CN-122023356-A - Building nondestructive mapping method based on multi-mode data comprehensive analysis

CN122023356ACN 122023356 ACN122023356 ACN 122023356ACN-122023356-A

Abstract

The invention is suitable for the technical field of mapping, and provides a building nondestructive mapping method based on multi-mode data comprehensive analysis, which performs non-contact self-adaptive data acquisition through close-range photography and laser scanning cooperation; the method comprises the steps of carrying out fusion processing on multi-source data to construct a high-precision integrated three-dimensional model, taking a first model as a reference, utilizing an improved time sequence registration and semantic segmentation algorithm after periodical retesting to automatically extract micro-deformation information of key components, and finally generating a three-dimensional result with protection attributes and risk labels based on deformation analysis and outputting the three-dimensional result to a management platform. The invention solves the problems that the prior art is difficult to achieve nondestructive operation, high-precision information acquisition, intelligent deformation monitoring and achievement practicability, and realizes the full-flow nondestructive, efficient and intelligent operation from accurate digital gear establishment to long-term health monitoring of the historical building.

Inventors

LI PENG
GUI ZHIMING
TAN XIAOSONG
HE DEPING
JIANG BEN
PAN KE
HUANG LEI
SUN JIANHUA
HE CHAO
Nie Tuxiang
LI SHA

Assignees

重庆市测绘科学技术研究院

Dates

Publication Date: 20260512
Application Date: 20260202

Claims (10)

1. The building nondestructive mapping method based on multi-mode data comprehensive analysis is characterized by comprising the following steps of: S1, cooperatively acquiring close-range photogrammetry data and ground three-dimensional laser scanning data of a historical building in a non-contact mode through a close-range photogrammetry unit and a ground three-dimensional laser scanning unit; S2, constructing an integrated three-dimensional model through a multi-mode data fusion algorithm based on the close-range photogrammetry data and the ground three-dimensional laser scanning data; S3, retesting according to a monitoring period, taking the integrated three-dimensional model constructed by the first execution of the steps S1 to S2 as a reference period integrated three-dimensional model, and repeating the steps S1 to S2 in a subsequent monitoring period to obtain the retesting period integrated three-dimensional model; s4, registering the retest period integrated three-dimensional model and the reference period integrated three-dimensional model through a time sequence micro-deformation recognition algorithm, and extracting micro-deformation information of a key component; and S5, generating and outputting the achievement with the three-dimensional label based on the micro-deformation information.
2. The method of claim 1, wherein the collaborative acquisition in S1 comprises: based on the identification of building materials and component structures, the laser power and the point cloud density of the ground three-dimensional laser scanning unit and the shooting angle of the close-range photogrammetry unit are adaptively adjusted.
3. The method according to claim 2, wherein the adaptive adjustment specifically comprises: when the wood member is identified, controlling the laser power of the ground three-dimensional laser scanning unit to be not higher than 5mW, and controlling the point cloud density to be not lower than 800 points/square centimeter; and when the three-dimensional laser scanning unit is identified as a masonry member, controlling the ground three-dimensional laser scanning unit to adopt conventional laser power, wherein the point cloud density is not lower than 500 points per square centimeter.
4. The method according to claim 1, wherein the multi-modal data fusion algorithm in S2 performs the following process: carrying out statistical filtering and radius filtering on the ground three-dimensional laser scanning data to remove noise; performing distortion correction on the close-range photogrammetry data; And (3) merging the denoised point cloud data with the corrected image data based on a RANSAC algorithm to generate a structure and texture integrated three-dimensional model.
5. The method of claim 1, wherein in S4, registration is performed using a temporal micro-deformation recognition algorithm, employing the introduction of semantic weighting factors The registration error objective function is: E=min R,t ‧‖R Pi +t-q i ‖ 2 ; Wherein: r is 3 multiplied by 3 rotation matrix, describing the space rotation relation of two-stage point clouds; t is 3 multiplied by 1 translation vector, describing the space translation relation of two-stage point clouds; p i three-dimensional coordinates (x i ,y i ,z i ) of an ith point in the reference period point cloud; Q i the three-dimensional coordinates (x i ′,y i ′,z i ') of the ith point in the retest period point cloud; W i , semantic weight factor corresponding to the ith point; N, the total point number of the point clouds participating in registration; And E, registering error sum.
6. The method of claim 5, wherein the micro-deformation information extracted in S4 includes a cumulative deformation distance Δdi and a deformation rate vi of the monitoring point, and the calculation formula is: Δdi=√((xi′−xi) 2 +(yi′−yi) 2 +(zi′−zi) 2 ); vi=Δdi/T; Wherein: Δdi, cumulative deformation distance of the i-th point; vi, deformation rate of the ith point; t is the time interval between two data acquisitions.
7. The method of claim 6, wherein extracting micro-deformation information of the key member further comprises: identifying a key component area from the integrated three-dimensional model by using a U-Net semantic segmentation algorithm; And extracting the deformation amount by calculating the differential point cloud of the key component area.
8. The method of claim 1, wherein the three-dimensional tag in S5 comprises a component protection level, a component material weakness level, and a deformation risk level determined based on the micro deformation information.
9. The method of claim 1, wherein the outcome of the S5 output comprises: an integrated three-dimensional model file, a deformation thermodynamic diagram and a monitoring report attached with the three-dimensional label; the achievements are synchronized to a historical building protection management platform through a standardized data interface.
10. The method of claim 1, wherein S1 is preceded by the preliminary steps of laying out at least four control points around the perimeter of the target building and calibrating the three-dimensional coordinates of each control point using a high-precision total station.

Description

Building nondestructive mapping method based on multi-mode data comprehensive analysis Technical Field The invention belongs to the technical field of mapping, and particularly relates to a building nondestructive mapping method based on multi-mode data comprehensive analysis. Background The historical architecture is used as a non-renewable cultural heritage, and the accurate mapping and long-term health monitoring are the basis for implementing scientific protection and preventive maintenance. Currently, related technical solutions in this field mainly surround the acquisition, processing and application expansion of high-precision geometric information and realistic texture information. Mapping schemes commonly employed in the industry can be divided into two categories, contact and non-contact. The contact measurement mainly depends on devices such as a total station, a tape measure and the like, and dimension and deformation data are obtained by directly contacting a building body. The non-contact measurement mainly adopts a single technical path, namely, only a close-range photogrammetry technology is used for acquiring high-resolution images and reconstructing a three-dimensional model, or only a ground three-dimensional laser scanning technology is used for acquiring high-density three-dimensional point cloud data. These techniques provide the underlying digital means for building records. However, the prior art solutions described above still have comprehensive technical limitations when applied to fragile, complex historic buildings. Contact operation inevitably causes physical wear or contamination to the painted, wooden or masonry surface, against the first principle of protection. The single non-contact technology is difficult to meet the dual requirements of millimeter-level geometric precision and submillimeter-level texture details, so that the digital file has a short message board. In addition, in the process from data acquisition to analysis application, the self-adaptive optimization of building material and structure specificity is generally lacking, the multi-period data comparison precision is insufficient, early and tiny deformation early warning is difficult to realize, the finally generated result is mainly a simple geometric model, and the method is disjointed with the actual protection management, risk rating and repair decision requirement, and has limited practicability. Disclosure of Invention The invention aims to provide a building nondestructive mapping method based on multi-mode data comprehensive analysis, and aims to solve the technical problems in the prior art determined in the background art. The invention is realized in such a way that a building nondestructive mapping method based on multi-mode data comprehensive analysis comprises the following core steps: Texture and structure data of the historical building are cooperatively collected in a non-contact mode through a close-range photogrammetry unit and a ground three-dimensional laser scanning unit. In the process, scanning and shooting parameters are dynamically adjusted according to real-time identification results of building materials and component structures through a self-adaptive acquisition parameter generation algorithm, so that lossless, efficient and targeted data acquisition is realized. And preprocessing and fusing the acquired multi-source data. The method comprises the steps of cleaning data through technical means such as point cloud denoising and image distortion correction, accurately registering and fusing high-precision laser point cloud and high-resolution images based on an RANSAC algorithm and the like, and constructing a structure and texture integrated three-dimensional model with millimeter-level geometric precision and sub-millimeter-level texture details, wherein the structure and texture integrated three-dimensional model is used as a reference for digital gear establishment and subsequent deformation analysis. And (5) performing time sequence micro deformation intelligent monitoring. After a new first-stage integrated three-dimensional model is obtained through periodic retesting, an improved ICP algorithm introducing a component semantic weight factor is adopted to carry out high-precision registration on multi-stage data, and the registration error of a key area is ensured to be less than or equal to 0.3mm. And combining U-Net semantic segmentation and differential point cloud analysis, automatically and accurately identifying and quantifying micro deformation of the key component which is more than or equal to 1mm, and calculating the deformation rate. Generating a protection-oriented labeling result. And correlating the extracted micro deformation information with the attributes such as the protection level of the building, the fragility of the component and the like, attaching a three-dimensional label containing the risk level to the component in the three-dimensional model, aut