crane's operating environment in the global spatial model includes: selecting a target sensing area corresponding to the current operating state of the tower crane in the multi-level sensing area according to the current operating state of the tower crane; generating a multi-area fusion sensing result based on real-time data of the target sensing area; inputting the multi-area fusion sensing result to the hoisting safety control center for risk warning, path judgment, and target identification; and continuously looping the operations of dual radar module coordinated rotation, dynamic task allocation, data acquisition, field-of-view fusion processing, spatial mapping, and generation of fusion sensing results to achieve continuous monitoring of the tower crane's operating environment.

[0148] In some embodiments, generating multi-region fusion sensing results based on real-time data of the target sensing area includes: acquiring point cloud data and distance data of each monitoring sub-region within the target sensing area; performing coordinate unification processing on the data of each sub-region based on a preset spatial coordinate transformation algorithm to establish a unified data format including three-dimensional spatial coordinates, monitoring timestamps, and data confidence scores; matching and deduplicating duplicate data of the same monitoring target in different sub-regions through preset dynamic data association rules, retaining the valid data with the highest confidence scores; and performing feature fusion on the processed valid data according to a preset fusion strategy to generate multi-region fusion sensing results including target location information, motion trajectory information, and environmental attribute information, wherein the fusion strategy dynamically adjusts the data fusion weights according to the current tower crane operation status.

[0149] In some embodiments, inputting the multi-area fusion perception results to the hoisting safety control center for risk warning, path determination, and target identification includes: encrypting and standardizing the generated multi-area fusion perception results, and transmitting them to the hoisting safety control center in real time through a preset communication interface; after receiving the data, the hoisting safety control center calculates the risk level of obstacle distances, movement trends, and tower crane component positions within the target perception area based on a preset risk assessment model, and generates corresponding risk warning signals; it calls a preset path planning algorithm to analyze real-time environmental data of the hook path area, determines whether there is a collision risk in the current hoisting path, and outputs path adjustment suggestions; and it uses a target identification model deployed in the hoisting safety control center to identify and locate personnel and equipment in the ground landing area and platform operation area, generating target location coordinates and category information.

[0150] In some embodiments, the continuous cyclic execution of the operations of coordinated rotation of the dual radar modules, dynamic task allocation, data acquisition, field-of-view fusion processing, spatial mapping, and generation of fused perception results to achieve continuous monitoring of the tower crane's operating environment includes: setting a time interval threshold or event triggering condition for cyclic execution; starting a new round of cyclic operation when the preset time interval is reached or a change in the tower crane's operating status is detected; before each cycle begins, acquiring the current rotation angle of the dual radar modules, the equipment operating status, and the tower crane's real-time operating parameters as initial data for dynamic task allocation; during the cycle, caching and version management of the processed data to ensure that the monitoring data of two adjacent cycles have temporal continuity and spatial correlation; if abnormal data or equipment fault signals are detected during the cycle, triggering an emergency handling process, temporarily freezing the current monitoring task and maintaining the output of the last valid data until the anomaly is resolved and cyclic execution resumes.

[0151] In some embodiments, the step of dynamically allocating tasks to the dual radar modules based on their real-time rotation angle and monitoring range, so that the first and second radar modules respectively undertake real-time data acquisition tasks for different monitoring areas, includes: establishing a preset monitoring area priority mapping table, wherein the priority mapping table dynamically sorts the monitoring priorities of the boom area and the base area of ​​the hook path according to the current lifting load, hook height, and operation stage of the tower crane; calculating in real time the effective monitoring range corresponding to the current rotation angle of the first and second radar modules, generating a three-dimensional coordinate matrix of the radar field of view coverage area; and, based on the monitoring area priority mapping table and the radar field of view coverage area, using a task scheduling algorithm to allocate high-priority monitoring areas to the radar modules with more complete current field of view coverage, and allocating spare monitoring cycles to low-priority areas, wherein the task scheduling algorithm is based on a greedy strategy to prioritize satisfying the real-time data acquisition frequency requirements of high-priority areas.

[0152] In some embodiments, the point cloud data and distance data collected by the dual radar modules in each monitoring area are subjected to field-of-view fusion processing to remove redundant data in overlapping monitoring areas and generate fused monitoring data. This includes: establishing a spatial coordinate transformation relationship between the first radar module and the second radar module based on the installation position parameters and real-time rotation angle of the dual radar modules, and unifying the monitoring data of the first radar module and the second radar module into the tower crane coordinate system; performing time synchronization processing on the point cloud data after unifying the coordinate system, and removing invalid data with time differences exceeding a preset threshold according to the data acquisition timestamp; for overlapping monitoring areas, by calculating the spatial distance deviation and signal strength value of the point cloud data, setting a distance fusion threshold and strength filtering rules, retaining the valid data points with the highest signal strength and distance deviation within the threshold, and generating a non-redundant fused monitoring dataset.

[0153] In some embodiments, the step of constructing a spatial model based on the fused monitoring data, including the boom area, hook path, tower back, platform work area, and ground landing point, includes: processing the point cloud information in the fused monitoring data using a simultaneous positioning and mapping algorithm to generate a three-dimensional point cloud map of the tower crane's operating environment; marking the spatial boundaries and feature parameters of basic areas such as the boom area and hook path in the three-dimensional point cloud map according to preset area division rules, and establishing digital labels for each area; dynamically correcting the three-dimensional point cloud map using real-time updated fused monitoring data, and triggering a local map update mechanism when environmental feature changes are detected to ensure the consistency between the global spatial model and the actual operating environment.

[0154] In some embodiments, dividing the global spatial model into multi-level sensing areas includes: dividing the global spatial model into a core sensing layer, an early warning sensing layer, and a regular sensing layer according to monitoring accuracy and risk level requirements; the core sensing layer covers high-risk areas such as the hook path and platform operation area, sets the highest monitoring frequency and the minimum data acquisition interval, and captures target changes with millimeter-level accuracy in real time; the early warning sensing layer covers medium-risk areas such as the boom area and ground landing point, adopts a dynamic monitoring frequency, and automatically adjusts the data acquisition interval according to the tower crane's movement speed; the regular sensing layer covers low-risk areas such as behind the tower crane, sets a basic monitoring frequency, and periodically scans environmental features; the division boundaries and monitoring parameters of each level of sensing area are automatically adjusted according to the tower crane's operation mode to form an adaptive multi-level sensing system.

[0155] This application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, causes the processor to implement the steps of the tower crane multi-area fusion sensing method based on a rotating radar module as provided in any embodiment of this application.

[0156] The computer-readable storage medium may be an internal storage unit of the controller as described in the foregoing embodiments, such as the hard disk or memory of the controller. Alternatively, the computer-readable storage medium may be an external storage device of the controller, such as a plug-in hard disk, Smart Media Card (SMC), Secure Digital (SD) card, or Flash Card.

[0157] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in this application, and these modifications or substitutions should all be covered within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A multi-area fusion sensing method for tower cranes based on a rotating radar module, characterized in that, include: The dual radar modules are controlled to rotate in tandem at a preset speed, so that the monitoring range of the first radar module and the second radar module respectively covers different foundation areas of the tower crane. The foundation areas include at least two areas among the boom area, hook path, behind the tower, platform working area, and ground landing point; the dual radar modules include the first radar module and the second radar module. Based on the real-time rotation angle and monitoring range of the dual radar modules, dynamic task allocation is performed on the dual radar modules, so that the first radar module and the second radar module respectively undertake the real-time data acquisition task of different monitoring areas, wherein the monitoring area is any one or more combinations of the basic areas. The dual radar modules are controlled to collect real-time data from the corresponding areas according to the assigned monitoring tasks, and to obtain point cloud data and distance data for each monitoring area; the point cloud data and distance data collected by the dual radar modules for each monitoring area are subjected to field-of-view fusion processing to remove redundant data from repeated monitoring areas and generate fused monitoring data; Based on the fused monitoring data, spatial mapping is performed to construct a global spatial model that includes the boom area, hook path, tower back, platform work area, and ground landing point. Multi-level sensing areas are then divided within the global spatial model to enable continuous monitoring of the tower crane's operating environment. The continuous monitoring of the tower crane's operating environment within the global spatial model includes: selecting a target sensing area corresponding to the current operating state of the tower crane within the multi-level sensing area based on the current operating state; generating a multi-area fusion sensing result based on real-time data from the target sensing area; inputting the multi-area fusion sensing result to the hoisting safety control center for risk warning, path determination, and target identification; and continuously looping the operations of dual radar module coordinated rotation, dynamic task allocation, data acquisition, field-of-view fusion processing, spatial mapping, and generation of fusion sensing results to achieve continuous monitoring of the tower crane's operating environment. The process of generating multi-region fusion sensing results based on real-time data from the target sensing area includes: acquiring point cloud data and distance data of each monitoring sub-region within the target sensing area; performing coordinate unification processing on the data of each sub-region based on a preset spatial coordinate transformation algorithm to establish a unified data format containing three-dimensional spatial coordinates, monitoring timestamps, and data confidence scores; matching and deduplicating duplicate data of the same monitoring target in different sub-regions through preset dynamic data association rules, retaining the valid data with the highest confidence score; and performing feature fusion on the processed valid data according to a preset fusion strategy to generate multi-region fusion sensing results containing target location information, motion trajectory information, and environmental attribute information, wherein the fusion strategy dynamically adjusts the data fusion weights according to the current tower crane operation status.

2. The method according to claim 1, characterized in that, The step of inputting the multi-region fusion perception results to the hoisting safety control center for risk warning, path determination, and target identification includes: The generated multi-area fusion perception results are encrypted and format-standardized, and then transmitted in real time to the hoisting safety control center through a preset communication interface. After receiving the data, the hoisting safety control center calculates the risk level of the obstacle distance, movement trend and tower crane component position in the target perception area based on a preset risk assessment model, and generates corresponding risk warning signals. The system calls a preset path planning algorithm to analyze real-time environmental data of the hook path area, determine whether there is a collision risk in the current hoisting path, and output path adjustment suggestions; through the target recognition model deployed in the hoisting safety control center, it identifies and locates personnel and equipment in the ground landing area and platform operation area, and generates target location coordinates and category information.

3. The method according to claim 1, characterized in that, The continuous cyclic execution of the dual radar module's coordinated rotation, dynamic task allocation, data acquisition, field-of-view fusion processing, spatial mapping, and generation of fused perception results enables continuous monitoring of the tower crane's operating environment, including: Set a time interval threshold or event trigger condition for the loop execution. When the preset time interval is reached or a change in the tower crane's operating status is detected, start a new round of loop operation. Before each cycle begins, the current rotation angle of the dual radar modules, the equipment operating status, and the real-time operating parameters of the tower crane are obtained as initial data for dynamic task allocation. During the loop, the processed data is cached and version managed to ensure that the monitoring data of two adjacent loops have temporal continuity and spatial correlation. If abnormal data or equipment failure signals are detected during the loop, the emergency handling procedure is triggered, the current monitoring task is temporarily frozen and the output of the last valid data is maintained until the abnormality is eliminated and the loop execution resumes.

4. The method according to claim 1, characterized in that, The step of dynamically allocating tasks to the dual radar modules based on their real-time rotation angle and monitoring range, so that the first radar module and the second radar module respectively undertake real-time data acquisition tasks for different monitoring areas, includes: Establish a preset monitoring area priority mapping table. The priority mapping table dynamically sorts the monitoring priorities of the boom area and the base area of ​​the hook path according to the current lifting load, hook height and operation stage of the tower crane. The effective monitoring range corresponding to the current rotation angle of the first and second radar modules is calculated in real time, and a three-dimensional coordinate matrix of the radar field of view coverage area is generated. Based on the monitoring area priority mapping table and the radar field of view coverage area, a task scheduling algorithm is used to allocate high-priority monitoring areas to radar modules with more complete current field of view coverage, and to allocate spare monitoring cycles to low-priority areas. The task scheduling algorithm is based on a greedy strategy to prioritize meeting the real-time data acquisition frequency requirements of high-priority areas.

5. The method according to claim 1, characterized in that, The point cloud data and distance data of each monitoring area collected by the dual radar modules are subjected to field-of-view fusion processing to remove redundant data from overlapping monitoring areas and generate fused monitoring data, including: Based on the installation position parameters and real-time rotation angle of the dual radar modules, a spatial coordinate transformation relationship between the first radar module and the second radar module is established, and the monitoring data of the first radar module and the second radar module are unified into the tower crane coordinate system. Time synchronization processing is performed on the point cloud data after unifying the coordinate system, and invalid data with time differences exceeding a preset threshold is removed based on the data acquisition timestamp; For areas with repeated monitoring, the spatial distance deviation and signal strength values ​​of point cloud data are calculated, and distance fusion thresholds and strength filtering rules are set. Valid data points with the highest signal strength and distance deviation within the threshold are retained to generate a non-redundant fused monitoring dataset.

6. The method according to claim 1, characterized in that, The spatial mapping based on the fused monitoring data constructs a global spatial model including the boom area, hook path, behind the tower, platform work area, and ground landing point, including: Simultaneous localization and mapping (SMR) algorithms are used to process point cloud information from fused monitoring data to generate a 3D point cloud map of the tower crane's operating environment. According to the preset regional division rules, the spatial boundaries and feature parameters of the basic regions corresponding to the boom region and hook path are marked in the three-dimensional point cloud map, and digital labels for each region are established. The 3D point cloud map is dynamically corrected by real-time updated fusion monitoring data. When changes in environmental features are detected, a local map update mechanism is triggered to ensure the consistency between the global spatial model and the actual working environment.

7. The method according to claim 1, characterized in that, The process of dividing the global spatial model into multi-level perception regions includes: Based on monitoring accuracy and risk level requirements, the overall spatial model is divided into a core sensing layer, an early warning sensing layer, and a regular sensing layer. The core sensing layer covers high-risk areas corresponding to the hook path and platform operation area, setting the highest monitoring frequency and minimum data acquisition interval to capture target changes with millimeter-level precision in real time. The early warning sensing layer covers medium-risk areas corresponding to the boom area and ground landing point, using a dynamic monitoring frequency that automatically adjusts the data acquisition interval according to the tower crane's movement speed. The regular sensing layer covers low-risk areas behind the tower crane, setting a basic monitoring frequency and periodically scanning environmental features. The boundaries and monitoring parameters of each sensing layer are automatically adjusted according to the tower crane's operation mode, forming an adaptive multi-level sensing system.

8. A tower crane multi-area fusion sensing system based on a rotating radar module, characterized in that, include: A rotation control unit is used to control the dual radar modules to rotate in tandem at a preset speed, so that the monitoring range of the first radar module and the second radar module respectively covers different foundation areas of the tower crane. The foundation areas include at least two areas among the boom area, hook path, behind the tower, platform working area, and ground landing point. The dual radar modules include a first radar module and a second radar module. The task allocation unit is used to dynamically allocate tasks to the dual radar modules according to the real-time rotation angle and monitoring range of the dual radar modules, so that the first radar module and the second radar module respectively undertake the real-time data acquisition tasks of different monitoring areas, wherein the monitoring area is any one or more combinations of the basic areas. The data acquisition unit is used to control the dual radar modules to collect data in real time in the corresponding areas according to the assigned monitoring tasks, and to acquire point cloud data and distance data of each monitoring area; the point cloud data and distance data of each monitoring area collected by the dual radar modules are processed by field-of-view fusion to remove redundant data of repeated monitoring areas and generate fused monitoring data. The continuous monitoring unit is used to perform spatial mapping based on the fused monitoring data, construct a global spatial model including the boom area, hook path, behind the tower, platform work area and ground landing point, and divide the global spatial model into multi-level sensing areas to realize continuous monitoring of the tower crane's working environment in the global spatial model. The continuous monitoring of the tower crane's operating environment within the global spatial model includes: selecting a target sensing area corresponding to the current operating state of the tower crane within the multi-level sensing area based on the current operating state; generating a multi-area fusion sensing result based on real-time data from the target sensing area; inputting the multi-area fusion sensing result to the hoisting safety control center for risk warning, path determination, and target identification; and continuously looping the operations of dual radar module coordinated rotation, dynamic task allocation, data acquisition, field-of-view fusion processing, spatial mapping, and generation of fusion sensing results to achieve continuous monitoring of the tower crane's operating environment. The process of generating multi-region fusion sensing results based on real-time data from the target sensing area includes: acquiring point cloud data and distance data of each monitoring sub-region within the target sensing area; performing coordinate unification processing on the data of each sub-region based on a preset spatial coordinate transformation algorithm to establish a unified data format containing three-dimensional spatial coordinates, monitoring timestamps, and data confidence scores; matching and deduplicating duplicate data of the same monitoring target in different sub-regions through preset dynamic data association rules, retaining the valid data with the highest confidence score; and performing feature fusion on the processed valid data according to a preset fusion strategy to generate multi-region fusion sensing results containing target location information, motion trajectory information, and environmental attribute information, wherein the fusion strategy dynamically adjusts the data fusion weights according to the current tower crane operation status.

Images