CN-121981886-A - Method, system, equipment and medium for splicing panoramic images of inner wall of tunnel

CN121981886ACN 121981886 ACN121981886 ACN 121981886ACN-121981886-A

Abstract

The invention relates to the technical field of image processing and provides a method, a system, equipment and a medium for splicing panoramic images of the inner wall of a tunnel, wherein the method comprises the following steps of S1, acquiring point cloud data shot by a laser radar, multi-view images shot by a camera, and calibration data of the laser radar and the camera; the method comprises the steps of S2, geometrically reconstructing the inner wall of a tunnel according to point cloud data to obtain a fitting cylinder, S3, generating continuous multi-section transverse strip plane unfolding images according to the fitting cylinder, the multi-view images and calibration data, wherein the lane width direction is transverse, and S4, longitudinally splicing the continuous multi-section transverse strip plane unfolding images to obtain a panoramic image of the inner wall of the tunnel, wherein the driving direction is longitudinal. The scheme can improve the accuracy and the high efficiency of the panoramic image detection of the inner wall of the tunnel.

Inventors

ZHOU LANGMING
CAO DONG
YIN LUBING
YAN ZHENYU
CHEN PAN
CHEN LEI

Assignees

湖南科天健光电技术有限公司

Dates

Publication Date: 20260505
Application Date: 20260124

Claims (10)

1. The tunnel inner wall panoramic image stitching method is characterized by comprising the following steps of: s1, acquiring point cloud data shot by a laser radar, multi-view images shot by a camera, and calibration data of the laser radar and the camera; s2, performing geometric reconstruction on the inner wall of the tunnel according to the point cloud data to obtain a fitting cylinder; s3, generating continuous multi-section transverse stripe plane unfolding images according to the fitting cylinder, the multi-view images and the calibration data, wherein the width direction of the lane is transverse; and S4, longitudinally splicing the continuous multi-section transverse strip plane unfolding images to obtain a panoramic image of the inner wall of the tunnel, wherein the running direction is longitudinal.
2. The method for stitching panoramic images of an inner wall of a tunnel according to claim 1, wherein in step S2, performing geometric reconstruction on the inner wall of the tunnel according to the point cloud data to obtain a fitted cylinder comprises: converting the point cloud data into a triangular grid model of the closed surface by adopting delaunay grid construction algorithm; fitting the triangular mesh model by using a cylinder fitting algorithm to obtain a fitting cylinder.
3. The method for stitching panoramic images of an inner wall of a tunnel according to claim 1, wherein in step S3, the generating continuous multi-segment transverse strip plane expanded images from the fitted cylinder, the multi-view images and the calibration data comprises: S31, carrying out distortion correction on each single-view image of multiple views; S32, dividing single-view image surface elements; S33, projecting each single-view image bin with multiple views to a space coordinate system where the fitting cylinder is located; S34, axially expanding the projected cylindrical surface of the cylinder; s35, homogenizing and cutting the unfolded image to obtain a section of transverse stripe plane unfolded image; And S36, repeating the steps S31 to S35 to obtain a multi-section transverse strip plane development diagram.
4. A method of stitching panoramic images of an inner wall of a tunnel according to claim 3, wherein in step S33, projecting the single view image bin into a spatial coordinate system in which the fitting cylinder is located comprises: s331, converting calibration data into a space coordinate system where the fitting cylinder is positioned to obtain a space point corresponding to the bin point; s332, performing straight line-cylindrical surface intersection measurement and visibility verification to obtain a bin coordinate under a space coordinate system; s333, calculating homonymous points.
5. The method for stitching panoramic images of an inner wall of a tunnel according to claim 4, wherein in step S34, the axially expanding the projected cylindrical surface comprises: s341, defining a cylindrical axial expansion diagram; s342, carrying out plane expansion on the three-dimensional surface element according to the homonymous corner points; s343, calculating the homography of the surface element; S344, performing surface element homography transformation; S345, recording a multi-view surface element development diagram.
6. The method for stitching panoramic images of an inner wall of a tunnel according to claim 3, wherein in step S35, the homogenizing the unfolded image comprises: s351, calculating affine transformation relations between adjacent view angle bin expansion graphs; s352, carrying out chain type compound calculation on affine transformation relation; s353, carrying out affine transformation and coordinate normalization; S354, performing texture mapping.
7. The method for stitching panoramic images of an inner wall of a tunnel according to claim 1, wherein in step S4, the stitching the continuous multi-segment horizontal stripe plane deployment map longitudinally comprises: s41, detecting offset of the transverse stripe plane unfolding images of the adjacent sections; S42, accumulating offset among all segments to construct a global transformation matrix sequence of all segments; S43, performing waveform correction to obtain a waveform compensation matrix; S44, compounding the waveform compensation matrix with the global transformation matrix to obtain final global transformation containing waveform compensation; and S45, carrying out longitudinal splicing based on the final global transformation to obtain the panoramic image of the inner wall of the tunnel.
8. The system for splicing the panoramic image of the inner wall of the tunnel is characterized by comprising a data acquisition module, a cylinder fitting module, a transverse strip generating module and a panoramic splicing module; The data acquisition module is used for acquiring point cloud data shot by the laser radar, multi-view images shot by the camera, and calibration data of the laser radar and the camera; the cylinder fitting module is used for geometrically reconstructing the inner wall of the tunnel according to the point cloud data to obtain a fitting cylinder; The transverse strip generating module is used for generating continuous multi-section transverse strip plane unfolding images according to the fitting cylinder, the multi-view images and the calibration data, wherein the lane width direction is transverse; The panoramic stitching module is used for stitching the continuous multi-section transverse strip plane unfolding images longitudinally to obtain a panoramic image of the inner wall of the tunnel, wherein the driving direction is longitudinal.
9. An electronic device, comprising: One or more processors; A storage means for storing one or more programs; The one or more programs, when executed by the one or more processors, cause the one or more processors to implement the tunnel inner wall panoramic image stitching method of any of claims 1-7.
10. A computer readable medium having stored thereon a computer program, wherein the program when executed by a processor implements the tunnel inner wall panoramic image stitching method according to any one of claims 1-7.

Description

Method, system, equipment and medium for splicing panoramic images of inner wall of tunnel Technical Field The present invention relates to the field of image processing technologies, and in particular, to a method, a system, an apparatus, and a medium for stitching panoramic images of an inner wall of a tunnel. Background The tunnel is used as a core component of infrastructure such as traffic, water conservancy and the like, and the integrity of the inner wall result is directly related to operation safety. With the development of intelligent infrastructure technology, machine vision technology is adopted to collect images of the inner wall of a tunnel and splice the images to generate a panoramic image, so that the intelligent infrastructure technology becomes a mainstream technical means of automatic inspection of the tunnel. The existing tunnel inner wall panoramic image stitching method is mainly divided into two types, namely pure image feature matching stitching and geometric model-based stitching. The pure image feature matching and splicing method takes SIFT (Scale-INVARIANT FEATURE TRANSFORM, scale invariant feature transformation), ORB (Oriented FAST and Rotated BRIEF, direction acceleration segmentation test feature and rotation binary steady independent basic feature) and other algorithms as cores, and completes image alignment and splicing by extracting image feature points, matching corresponding point pairs and solving homography matrixes, but the problems of sparse texture, uneven illumination, repeated structure and the like commonly exist on the inner wall of a tunnel, which easily cause insufficient feature point extraction or mismatching, further cause defects of splicing dislocation, ghost, obvious joints and the like, and meanwhile, the method does not consider the geometric characteristics of a cylindrical curved surface of the tunnel, is difficult to eliminate perspective distortion and annular distortion of images, and the splicing precision cannot meet the engineering level detection requirement. In the splicing method based on the geometric model, point cloud data of the inner wall of a tunnel is collected by means of a laser radar, a three-dimensional geometric model of the tunnel is built, and then multi-view images are mapped to the surface of the model to realize splicing. The method can improve the splicing precision by utilizing the geometric constraint of the point cloud, but has two technical bottlenecks that firstly, the registration of the point cloud and the image depends on high-precision external parameter calibration, the calibration process is complex and error is easy to introduce to influence the fusion effect, secondly, the data volume of the point cloud is huge, the pretreatment and the model construction are long in time consumption, the real-time requirement of tunnel inspection is difficult to meet, and furthermore, the single-point cloud geometric constraint-dependent splicing method cannot fully utilize the texture information of the image, so that the generated panoramic image has defects in detail expression. Part of the prior art tries to combine images and point cloud data to perform multi-mode fusion splicing, but mostly adopts a serial architecture of 'calibration before fusion', so that the problems of difficult data synchronization, error accumulation, poor robustness and the like exist, and a fusion strategy is not optimized for a special scene of the inner wall of a tunnel, so that the splicing precision, efficiency and texture quality of a panoramic image are difficult to consider. In addition, the prior art lacks the systematic design of strip type plane expansion and longitudinal splicing of the cylindrical curved surface of the tunnel inner wall, and continuous and high-precision panoramic reconstruction of the long-distance tunnel inner wall cannot be realized. In summary, the existing tunnel inner wall panoramic image splicing method has obvious short plates in the aspects of low texture scene adaptability, geometric distortion correction capability, multi-mode data fusion efficiency and the like, and cannot meet urgent requirements of tunnel engineering health monitoring on high-precision, real-time and high-robustness panoramic splicing technology. Therefore, it is needed to provide a method, a system, a device and a medium for stitching panoramic images of tunnel inner wall, so as to improve the accuracy and the efficiency of detecting panoramic images of tunnel inner wall. The above information disclosed in the background section is only for enhancement of understanding of the background of the application and therefore it may contain information that does not form the prior art that is already known to a person of ordinary skill in the art. Disclosure of Invention The invention mainly aims to solve the problems of low detection precision and low efficiency of a panoramic image of the inner wall of a tunnel, and provides a method, a sy