Wearable three-dimensional laser scanning system underground passage rapid mapping method

Abstract

The invention discloses a wearable three-dimensional laser scanning system underground passage rapid mapping method, which comprises the following steps of: A, constructing wearable equipment, and integrating a scanning module, an inertial navigation module, a panoramic image module, a UWB (Ultra Wideband) positioning module and an edge calculation unit on a backpack bracket; and B, dynamic data adjustment: dynamically adjusting the scanning frequency according to the advancing speed through a self-adaptive point cloud density adjustment technology. The scanning module, the inertial navigation module, the panoramic image module, the UWB positioning module and the edge calculation unit are integrated on the backpack support, the portability of the equipment is improved, and the equipment can be easily worn and freely move in an underground passage for data acquisition, so that the flexibility and efficiency of mapping are improved, and the practicability of the equipment is improved. Through cooperation of the scanning module, the inertial navigation module and the panoramic image module, high-precision positioning and attitude information is provided for data acquisition, and the accuracy of subsequent mapping is ensured.

Description

Technical Field

[0001] The present invention relates to the technical field of three-dimensional laser scanning equipment for underground passages, and specifically provides a method for rapidly mapping underground passages using a wearable three-dimensional laser scanning system. Background Art

[0002] Three-dimensional laser scanning equipment for underground passages is a measurement tool that can quickly and accurately obtain the three-dimensional spatial information of underground passages. It creates a three-dimensional model of the passage by emitting laser beams and detecting the reflected light, and is widely used in fields such as building surveying and mapping, railway tunnel detection, and mine roadway measurement.

[0003] In the traditional process of mapping underground passages, total stations or GPS-RTK technologies are mainly used for data collection and mapping. However, these methods have some significant limitations, such as relatively low work efficiency, high dependence on environmental conditions, and high labor costs. Although backpack-type or vehicle-mounted three-dimensional laser scanning systems have emerged on the market, and these systems have indeed contributed to improving data collection efficiency, in practical applications, especially in narrow and complex multi-branch underground passages, they still face technical bottlenecks such as positioning drift and point cloud stitching errors. These problems limit the wide application and accuracy of three-dimensional laser scanning systems in underground passage surveying and mapping.

[0004] Therefore, a method for rapidly mapping underground passages using a wearable three-dimensional laser scanning system is provided. Summary of the Invention

[0005] The purpose of the present invention is to provide a method for rapidly mapping underground passages using a wearable three-dimensional laser scanning system, which has the advantages of providing measurement accuracy and efficiency, and solves the problems of low efficiency and poor accuracy of traditional methods.

[0006] To achieve the above purpose, the present invention provides the following technical solution: A method for rapidly mapping underground passages using a wearable three-dimensional laser scanning system, and the mapping method includes:

[0007] A. Construction of the wearable device, which integrates a scanning module, an inertial navigation module, a panoramic imaging module, a UWB positioning module, and an edge computing unit on a backpack-type bracket;

[0008] B. Dynamic data adjustment, by means of an adaptive point cloud density adjustment technology, dynamically adjusts the scanning frequency according to the traveling speed;

[0009] C. Multi-source data fusion, uses an improved LOAM algorithm for rough point cloud registration, introduces human kinematic constraints to reduce stitching errors, and combines Kalman filtering to fuse IMU and foot pressure sensor data to eliminate vibration interference;

[0010] D. Real-time modeling and output. The embedded GPU processor performs point cloud lightweight compression and BIM parametric mapping to generate a LOD3-level 3D model and output files in IFC or Revit format. By integrating the point cloud topology and the BIM parametric component library, a true-color 3D model is automatically generated.

[0011] Preferably, in step A, the scanning module uses a line array lidar, whose scanning rate is adjustable from 1 million to 2 million points per second, and the ranging accuracy is ±2 cm / 10 m.

[0012] Preferably, in step A, the inertial navigation module uses a six-axis IMU, whose attitude accuracy is ±0.01° and it is networked with the UWB positioning module.

[0013] Preferably, in step A, the edge computing unit uses an NVIDIA Jetson AGX Xavier embedded GPU, which supports CUDA accelerated computing.

[0014] Preferably, in step B, the dynamic data acquisition includes a blind area compensation mechanism, which automatically repairs the scanning blind area through the overlap rate of adjacent site clouds, and the overlap rate ≥ 30%. In step B, a semantic segmentation network based on the PointNet++ architecture is used to identify segment joints and cable bracket facilities, and the leakage area identification accuracy ≥ 95%.

[0015] Preferably, the improved LOAM algorithm in step C is based on the joint angle compensation formula of the human kinematics model:

[0016] 1) Algorithm optimization item: Introduce human kinematics constraints to correct the point cloud distortion equation:

[0017]

[0018] where ωIMU is the inertial navigation angular velocity and θк is the joint angle;

[0019] 2) Key parameters:

[0020] The number of scan matching iterations: 50 times / frame, the pose optimization convergence threshold: translation 0.01 m, rotation 0.1°, the map update frequency is 10 Hz, and the embedded GPU is used for acceleration.

[0021] Preferably, the dynamic somatosensory compensation algorithm is as follows:

[0022] 1) Multi-sensor fusion:

[0023] State equation:

[0024] x к =F кχκ-1 +B κUκ +ω k

[0025] Observation equation:

[0026] z κ = H κχκ-1 + u k

[0027] where χ = [θ, ω, α]T is the state vector, and F κ is the human kinematics matrix;

[0028] 2) Implementation process:

[0029] a) IMU data preprocessing: Low-pass filtering, cut-off frequency 10 Hz;

[0030] b) Foot pressure sensor calibration: Nonlinear correction, polynomial fitting;

[0031] c) Kalman gain update frequency: 100 Hz;

[0032] d) Vibration suppression effect: High-frequency vibration (>5 Hz) attenuation ≥ 90%.

[0033] Preferably, the semantic segmentation model of PointNet++:

[0034] 1) Network architecture:

[0035] Input: Original point cloud (n×3 coordinates) + intensity value;

[0036] Feature extraction layer: 4-level MSG module;

[0037] Output: 12 types of underground facility labels, segment joints, cable brackets, leakage areas;

[0038] 2) Training strategy:

[0039] Dataset: 500 groups of underground passage point clouds annotated in COCO format;

[0040] Transfer learning: Fine-tuning on the ShapeNet pre-trained model;

[0041] Optimizer: Adam, learning rate 1e-4, decay rate 0.95;

[0042] Recognition accuracy: mAP@0.5 = 95.2%.

[0043] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0044] 1. Through this method, the present invention integrates a scanning module, an inertial navigation module, a panoramic imaging module, a UWB positioning module, and an edge computing unit on a backpack-type bracket, improving the portability of the device. It can be easily worn and move freely in underground passages for data collection, getting rid of the limitations of traditional large-scale surveying equipment, enhancing the flexibility and efficiency of mapping, and enabling quick entry into different underground passage scenarios to carry out work. The cooperation of the scanning module, inertial navigation module, and panoramic imaging module provides high-precision positioning and attitude information for data collection, ensuring the accuracy of subsequent mapping.

[0045] 2. Through this method, the present invention, by means of an adaptive point cloud density adjustment technology, dynamically adjusts the scanning frequency according to the traveling speed, enabling the system to better adapt to different working scenarios. When the traveling speed is relatively fast, the scanning frequency is appropriately increased to ensure data integrity. When the traveling speed is relatively slow, the scanning frequency is decreased to reduce data redundancy and improve data processing efficiency. By executing point cloud lightweight compression and BIM parametric mapping through an embedded GPU processor, it can quickly generate a LOD3-level 3D model and output files in IFC or Revit format. The real-time modeling and output function enables staff to promptly obtain the 3D model of the underground passage. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 It is a schematic diagram of the system flow of the present invention;

[0047] Figure 2 It is a schematic diagram of the point cloud data processing flow of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0048] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0049] The present invention provides a technical solution, a method for rapid mapping of underground passages by a wearable three-dimensional laser scanning system. The mapping method includes:

[0050] A. Construction of a wearable device, which integrates a scanning module, an inertial navigation module, a panoramic imaging module, a UWB positioning module, and an edge computing unit on a backpack-type bracket;

[0051] B. Dynamic data adjustment, by means of an adaptive point cloud density adjustment technology, dynamically adjusts the scanning frequency according to the traveling speed;

[0052] C. Multi-source data fusion, using an improved LOAM algorithm for rough point cloud registration, introducing human kinematic constraints to reduce stitching errors, and combining Kalman filtering to fuse IMU and foot pressure sensor data to eliminate vibration interference;

[0053] D. Real-time modeling and output, performing lightweight compression of point clouds and BIM parametric mapping through an embedded GPU processor, generating a LOD3-level 3D model and outputting IFC or Revit format files, and automatically generating a true-color 3D model by fusing the point cloud topological structure and the BIM parametric component library.

[0054] Embodiment 2:

[0055] In Embodiment 1, add the following process:

[0056] In step A, the scanning module uses a matrix lidar, whose scanning rate is adjustable from 1 million to 2 million points per second, and the ranging accuracy is ±2 cm / 10 m.

[0057] Its mapping method includes the following steps:

[0058] A. Construction of a wearable device, integrating a scanning module, an inertial navigation module, a panoramic imaging module, a UWB positioning module, and an edge computing unit on a backpack support;

[0059] B. Dynamic data adjustment, dynamically adjusting the scanning frequency according to the traveling speed through an adaptive point cloud density adjustment technology;

[0060] C. Multi-source data fusion, using an improved LOAM algorithm for rough point cloud registration, introducing human kinematic constraints to reduce stitching errors, and combining Kalman filtering to fuse IMU and foot pressure sensor data to eliminate vibration interference;

[0061] D. Real-time modeling and output, performing lightweight compression of point clouds and BIM parametric mapping through an embedded GPU processor, generating a LOD3-level 3D model and outputting IFC or Revit format files, and automatically generating a true-color 3D model by fusing the point cloud topological structure and the BIM parametric component library.

[0062] Embodiment 3:

[0063] In Embodiment 2, add the following process:

[0064] In step A, the inertial navigation module uses a six-axis IMU, whose attitude accuracy is ±0.01° and it is networked with the UWB positioning module.

[0065] Its mapping method includes the following steps:

[0066] A. Construction of a wearable device, which integrates a scanning module, an inertial navigation module, a panoramic imaging module, a UWB positioning module, and an edge computing unit on a backpack-type bracket;

[0067] B. Dynamic data adjustment, through the adaptive point cloud density adjustment technology, dynamically adjusts the scanning frequency according to the traveling speed;

[0068] C. Multi-source data fusion, uses an improved LOAM algorithm for rough point cloud registration, introduces human kinematic constraints to reduce stitching errors, and combines Kalman filtering to fuse IMU and foot pressure sensor data to eliminate vibration interference;

[0069] D. Real-time modeling and output, through an embedded GPU processor, performs lightweight compression of the point cloud and BIM parametric mapping, generates a LOD3-level three-dimensional model and outputs an IFC or Revit format file, and automatically generates a true-color three-dimensional model by fusing the point cloud topological structure and the BIM parametric component library.

[0070] Example 4:

[0071] In Example 3, add the following process:

[0072] In step A, the edge computing unit uses an NVIDIA Jetson AGX Xavier embedded GPU, which supports CUDA accelerated computing.

[0073] Its mapping method includes the following steps:

[0074] A. Construction of a wearable device, which integrates a scanning module, an inertial navigation module, a panoramic imaging module, a UWB positioning module, and an edge computing unit on a backpack-type bracket;

[0075] B. Dynamic data adjustment, through the adaptive point cloud density adjustment technology, dynamically adjusts the scanning frequency according to the traveling speed;

[0076] C. Multi-source data fusion, uses an improved LOAM algorithm for rough point cloud registration, introduces human kinematic constraints to reduce stitching errors, and combines Kalman filtering to fuse IMU and foot pressure sensor data to eliminate vibration interference;

[0077] D. Real-time modeling and output, through an embedded GPU processor, performs lightweight compression of the point cloud and BIM parametric mapping, generates a LOD3-level three-dimensional model and outputs an IFC or Revit format file, and automatically generates a true-color three-dimensional model by fusing the point cloud topological structure and the BIM parametric component library.

[0078] Example 5:

[0079] In Example 4, add the following process:

[0080] The dynamic data acquisition in step B includes a blind area compensation mechanism, which automatically repairs the scanning blind area through the cloud overlap rate of adjacent stations, and the overlap rate ≥ 30%. In step B, a semantic segmentation network based on the PointNet++ architecture is used to identify the segment joints and cable support facilities, and the recognition accuracy of the leakage area is ≥ 95%.

[0081] Its mapping method includes the following steps:

[0082] A. Construction of a wearable device, which integrates a scanning module, an inertial navigation module, a panoramic image module, a UWB positioning module, and an edge computing unit on a backpack bracket;

[0083] B. Dynamic data adjustment, through the adaptive point cloud density adjustment technology, dynamically adjusts the scanning frequency according to the traveling speed;

[0084] C. Multi-source data fusion, using an improved LOAM algorithm for rough point cloud registration, introducing human kinematic constraints to reduce stitching errors, and combining Kalman filtering to fuse IMU and foot pressure sensor data to eliminate vibration interference;

[0085] D. Real-time modeling and output, through an embedded GPU processor to perform point cloud lightweight compression and BIM parametric mapping, generate a LOD3-level three-dimensional model and output IFC or Revit format files, and automatically generate a true-color three-dimensional model by fusing the point cloud topological structure and the BIM parametric component library.

[0086] Example 6:

[0087] In Example 5, add the following process:

[0088] The improved LOAM algorithm in step C is based on the joint angle compensation formula of the human kinematic model:

[0089] 1) Algorithm optimization items, introducing human kinematic constraints to correct the point cloud distortion equation:

[0090]

[0091] where ωIMU is the inertial navigation angular velocity and θκ is the joint angle;

[0092] 2) Key parameters:

[0093] The number of scanning matching iterations: 50 times / frame, the pose optimization convergence threshold: translation 0.01m, rotation 0.1°, the map update frequency is 10Hz, and embedded GPU acceleration is used.

[0094] Its mapping method includes the following steps:

[0095] A. Construction of a wearable device, integrating a scanning module, an inertial navigation module, a panoramic imaging module, a UWB positioning module, and an edge computing unit on a backpack-type bracket;

[0096] B. Dynamic data adjustment, dynamically adjusting the scanning frequency according to the traveling speed through an adaptive point cloud density adjustment technology;

[0097] C. Multi-source data fusion, using an improved LOAM algorithm for rough point cloud registration, introducing human kinematic constraints to reduce stitching errors, and combining Kalman filtering to fuse IMU and foot pressure sensor data to eliminate vibration interference;

[0098] D. Real-time modeling and output, performing point cloud lightweight compression and BIM parametric mapping through an embedded GPU processor, generating a 3D model at LOD3 level and outputting files in IFC or Revit format, and automatically generating a true-color 3D model by fusing the point cloud topology and the BIM parametric component library.

[0099] Example Seven:

[0100] In Example Six, add the following process:

[0101] The dynamic somatosensory compensation algorithm is as follows:

[0102] 1) Multi-sensor fusion:

[0103] State equation:

[0104] x κ = F κχκ-1 + B κUκ + ω k

[0105] Observation equation:

[0106] z κ = H κχκ-1 + u k

[0107] where χ = [θ, ω, α]T is the state vector, and F κ is the human kinematic matrix;

[0108] 2) Implementation process:

[0109] a) IMU data preprocessing: low-pass filtering, cut-off frequency 10Hz;

[0110] b) Foot pressure sensor calibration: non-linear correction, polynomial fitting;

[0111] c) Kalman gain update frequency: 100Hz;

[0112] d) Vibration suppression effect: attenuation of high-frequency vibration (>5Hz) ≥90%.

[0113] Its mapping method includes the following steps:

[0114] A. Construction of a wearable device, which integrates a scanning module, an inertial navigation module, a panoramic imaging module, a UWB positioning module, and an edge computing unit on a backpack bracket;

[0115] B. Dynamic data adjustment, dynamically adjusting the scanning frequency according to the traveling speed through an adaptive point cloud density adjustment technology;

[0116] C. Multi-source data fusion, using an improved LOAM algorithm for rough point cloud registration, introducing human kinematic constraints to reduce stitching errors, and combining Kalman filtering to fuse IMU and foot pressure sensor data to eliminate vibration interference;

[0117] D. Real-time modeling and output, performing point cloud lightweight compression and BIM parametric mapping through an embedded GPU processor, generating a LOD3-level three-dimensional model and outputting an IFC or Revit format file, and automatically generating a true-color three-dimensional model by fusing the point cloud topological structure and the BIM parametric component library.

[0118] Example VIII:

[0119] In Example VII, add the following process:

[0120] Semantic segmentation model of PointNet++:

[0121] 1) Network architecture:

[0122] Input: Original point cloud (n×3 coordinates) + intensity value;

[0123] Feature extraction layer: 4-level MSG module;

[0124] Output: 12 types of underground facility labels, segment joints, cable brackets, leakage areas;

[0125] 2) Training strategy:

[0126] Dataset: 500 groups of underground passage point clouds annotated in COCO format;

[0127] Transfer learning: Fine-tuning on a ShapeNet pre-trained model;

[0128] Optimizer: Adam, learning rate 1e-4, decay rate 0.95;

[0129] Recognition accuracy: mAP@0.5 = 95.2%.

[0130] Its mapping method includes the following steps:

[0131] A. Construction of a wearable device, which integrates a scanning module, an inertial navigation module, a panoramic imaging module, a UWB positioning module, and an edge computing unit on a backpack-type bracket;

[0132] B. Dynamic data adjustment, through an adaptive point cloud density adjustment technology, dynamically adjusts the scanning frequency according to the traveling speed;

[0133] C. Multi-source data fusion, uses an improved LOAM algorithm for rough point cloud registration, introduces human kinematic constraints to reduce stitching errors, and combines Kalman filtering to fuse IMU and foot pressure sensor data to eliminate vibration interference;

[0134] D. Real-time modeling and output, through an embedded GPU processor, performs point cloud lightweight compression and BIM parametric mapping, generates a LOD3-level three-dimensional model and outputs an IFC or Revit format file, and automatically generates a true-color three-dimensional model by fusing the point cloud topological structure and the BIM parametric component library.

[0135] It should be noted that in this article, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements not only includes those elements, but also includes other elements not expressly listed, or also includes elements inherent to such process, method, article or device. Without further limitation, an element defined by the statement "including a..." does not exclude the existence of additional identical elements in the process, method, article or device including the said element.

[0136] Although the embodiments of the present invention have been shown and described, for those of ordinary skill in the art, it can be understood that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the present invention, and the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for rapid mapping of underground passages using a wearable three-dimensional laser scanning system, characterized in that: The mapping method includes: A. Construction of a wearable device, which integrates a scanning module, an inertial navigation module, a panoramic imaging module, a UWB positioning module, and an edge computing unit on a backpack bracket; B. Dynamic data adjustment, dynamically adjusting the scanning frequency according to the traveling speed through an adaptive point cloud density adjustment technology; C. Multi-source data fusion, using an improved LOAM algorithm for rough point cloud registration, introducing human kinematic constraints to reduce stitching errors, and combining Kalman filtering to fuse IMU and foot pressure sensor data to eliminate vibration interference; D. Real-time modeling and output, performing point cloud lightweight compression and BIM parametric mapping through an embedded GPU processor, generating a LOD3-level three-dimensional model and outputting files in IFC or Revit format, and automatically generating a true-color three-dimensional model by fusing the point cloud topological structure and the BIM parametric component library.

2. A method for rapid mapping of underground passages using the wearable three-dimensional laser scanning system according to claim 1, characterized in that: In step A, the scanning module uses a planar array lidar, whose scanning rate is adjustable from 1 million to 2 million points per second, and the ranging accuracy is ±2 cm / 10 m.

3. A method for rapid mapping of underground passages using a wearable three-dimensional laser scanning system according to claim 1, characterized in that: In step A, the inertial navigation module uses a six-axis IMU, whose attitude accuracy is ±0.01° and it forms a network with the UWB positioning module.

4. A method for rapid mapping of underground passages using a wearable three-dimensional laser scanning system according to claim 1, characterized in that: In step A, the edge computing unit uses an NVIDIA Jetson AGX Xavier embedded GPU, which supports CUDA accelerated operation.

5. A method for rapid mapping of underground passages using the wearable three-dimensional laser scanning system according to claim 1, characterized in that: In step B, the dynamic data acquisition includes a blind area compensation mechanism, automatically repairing the scanning blind area through the overlap rate of adjacent site point clouds, and the overlap rate ≥ 30%. In step B, a semantic segmentation network based on the PointNet++ architecture is used to identify segment joints and cable bracket facilities, and the leakage area identification accuracy ≥ 95%.

6. A method for rapid mapping of underground channels using the wearable three-dimensional laser scanning system according to claim 1, characterized in that: In step C, the improved LOAM algorithm is based on the joint angle compensation formula of the human kinematic model: 1) Algorithm optimization item, introducing human kinematic constraints and correcting the point cloud distortion equation: where ωIMU is the inertial navigation angular velocity, and θ κ is the joint angle; 2) Key parameters: Scanning matching iteration times: 50 times / frame, pose optimization convergence threshold: translation 0.01 m, rotation 0.1°, the map update frequency is 10 Hz, and an embedded GPU is used for acceleration.

7. A method for rapid mapping of underground channels using the wearable three-dimensional laser scanning system according to claim 1, characterized in that: The dynamic somatosensory compensation algorithm is as follows: 1) Multi-sensor fusion: State equation: x κ = F κχκ-1 + B κUκ + ω k Observation equation: z κ = H кχκ-1 + u k where χ = [θ, ω, α]T is the state vector, and F к is the human kinematics matrix; 2) Implementation process: a) IMU data preprocessing: low-pass filtering, cut-off frequency 10 Hz; b) Calibration of the foot pressure sensor: non-linear correction, polynomial fitting; c) Kalman gain update frequency: 100 Hz; d) Vibration suppression effect: high-frequency vibration (>5 Hz) attenuation ≥ 90%.

8. A method for rapid mapping of underground passages using the wearable three-dimensional laser scanning system according to claim 1, characterized in that: The semantic segmentation model of PointNet++: 1) Network architecture: Input: original point cloud (n×3 coordinates) + intensity value; Feature extraction layer: 4-level MSG module; Output: 12 types of underground facility labels, segment joints, cable brackets, leakage areas; 2) Training strategy: Dataset: 500 groups of underground passage point clouds annotated in COCO format; Transfer learning: fine-tuning on the ShapeNet pre-trained model; Optimizer: Adam learning rate 1e-4, decay rate 0.95; Recognition accuracy: mAP@0.5 = 95.2%.

Images