CN-121982230-A - Monocular vision TBM tunnel surrounding rock fracture network three-dimensional reconstruction system

CN121982230ACN 121982230 ACN121982230 ACN 121982230ACN-121982230-A

Abstract

The invention relates to a monocular vision TBM tunnel surrounding rock fracture network three-dimensional reconstruction system, which is characterized in that an image acquisition module is used for acquiring a surrounding rock surface two-dimensional image, an image preprocessing module is used for carrying out fracture recognition, binarization processing, hole self-adaptive filling and skeleton extraction on the image, a space analysis module is used for extracting depth information of each pixel point in the image by combining geometric assumption that a TBM tunnel section is a regular smooth cylinder and internal parameters and external parameters of an industrial camera array through coordinate conversion, a fracture dimension conversion module is used for carrying out intersection recognition and separation on the extracted skeleton pixel points and converting the surrounding rock surface two-dimensional image into a poisson disc representation in a three-dimensional space, and a fracture network reconstruction module is used for polymerizing the poisson disc into a complete fracture surface to generate a TBM tunnel surrounding rock fracture network three-dimensional model. The invention realizes the three-dimensional reconstruction of the tunnel surrounding rock fracture network with high precision, and is suitable for the dynamic operation environment of continuous tunneling of TBM.

Inventors

ZHANG CHONGCHONG
ZHENG KAIGE
FU WEI
HUANG YONG
GAO BIN
LIAO JUN

Assignees

中铁第一勘察设计院集团有限公司
中国铁建股份有限公司

Dates

Publication Date: 20260505
Application Date: 20251225

Claims (10)

1. Monocular vision TBM tunnel country rock crack network three-dimensional reconstruction system, its characterized in that: the system comprises an image acquisition module, an image preprocessing module, a space analysis module, a fracture dimension conversion module and a fracture network reconstruction module: The image acquisition module is used for continuously acquiring two-dimensional images of the surrounding rock surface through the industrial camera array; the image preprocessing module is used for carrying out crack identification, binarization processing, hole self-adaptive filling and skeleton extraction on the two-dimensional image of the surrounding rock surface; the space analysis module is used for extracting depth information of each pixel point in the two-dimensional image of the surrounding rock surface by coordinate transformation and combining geometric assumption that the TBM tunnel section is a regular smooth cylinder and internal parameters and external parameters of the industrial camera array; the fracture dimension conversion module is used for identifying and separating intersection points of the extracted skeleton pixel points and converting the two-dimensional image of the surrounding rock surface into a poisson disc representation in a three-dimensional space; And the fracture network reconstruction module is used for polymerizing the scattered poisson discs into a complete fracture surface according to a space communication relation to generate a complete TBM tunnel surrounding rock fracture network three-dimensional model.
2. The monocular vision TBM tunnel surrounding rock fracture network three-dimensional reconstruction system of claim 1, wherein: The image acquisition module comprises an industrial camera array and a synchronous trigger; the industrial camera array is deployed in an L1 safety zone behind the TBM shield, and automatically shoots two-dimensional images of the surrounding rock surface at fixed time intervals or distance intervals in the TBM tunneling process; The synchronous trigger is used for realizing time synchronous shooting of multiple cameras.
3. The monocular vision TBM tunnel surrounding rock fracture network three-dimensional reconstruction system of claim 2, wherein: The image preprocessing module comprises a crack identification unit, a binarization processing unit, a hole self-adaptive filling unit and a skeleton extraction unit; the crack identification unit is used for preprocessing the two-dimensional image of the surrounding rock surface, including graying, denoising and contrast enhancement; The binarization processing unit is used for converting the two-dimensional image of the surrounding rock surface into a binarization image by adopting a self-adaptive threshold segmentation method, wherein a crack pixel is marked as 1, and a background pixel is marked as 0; the hole self-adaptive filling unit is used for marking the connected domain of the binarized image, calculating the area of the connected domain, judging whether the connected domain is a hole or not, and modifying pixel values to fill the hole; the skeleton extraction unit is used for extracting the central line skeleton of the crack based on a Zhang-Suen refinement algorithm of distance transformation.
4. A monocular vision TBM tunnel surrounding rock fracture network three-dimensional reconstruction system as claimed in claim 3, wherein: Carrying out connected domain marking on the binarized image, judging whether the connected domain is a hole after calculating the area of the connected domain, and modifying pixel values to carry out hole filling, wherein the method comprises the following steps: Carrying out connected domain marking on the binarized image, and marking all connected domains in the binarized image as different labels; calculating the area of each connected domain, and setting an area threshold; judging whether the connected domain is a hole or not according to the connected domain with the area smaller than the area threshold value, namely checking whether the periphery of the connected domain is completely surrounded by slit pixels or not; and modifying all pixel values of the connected domain determined as the hole into 1 to realize hole filling.
5. The monocular vision TBM tunnel surrounding rock fracture network three-dimensional reconstruction system of claim 4, wherein: The space analysis module comprises a coordinate conversion unit, a camera calibration unit and a depth calculation unit; the coordinate conversion unit is used for converting pixel coordinates to image physical coordinates, converting the image physical coordinates to camera coordinates and converting the camera coordinates to world coordinates; the camera calibration unit is used for confirming and calibrating internal parameters and external parameters of the industrial camera array; the depth calculation unit is used for calculating the actual distance from each pixel point to the camera, namely the depth information of each pixel point by using a cylinder fitting method based on the geometric assumption that the TBM tunnel section is a regular smooth cylinder.
6. The monocular vision TBM tunnel surrounding rock fracture network three-dimensional reconstruction system of claim 5, wherein: Calculating the actual distance from each pixel point to the camera comprises the following steps: calculating the ray direction corresponding to the pixel according to the pixel coordinates; calculating the intersection point of the ray and the cylindrical surface, wherein the distance from the intersection point to the camera is the depth; For the crack area which is not on the cylindrical surface, a bilinear interpolation method is adopted, and the depth of the pixel is calculated according to the depth value interpolation of surrounding pixels.
7. The monocular vision TBM tunnel surrounding rock fracture network three-dimensional reconstruction system of claim 6, wherein: The fracture dimension conversion module comprises an intersection point identification unit, a fracture separation unit, a plane fitting unit and a disc parameter calculation unit; The intersection point identification unit is used for carrying out convolution operation by adopting a fracture intersection point identification operator to identify and eliminate fracture intersection points; The crack separation unit is used for processing the identified crack intersection points, and completely modifying the pixel values in the crack intersection point pixels and 8 adjacent areas thereof to 0 so as to realize the separation of the intersecting cracks; the plane fitting unit is used for fitting a plane where the crack is located by adopting an SVD singular value decomposition method, and taking a right singular vector corresponding to the minimum singular value as a unit normal vector of the plane; The disc parameter calculation unit is used for completing parameter calculation of poisson disc representation, taking a point set center point as a disc center, calculating the distance between the head end and the tail end of the crack as the disc diameter, and taking a unit normal vector as a disc normal vector.
8. The monocular vision TBM tunnel surrounding rock fracture network three-dimensional reconstruction system of claim 7, wherein: fitting a plane where the crack is located by adopting an SVD singular value decomposition method, taking a right singular vector corresponding to a minimum singular value as a unit normal vector of the plane, and comprising the following steps: Skeleton pixel points of each independent crack section form a three-dimensional point set; Calculating a three-dimensional point set center point; Constructing an offset vector matrix of each point in the point set and the central point; SVD singular value decomposition is carried out on the matrix; the right singular vector corresponding to the minimum singular value is the unit normal vector of the fitting plane.
9. The monocular vision TBM tunnel surrounding rock fracture network three-dimensional reconstruction system of claim 8, wherein: The fracture network reconstruction module comprises a coupling clustering unit, a plane optimizing unit and a model generating unit; the coupling clustering unit is used for dividing the disks with similar orientations and similar positions into the same group, and each group represents a complete fracture surface; the plane optimizing unit is used for maximizing the projection area of the inner discs of the group to the target plane by adopting a gradient descent method, so as to determine the optimal parameters of the complete fracture surface; the model generating unit is used for exporting parameters of all complete fracture surfaces into a general three-dimensional format and generating a visual model.
10. The monocular vision TBM tunnel surrounding rock fracture network three-dimensional reconstruction system of claim 9, wherein: grouping closely oriented and closely located disks into the same group includes: based on the orientation clustering of cosine similarity, calculating cosine similarity between unit normal vectors of any two discs and grouping by adopting a DBSCAN clustering algorithm; For each obtained group, calculating Euclidean distance between circle centers of inner discs in the group and further subdividing the Euclidean distance; After two-step clustering, the disks with similar orientations and similar positions are separated into the same group.

Description

Monocular vision TBM tunnel surrounding rock fracture network three-dimensional reconstruction system Technical Field The invention relates to the technical field of tunnel engineering detection, in particular to a monocular vision TBM tunnel surrounding rock fracture network three-dimensional reconstruction system. Background As a highly mechanized and automatic tunnel construction device integrating mechanical, electronic and hydraulic functions, the full-face tunnel boring machine (Tunnel Boring Machine, TBM) has the outstanding advantages of high construction speed, small environmental influence, high automation degree and the like, and has become the primary choice for long tunnel construction. However, TBM construction faces an important technical challenge, namely real-time perception and adaptive regulation of surrounding rock structure. During TBM tunnel excavation, the stress balance of surrounding rock is destroyed, which can lead to further expansion and fracture of structural faces. The accurate acquisition of three-dimensional parameters of the rock mass structural plane is important for evaluating the stability of surrounding rock of a tunnel and guiding the design of a construction scheme. The tunnel surrounding rock cracks are various cracks, fracture surfaces or weak surfaces existing in the rock mass around the tunnel, and are mutually intersected and connected to form a complex crack network. The acquisition of fracture network information is a core premise for realizing safe, efficient and intelligent tunneling of TBM. Specifically, tunneling performance can be accurately predicted and optimized through tunnel surrounding rock fracture network information, and core parameters such as cutter head rotating speed, thrust, torque and the like can be dynamically adjusted according to stratum fracture characteristics, so that optimal tunneling efficiency is realized while equipment is protected. Meanwhile, based on tunnel surrounding rock fracture network information, important construction risks can be effectively prevented and controlled, surrounding rock can be quickly loosened and collapsed after excavation in a fracture network with fault fracture zones or extremely broken fracture zones, TBM shields are wrapped or blocked, machines are trapped, the fracture network is mastered in advance, and reinforcing measures can be adopted or tunneling modes can be adjusted in advance. However, the problem that the surrounding rock fracture sensing system is arranged in the TBM tunnel is mainly that the first is insufficient in space information extraction, the existing method is mostly limited to a two-dimensional image layer, space depth information of a fracture cannot be obtained, an effective technology is lacked to convert a two-dimensional pixel identification result into a three-dimensional structure surface shape containing real depth information, the identification result is difficult to directly use for the shape analysis and the connection relation judgment of the three-dimensional structure surface, the second is poor in instantaneity, the data processing flow is long, and urgent requirements for real-time and rapid analysis of surrounding rock states in a TBM dynamic construction scene are difficult to meet. Therefore, there is a need to propose new measures to overcome the above drawbacks. Disclosure of Invention The invention aims to provide a monocular vision TBM tunnel surrounding rock fracture network three-dimensional reconstruction system, which aims to solve the problems that the prior art is limited to a two-dimensional image layer, has poor instantaneity and the like. In order to achieve the above purpose, the technical scheme adopted by the invention is as follows: the system comprises an image acquisition module, an image preprocessing module, a space analysis module, a fracture dimension conversion module and a fracture network reconstruction module: The image acquisition module is used for continuously acquiring two-dimensional images of the surrounding rock surface through the industrial camera array; the image preprocessing module is used for carrying out crack identification, binarization processing, hole self-adaptive filling and skeleton extraction on the two-dimensional image of the surrounding rock surface; the space analysis module is used for extracting depth information of each pixel point in the two-dimensional image of the surrounding rock surface by coordinate transformation and combining geometric assumption that the TBM tunnel section is a regular smooth cylinder and internal parameters and external parameters of the industrial camera array; the fracture dimension conversion module is used for identifying and separating intersection points of the extracted skeleton pixel points and converting the two-dimensional image of the surrounding rock surface into a poisson disc representation in a three-dimensional space; And the fracture network reconstruction mod