Intelligent detection system and method for road surface closure facility

Abstract

The intelligent detection system comprises a field detection device and a cloud analysis system, the field detection device carries a laser radar and an optical camera through a horizontally rotatable holder, the laser radar and the optical camera are coaxial and consistent in view field, and point cloud data and image data are synchronously collected; the control module is responsible for instruction receiving, equipment driving and data caching and uploading; the cloud analysis system realizes optical camera and laser radar coordinate system conversion, road closure element identification, point cloud optimization clustering, spatial positioning and multi-dimensional early warning through five modules of joint calibration, target detection, point cloud processing, road closure facility element extraction and monitoring early warning. The method does not need to depend on a GPS, adapts to multiple scenes such as fixed road closure, dynamic road closure and inspection, solves the problems of low efficiency, insufficient precision of a single sensor and poor scene adaptation of traditional monitoring, and improves the safety, management efficiency and data traceability of road closure operation through a closed loop of real-time early warning, dynamic tracking and data storage.

Description

Technical Field

[0001] This invention relates to the field of municipal maintenance technology, and in particular to an intelligent detection system and method for road closure facilities. Background Technology

[0002] With the expansion of municipal maintenance operations and the upgrading of traffic safety requirements, real-time monitoring of on-site road closure facilities (including cones, signs, and construction vehicles) has become crucial to ensuring construction safety and smooth traffic flow. Traditional manual monitoring methods are inefficient, susceptible to environmental interference, and difficult to achieve full-time, full-coverage monitoring. While single-laser monitoring can acquire three-dimensional location information, its accuracy in identifying facility types is insufficient; single-camera monitoring lacks spatial coordinate data and cannot meet the need to determine the compliance of facility locations during road closure operations.

[0003] Furthermore, some existing solutions rely on GPS positioning or single monitoring devices, resulting in complex deployments and data silos, failing to achieve unified identification and behavior detection of maintenance equipment. Additionally, existing systems are mostly designed for fixed-location road closures, exhibiting poor adaptability to scenarios such as inspection operations and dynamic road closures, and lacking a complete closed loop of "real-time early warning + data storage," making it difficult to meet the industry's core needs for safety risk prevention and control, improved management efficiency, and accountability. Summary of the Invention

[0004] This invention provides an intelligent detection system and method for road closure facilities to solve the above-mentioned technical problems.

[0005] To address the aforementioned technical problems, this invention provides an intelligent detection system for road closure facilities, comprising on-site detection equipment and a cloud-based analysis system. The on-site testing equipment includes a horizontally rotatable gimbal, a lidar, an optical camera, and a control module. The lidar is mounted on the gimbal, and the optical camera is fixed above the lidar. The lidar and the optical camera are coaxially arranged and have the same field of view, respectively collecting point cloud data and image data. The control module is used to receive instructions from the cloud analysis system, drive the gimbal to rotate, control the lidar and optical camera to work synchronously, and cache and upload data to the cloud analysis system. The cloud-based analysis system includes a joint calibration module, a target detection module, a point cloud processing module, a road closure facility element extraction module, and a monitoring and early warning module. The joint calibration module is used to establish the transformation relationship between the camera coordinate system and the radar coordinate system, and to generate and store calibration files; The target detection module uses a target detection model to identify road closure elements in the image data and outputs bounding box coordinates, target category, and confidence level. The point cloud processing module performs rotation transformation, stitching and optimization processing on multiple sets of point cloud data according to a preset rotation angle. The road closure facility element extraction module converts the image bounding box into three-dimensional spatial parameters in the lidar coordinate system, generates a three-dimensional bounding box, clusters the point cloud within the three-dimensional bounding box, and calculates the centroid coordinates of the cluster as the spatial positioning coordinates of the road closure facility. The monitoring and early warning module is used to preset the facility placement spacing threshold, compare real-time coordinate data to determine violations and issue alarms; track the movement trajectory of engineering vehicles, workers and facilities, define safe areas with cone coordinates, and trigger alarms and push early warning information when personnel approach or vehicles enter the scene.

[0006] Preferably, the field testing equipment further includes a power supply module, which includes a removable battery pack and a power management chip for supplying power to the various components of the field testing equipment.

[0007] Preferably, the field testing equipment also includes an external protection module, which encapsulates the pan-tilt unit, lidar, optical camera, and control module into a single unit for physical protection in outdoor working environments.

[0008] Preferably, the outer protection module is mounted on a tripod via bolts, and the tripod is used to adjust the height and level of the field testing equipment.

[0009] Preferably, the joint calibration module adopts the Zhang Zhengyou calibration method, which solves the camera intrinsic parameters, distortion vector and extrinsic parameter matrices of the lidar and optical camera by collecting corresponding points in image data and point cloud data, and establishes the transformation relationship between the camera coordinate system and the radar coordinate system.

[0010] Preferably, the target detection module filters effective road closure elements based on preset key prompts for road closure scenarios, and the key prompts include at least traffic cones, signs, engineering vehicles, and workers.

[0011] Preferably, the point cloud processing module uses a voxel downsampling algorithm to optimize and filter the stitched point cloud data, and the size of the voxel grid is set to 100mm×100mm×100mm to remove point cloud noise and redundant data.

[0012] Preferably, the road closure facility element extraction module uses the Euclidean clustering algorithm to cluster the point cloud within the three-dimensional bounding box, and sets the clustering threshold to 800mm.

[0013] The present invention also provides an intelligent detection method for road closure facilities, applied to the intelligent detection system for road closure facilities as described above, comprising the following steps: Step 1: Deploy the on-site detection equipment and adjust its installation posture so that the field of view of the lidar and the optical camera covers the target road closure area; Step 2: Complete the joint calibration of the optical camera and lidar through the joint calibration module, and generate and store the calibration file; Step 3: The control module receives the scanning command from the cloud analysis system, drives the gimbal to rotate, and synchronously controls the lidar and optical camera to collect point cloud data and image data, caches and uploads them to the cloud analysis system; Step 4: The target detection module identifies road closure elements in the image data and outputs bounding box coordinates, target category, and confidence level; Step 5: The point cloud processing module performs rotation transformation, stitching, and optimization filtering on multiple sets of point cloud data; Step 6: The road closure facility element extraction module converts the image bounding box into a three-dimensional bounding box, clusters the point cloud data within the box, calculates the centroid coordinates, and obtains the spatial positioning coordinates of the road closure facility. Step 7: The monitoring and early warning module compares the real-time positioning coordinates with the preset threshold, tracks the dynamic trajectory, and triggers an alarm and pushes early warning information for violations, personnel approaching the safety boundary, or vehicles entering the area.

[0014] Preferably, in step 3, the scanning strategy used for data acquisition is as follows: Step 3.1: The initial angle of the gimbal is 0°, the lidar continuously scans for 20 seconds, synchronously collecting point cloud data, and the optical camera captures one image; Step 3.2: The gimbal rotates sequentially to 60°, 120°, and 180°, repeating the above 20-second scan and one image capture operation each time it stops; Step 3.3: The gimbal is reset to 0° and rotated to 180° in 30° increments at 5-second intervals. The optical camera takes one image each time it rotates. Step 3.4: Return the gimbal to 0° to complete one round of data acquisition.

[0015] Compared with existing technologies, the intelligent detection system and method for road closure facilities provided by this invention have the following advantages: 1. This invention acquires data synchronously and coaxially with a lidar and an optical camera, and achieves multi-source data fusion. It retains the three-dimensional spatial positioning advantage of point clouds and enhances the identification of road closure facility types by leveraging image texture features, thus solving the pain points of "inaccurate positioning" or "incomplete identification" of a single sensor. 2. This invention eliminates the dependence on GPS signals and achieves spatial positioning through laser-vision fusion technology. It can still work stably in obstructed scenarios such as tunnels and under elevated roads. At the same time, it is compatible with various municipal maintenance operation scenarios such as fixed road closures, dynamic road closures, and inspections. It is flexible in deployment and does not require complicated debugging. 3. This invention constructs a complete closed loop of "data collection - intelligent analysis - real-time early warning - dynamic tracking - data storage". It can provide real-time early warning of risks such as illegal placement of facilities, personnel approaching safety boundaries, and vehicles breaking in through preset thresholds. It can also track the movement trajectory of facilities, vehicles and personnel, providing complete data support for accident handling and accountability. Attached Figure Description

[0016] Figure 1 This is an architecture diagram of an intelligent detection system for road closure facilities according to a specific embodiment of the present invention; Figure 2 This is a schematic diagram of the installation of the field testing equipment in a specific embodiment of the present invention; Figure 3 This is a flowchart of an intelligent detection method for road closure facilities according to a specific embodiment of the present invention; Figure 4a This is the image detection result of a cone in a specific embodiment of the present invention; Figure 4b This is the point cloud detection result of a cone in a specific embodiment of the present invention.

[0017] In the diagram: 100-On-site testing equipment, 110-Pan-Tilt unit, 111-Motor, 120-LiDAR, 130-Optical camera, 140-Control module, 150-Power supply module, 160-External protection module, 170-Tripod, 200-Cloud analysis system, 210-Joint calibration module, 220-Target detection module, 230-Point cloud processing module, 240-Road closure facility element extraction module, 250-Monitoring and early warning module. Detailed Implementation

[0018] To illustrate the technical solutions of the invention in more detail, specific embodiments are listed below to demonstrate the technical effects; it should be emphasized that these embodiments are used to illustrate the invention and not to limit the scope of the invention.

[0019] The intelligent detection system for road closure facilities provided by this invention, such as Figure 1 and Figure 2 As shown, it includes on-site testing equipment 100 and a cloud analysis system 200, wherein: The on-site testing equipment 100 includes a horizontally rotatable gimbal 110 (rotation angle range 0-360°), a lidar 120, an optical camera 130, and a control module 140. The lidar 120 is mounted on the gimbal 110, and the optical camera 130 is fixed above the lidar 120. The lidar 120 and the optical camera 130 are coaxially arranged and have the same field of view, respectively acquiring point cloud data and image data. This coaxial and consistent installation method ensures the temporal and spatial matching of point cloud data and image data, providing an accurate data foundation for subsequent multi-source data fusion and avoiding detection errors caused by data misalignment.

[0020] The control module 140 is used to receive instructions (rotation angle and rotation time interval of gimbal 110) from the cloud analysis system 200, drive the gimbal 110 to rotate via motor 111, control the lidar 120 and optical camera 130 to work synchronously, read the timestamp information of lidar 120 and optical camera 130, and cache and upload data to the cloud analysis system 200, so as to realize the automated operation of the on-site detection equipment 100 without the need for real-time manual operation. At the same time, the local caching function can cope with outdoor communication interruption scenarios to ensure that data is not lost.

[0021] The cloud-based analysis system 200 includes a joint calibration module 210, a target detection module 220, a point cloud processing module 230, a road closure facility element extraction module 240, and a monitoring and early warning module 250.

[0022] The joint calibration module 210 is used to establish the transformation relationship between the camera coordinate system and the radar coordinate system, and to generate and store calibration files.

[0023] The target detection module 220 uses a target detection model to identify road closure elements in the image data and outputs bounding box coordinates, target category, and confidence level.

[0024] The point cloud processing module 230 performs rotation transformation, splicing and optimization processing on multiple sets of point cloud data according to a preset rotation angle.

[0025] The road closure facility element extraction module 240 converts the image bounding box into three-dimensional spatial parameters in the lidar coordinate system, generates a three-dimensional bounding box, clusters the point cloud within the three-dimensional bounding box, and calculates the centroid coordinates of the cluster as the spatial positioning coordinates of the road closure facility.

[0026] The monitoring and early warning module 250 is used to preset the facility placement spacing threshold, compare real-time coordinate data to determine violations and issue an alarm; track the movement trajectory of engineering vehicles, workers and facilities, define the safe area with cone coordinates, and trigger an alarm and push early warning information when personnel approach or vehicles enter the scene.

[0027] This invention achieves multi-source data fusion by coaxially and synchronously acquiring data through a lidar 120 and an optical camera 130. This retains the three-dimensional spatial positioning advantage of point clouds while enhancing the identification of road closure facility types by leveraging image texture features. Furthermore, this invention constructs a complete closed loop of "data acquisition - intelligent analysis - real-time early warning - dynamic tracking - data storage". It can provide real-time early warnings of risks such as illegal placement of facilities, personnel approaching safety boundaries, and vehicles entering through preset thresholds, and can also track the movement trajectories of facilities, vehicles, and personnel, providing complete data support for accident handling and accountability.

[0028] In some embodiments, please refer to the following: Figure 2 The field testing equipment 100 also includes a power supply module 150, which includes a detachable battery pack and a power management chip, for supplying power to the various components of the field testing equipment 100, taking into account both battery life and fast charging requirements.

[0029] In some embodiments, please refer to Figure 2 The field testing equipment 100 also includes an external protection module 160, which encapsulates the pan-tilt unit 110, lidar 120, optical camera 130, and control module 140 into one unit for physical protection in outdoor working environments, to resist interference from outdoor dust, rain, minor collisions, etc., to ensure stable operation of the equipment in harsh environments, and to reduce equipment failure rate and maintenance costs.

[0030] In some embodiments, please refer to Figure 2 The external protection module 160 is bolted to the tripod 170, which is used to adjust the height and level of the field detection equipment 100 to ensure that the field of view covers the target road closure area.

[0031] In some embodiments, the joint calibration module 210 adopts the Zhang Zhengyou calibration method, which solves the camera intrinsic parameters, distortion vector and extrinsic parameter matrices of the lidar 120 and the optical camera 130 by collecting corresponding points in the image data and point cloud data, and establishes the transformation relationship between the camera coordinate system and the radar coordinate system.

[0032] Specifically, the camera's parameters mainly include the intrinsic parameter matrix. Sum and distortion vectors ,in: , , The physical focal length of the camera. and The physical size of a pixel in the xy direction. Represents the pixel skew coefficient. The coordinates of the image center pixel are given. A transformation relationship between the camera coordinate system and the radar coordinate system is established using corresponding points in the image and point cloud (pixels and points representing the same target in both). The process of projecting the point cloud into the image can be achieved by... Expression. Among them Represents the camera intrinsic parameter matrix. T Represents the camera extrinsic parameter matrix. P To represent a point on a point cloud, This represents the depth information of a point in the camera coordinate system. 、v Represents the x and y coordinates in the pixel coordinate system. xyz This represents the three-dimensional coordinates of a point in the radar coordinate system. By selecting n sets of matching points (n>6), the extrinsic parameter matrix can be calculated using the least squares method and singular value decomposition algorithm. Finally, a calibration file is generated and stored.

[0033] In some embodiments, the target detection module 220 receives image data, uses an open-source target detection world model as its core algorithm, and sets key prompts for road closure scenarios. These key prompts include at least cones, signs, construction vehicles, and workers. The model can automatically identify all elements in the image and output the bounding box coordinates of each element. The system selects and outputs valid targets based on given keywords, including target category and confidence level. It accurately locks onto the monitoring target through keywords and avoids irrelevant targets interfering with the detection results.

[0034] In some embodiments, the point cloud processing module 230 uses a voxel downsampling algorithm to optimize and filter the stitched point cloud data. The size of the voxel grid is set to 100mm×100mm×100mm to remove point cloud noise and redundant data.

[0035] Specifically, the point cloud processing module 230 receives multiple sets of point cloud data, and for each set of point cloud... According to the pre-set rotation angle The point cloud is rotated, and the rotation matrix of the point cloud points is: The rotation of points in a point cloud is calculated using a rotation matrix: After completing each set of point cloud data After the rotation transformation, all the rotated point clouds By splicing the data together, a complete point cloud is formed. A voxel downsampling algorithm is used to optimize and filter the stitched point cloud, removing redundant points caused by stitching, and a voxel mesh is defined. To ensure processing accuracy, the extraction grid size is set to [size to be specified] based on the size of the cone. The unit is millimeters. Voxel downsampling significantly reduces the amount of point cloud data processed while removing noise.

[0036] In some embodiments, the road closure facility element extraction module 240 uses the Euclidean clustering algorithm to cluster the point cloud within the three-dimensional bounding box, and sets the clustering threshold to 800mm.

[0037] Specifically, the road closure facility element extraction module 240 receives image and point cloud data within the same time period, and combines this data with the rotation angle of the corresponding lidar 120. By calibrating the intrinsic and extrinsic parameters and using the stitched rotation matrix, the target bounding box output by the target detection module 220 is converted into three-dimensional spatial parameters in the lidar coordinate system. Based on the converted three-dimensional spatial parameters, the stitched complete point cloud is... D In this process, a corresponding 3D bounding box is generated for each target element. The bounding box is extended by 50mm in the x and y directions based on the transformed bounding box to cover the target edge point cloud; the z-direction is set to 0-3000mm (adapting to the height range of common elements such as workers, cones, signs, and construction vehicles), forming a closed 3D screening area. For the point cloud data within each 3D bounding box, Euclidean clustering is used for clustering, with a clustering threshold of 800mm. Point clouds with close proximity are grouped together, ensuring that each cluster corresponds to the complete point cloud data of a single target element, and eliminating scattered noise points within the 3D bounding box. For each valid cluster, its centroid coordinates are calculated. The formula is , , (in M (This represents the number of point clouds within the cluster). The centroid coordinates are the spatial positioning coordinates of the corresponding road closure facility elements, thus completing the extraction of road closure facility element coordinates.

[0038] Finally, the monitoring and early warning module 250 presets the placement spacing thresholds for various facilities (such as cone spacing of 5-8m) according to the road closure operation specifications, compares the real-time extracted coordinate data, and issues real-time alarms for violations; through data matching of each detection interval, it tracks the movement trajectory of engineering vehicles, the activity trajectory of workers, and the movement of facilities; and, using the detected cone coordinates as the operation boundary of the road closure area, it sets a safe area for road closure operations. When it detects workers approaching or moving vehicles entering the boundary, it immediately triggers an equipment alarm and pushes early warning information to the on-site management personnel terminal. The early warning information may include the target location, type, and real-time image.

[0039] This invention also provides an intelligent detection method for road closure facilities, applied to the intelligent detection system for road closure facilities described above, such as... Figure 3 As shown, it includes the following steps: Step 1: Deploy the on-site detection equipment 100 and adjust its installation posture so that the field of view of the lidar 120 and the optical camera 130 covers the target road closure area. Targeted posture adjustment ensures that there are no blind spots in the detection and avoids missed detection due to blind spots in the field of view, thus ensuring the comprehensiveness of the road closure area monitoring.

[0040] Step 2: The joint calibration module 210 completes the joint calibration of the optical camera 130 and the lidar 120, generating and storing the calibration file. Specifically, a checkerboard calibration board can be placed in a real scene to ensure full coverage of the device's field of view. The device is started to collect checkerboard data at different distances (3-5m) and angles (0°, 45°, 90°, 135°, 180°), with a cumulative collection of no less than 20 sets. After the data is uploaded to the cloud analysis system 200, the joint calibration module 210 is invoked to automatically solve for and save the camera's intrinsic and lidar extrinsic parameter matrices. This step only needs to be performed once, provided the device hardware structure remains unchanged.

[0041] Step 3: The control module 140 receives the scanning command from the cloud analysis system 200, drives the motor 111 to control the rotation of the gimbal 110, and simultaneously controls the lidar 120 and optical camera 130 to collect point cloud data and image data, caches and uploads them to the cloud analysis system 200. Specifically, the device can be placed near the center of the road closure area, and the height of the tripod 170 can be adjusted so that the device's field of view covers the entire road closure operation area, ensuring that the initial field of view of the lidar 120 and optical camera 130 is parallel to the road. After the device is started, the control module 140 receives the scanning command issued by the cloud analysis system 200 and automatically enters the rotation scanning mode.

[0042] Step 4: The target detection module 220 identifies the road closure elements in the image data and outputs the bounding box coordinates, target category, and confidence level (e.g., ...). Figure 4a The image detection results for the cone are shown below.

[0043] Step 5: The point cloud processing module 230 performs rotation transformation, stitching, and optimization filtering on multiple sets of point cloud data.

[0044] Step 6: The road closure facility element extraction module 240 converts the image bounding box into a 3D bounding box, clusters the point cloud data within the box, and calculates the centroid coordinates to obtain the spatial positioning coordinates of the road closure facility (e.g., ...). Figure 4b The image shows the point cloud detection results for the cone.

[0045] Step 7: The monitoring and early warning module 250 compares the real-time positioning coordinates with the preset threshold, tracks the dynamic trajectory, and triggers an alarm and pushes early warning information for illegal scenes, personnel approaching the safety boundary, or vehicles entering the scene.

[0046] Preferably, in step 3, the scanning strategy used for data acquisition is as follows: Step 3.1: The initial angle of the gimbal 110 is 0°, the lidar 120 continuously scans for 20 seconds, and simultaneously collects point cloud data (including spatial coordinates x_i, y_i, z_i and reflectivity reflect_i), and the optical camera 130 captures a 1440×1080 resolution color image. Step 3.2: The gimbal 110 rotates sequentially to 60°, 120°, and 180°, repeating the above 20s scanning and 1 image capture operation each time it stops; Step 3.3: The gimbal 110 is reset to 0° and rotated to 180° at 30° increments every 5 seconds. The optical camera 130 takes one image each time it rotates. Step 3.4: The PTZ 110 returns to 0°, completing one round of data acquisition. During the acquisition process, point cloud data is stored in CSV format and image data in JPG format, both named with "timestamp + device number". The data is uploaded to the cloud analysis system 200 in real time via the communication module. If communication is interrupted, local caching is activated, and automatic re-upload is performed upon resumption.

[0047] After each round of data collection and processing, the equipment remains stationary for 30 seconds before repeating the collection and analysis process to continuously monitor the entire road closure operation. All detection data, trajectory records, and early warning information are stored in the cloud in real time, forming complete data evidence to support safety traceability and accident handling.

[0048] In summary, the intelligent detection system and method for road closure facilities provided by the present invention includes on-site detection equipment 100 and cloud analysis system 200. By fusing multi-source data from lidar 120 and optical camera 130, and combining the five functional modules of cloud analysis system 200, the system achieves accurate detection and multi-dimensional early warning of road closure facilities.

[0049] Obviously, those skilled in the art can make various modifications and variations to the invention without departing from the spirit and scope of the invention. Therefore, if these modifications and variations fall within the scope of the claims of the invention and their equivalents, the invention is also intended to include these modifications and variations.

Claims

1. An intelligent detection system for road closure facilities, characterized in that, Including on-site testing equipment and cloud-based analysis systems, The on-site testing equipment includes a horizontally rotatable gimbal, a lidar, an optical camera, and a control module. The lidar is mounted on the gimbal, and the optical camera is fixed above the lidar. The lidar and the optical camera are coaxially arranged and have the same field of view, respectively collecting point cloud data and image data. The control module is used to receive instructions from the cloud analysis system, drive the gimbal to rotate, control the lidar and optical camera to work synchronously, and cache and upload data to the cloud analysis system. The cloud-based analysis system includes a joint calibration module, a target detection module, a point cloud processing module, a road closure facility element extraction module, and a monitoring and early warning module. The joint calibration module is used to establish the transformation relationship between the camera coordinate system and the radar coordinate system, and to generate and store calibration files; The target detection module uses a target detection model to identify road closure elements in the image data and outputs bounding box coordinates, target category, and confidence level. The point cloud processing module performs rotation transformation, stitching and optimization processing on multiple sets of point cloud data according to a preset rotation angle. The road closure facility element extraction module converts the image bounding box into three-dimensional spatial parameters in the lidar coordinate system, generates a three-dimensional bounding box, clusters the point cloud within the three-dimensional bounding box, and calculates the centroid coordinates of the cluster as the spatial positioning coordinates of the road closure facility. The monitoring and early warning module is used to preset the facility placement spacing threshold, compare real-time coordinate data to determine violations and issue alarms; track the movement trajectory of engineering vehicles, workers and facilities, define safe areas with cone coordinates, and trigger alarms and push early warning information when personnel approach or vehicles enter the scene.

2. The intelligent detection system for road closure facilities as described in claim 1, characterized in that, The field testing equipment also includes a power supply module, which includes a removable battery pack and a power management chip, for supplying power to the various components of the field testing equipment.

3. The intelligent detection system for road closure facilities as described in claim 1, characterized in that, The on-site testing equipment also includes an external protection module, which encapsulates the pan-tilt unit, lidar, optical camera, and control module into a single unit for physical protection in outdoor working environments.

4. The intelligent detection system for road closure facilities as described in claim 3, characterized in that, The external protection module is mounted on a tripod via bolts, and the tripod is used to adjust the height and level of the on-site testing equipment.

5. The intelligent detection system for road closure facilities as described in claim 1, characterized in that, The joint calibration module adopts the Zhang Zhengyou calibration method, which solves the camera intrinsic parameters, distortion vector and extrinsic parameter matrices of lidar and optical camera by collecting corresponding points in image data and point cloud data, and establishes the transformation relationship between the camera coordinate system and the radar coordinate system.

6. The intelligent detection system for road closure facilities as described in claim 1, characterized in that, The target detection module filters effective road closure elements based on preset key prompts for road closure scenarios. The key prompts include at least traffic cones, signs, engineering vehicles, and workers.

7. The intelligent detection system for road closure facilities as described in claim 1, characterized in that, The point cloud processing module uses a voxel downsampling algorithm to optimize and filter the stitched point cloud data. The size of the voxel grid is set to 100mm×100mm×100mm to remove point cloud noise and redundant data.

8. The intelligent detection system for road closure facilities as described in claim 1, characterized in that, The road closure facility element extraction module uses the Euclidean clustering algorithm to cluster the point cloud within the three-dimensional bounding box, setting the clustering threshold to 800mm.

9. A method for intelligent detection of road closure facilities, applied to the intelligent detection system for road closure facilities as described in any one of claims 1 to 8, characterized in that, Includes the following steps: Step 1: Deploy the on-site detection equipment and adjust its installation posture so that the field of view of the lidar and the optical camera covers the target road closure area; Step 2: Complete the joint calibration of the optical camera and lidar through the joint calibration module, and generate and store the calibration file; Step 3: The control module receives the scanning command from the cloud analysis system, drives the gimbal to rotate, and synchronously controls the lidar and optical camera to collect point cloud data and image data, caches and uploads them to the cloud analysis system; Step 4: The target detection module identifies road closure elements in the image data and outputs bounding box coordinates, target category, and confidence level; Step 5: The point cloud processing module performs rotation transformation, stitching, and optimization filtering on multiple sets of point cloud data; Step 6: The road closure facility element extraction module converts the image bounding box into a three-dimensional bounding box, clusters the point cloud data within the box, calculates the centroid coordinates, and obtains the spatial positioning coordinates of the road closure facility. Step 7: The monitoring and early warning module compares the real-time positioning coordinates with the preset threshold, tracks the dynamic trajectory, and triggers an alarm and pushes early warning information for violations, personnel approaching the safety boundary, or vehicles entering the area.

10. The intelligent detection method for road closure facilities as described in claim 9, characterized in that, In step 3, the scanning strategy used for data acquisition is as follows: Step 3.1: The initial angle of the gimbal is 0°, the lidar continuously scans for 20 seconds, synchronously collecting point cloud data, and the optical camera captures one image; Step 3.2: The gimbal rotates sequentially to 60°, 120°, and 180°, repeating the above 20-second scan and one image capture operation each time it stops; Step 3.3: The gimbal is reset to 0° and rotated to 180° in 30° increments at 5-second intervals. The optical camera takes one image each time it rotates. Step 3.4: Return the gimbal to 0° to complete one round of data acquisition.

Images