CN-121982578-A - In-service power transmission tower disease identification method and system based on multi-mode perception

CN121982578ACN 121982578 ACN121982578 ACN 121982578ACN-121982578-A

Abstract

The invention belongs to the field of power transmission tower modeling, and provides an in-service power transmission tower disease identification method and system based on multi-mode perception, which are used for acquiring images, laser point clouds and positioning data obtained by multi-angle surrounding shooting of a power transmission tower body and the periphery; the method comprises the steps of preprocessing acquired data, deleting invalid data, performing aerial triangulation based on preprocessed laser point cloud data and generating sparse point clouds by using preprocessed positioning data as beam adjustment constraint, constructing a three-dimensional reconstruction model of a tower body based on preprocessed images and the generated sparse point clouds, marking an interested region containing key components, and identifying and obtaining a disease detection result based on the interested region by using a pre-trained detection model. The invention realizes the depth fusion of the multi-mode data, can play the advantage of the texture definition of the image, and can obtain the integrated target of high-precision three-dimensional reconstruction of the power transmission tower and accurate disease identification and positioning by utilizing the geometric high-precision characteristic of the laser.

Inventors

LIU KAIYUE
CHEN ANXIN
SONG HONGZHU
CHU HONGMIN
WU XIAOMENG
JIA KEQIN
YAO XINLONG
MA RUISHENG
CAO SHUTING

Assignees

山东电力工程咨询院有限公司

Dates

Publication Date: 20260505
Application Date: 20251212

Claims (10)

1. The in-service power transmission tower disease identification method based on multi-mode sensing is characterized by comprising the following steps of: Acquiring an image, laser point clouds and positioning data obtained by surrounding shooting of a power transmission tower body and the periphery at multiple angles; preprocessing the acquired data, and deleting invalid data; Performing aerial triangulation based on the preprocessed laser point cloud data and generating sparse point cloud by using the preprocessed positioning data as beam adjustment constraint; constructing a three-dimensional reconstruction model of the tower body based on the preprocessed image and the generated sparse point cloud, and labeling an interesting area containing key components; and identifying and obtaining a disease detection result based on the region of interest by utilizing a pre-trained detection model.
2. The method for identifying the in-service power transmission tower diseases based on multi-mode perception according to claim 1, wherein the process of acquiring the images, the laser point clouds and the positioning data obtained by multi-angle surrounding shooting of the power transmission tower body and the periphery comprises the steps of carrying a visible light camera and a laser radar by using an unmanned plane, carrying out multi-angle surrounding shooting on the power transmission tower body and the periphery under the constraint of setting a flight height range, a heading overlapping range and a side overlapping range, simultaneously shooting key connection parts and detail parts of the power transmission tower at a short distance as characteristic areas, and synchronously acquiring high-precision positioning data to obtain an original image, a laser radar original data packet, base station observation data and positioning data.
3. The method for identifying the diseases of the in-service transmission tower based on multi-mode sensing as claimed in claim 1, wherein the process of preprocessing the acquired data comprises the following steps: Screening and removing invalid images with fuzzy, overexposure or insufficient overlapping degree, and checking consistency of integrity of shooting metadata and a time stamp; And carrying out stripe calculation, filtering and denoising on the laser radar original data to obtain a geographically registered point cloud, starting the image positioning metadata as beam adjustment constraint, carrying out aerial triangulation and generating a sparse point cloud.
4. The method for identifying the in-service power transmission tower diseases based on multi-mode perception according to claim 1, wherein in the process of preprocessing acquired data, the characteristic that repeated textures of angle steel and bolts of the power transmission tower are easy to cause automatic matching misjudgment is considered, a plurality of groups of rule features are marked as user connection points, each group covers more than a set number of images, and a consistency corresponding relation of the cross images is established so as to strengthen geometric constraint and remarkably reduce mismatching rate.
5. The method for identifying the in-service transmission tower diseases based on multi-mode sensing according to claim 1, wherein in the process of preprocessing acquired data, pseudo control points, control lines or structural planes extracted by laser point clouds are used as additional space strong constraints, namely, rough registration of an image model and the point clouds is finished by using positioning poses, then, a stable member of a transmission tower is used as an anchor, and fine registration of point-to-surface or surface-to-surface constraints is executed, so that beam adjustment is geometrically constrained by a laser radar and is dominated by images on textures, and local distortion which only depends on positioning data to appear in the bottom of a tower body and a repeated texture area is restrained.
6. The method for identifying the in-service power transmission tower diseases based on multi-mode perception according to claim 1, wherein when a three-dimensional reconstruction model of a tower body is constructed, a reconstruction task is created to output a high-resolution texture grid, pixels are set for texture image quality exceeding a set value and maximum texture size, sharpening is started to obtain a high-precision three-dimensional reconstruction scene of the power transmission tower, the reconstruction scene is subjected to semi-automatic segmentation, and based on the reference geometry of the power transmission tower only after segmentation, the center line, the space node coordinates and the connection relation among nodes of each rod piece are automatically identified and extracted through skeleton extraction and topology analysis of the tower body model, so that the three-dimensional reconstruction model of the tower body is constructed.
7. The method for identifying in-service transmission tower diseases based on multi-mode sensing according to claim 1, wherein the key components comprise bolt groups, hardware fittings and insulators.
8. The method for identifying the in-service transmission tower faults based on multi-mode sensing as claimed in claim 1, wherein the detection model is a big data model, and the state information and the three-dimensional coordinate information of the faults are obtained by utilizing the pre-trained big data model to identify defects including insulator breakage, hardware wear, rod rust, bolt missing and basic cracks based on the region of interest.
9. The method for identifying in-service transmission tower diseases based on multi-mode sensing as claimed in claim 1, further comprising the steps of: and (3) carrying out automatic space registration and data comparison on the apparent diseases and the three-dimensional coordinate information thereof identified by the big data model and the extracted rod piece structural data and space coordinates, and marking the defects on corresponding nodes of the three-dimensional reconstruction model of the tower body when the defects are successfully matched, so as to realize automatic mapping from visual disease identification to defects of the structural mechanical model.
10. An in-service transmission tower disease identification system based on multi-mode sensing is characterized by comprising: The multi-mode data acquisition module is configured to acquire images, laser point clouds and positioning data obtained by surrounding shooting of the power transmission tower body and the periphery at multiple angles; the preprocessing module is configured to preprocess the acquired data and delete invalid data; the sparse point cloud generation module is configured to perform aerial triangulation based on the preprocessed laser point cloud data and generate sparse point clouds by using the preprocessed positioning data as beam adjustment constraint; The interest region labeling module is configured to construct a three-dimensional reconstruction model of the tower body based on the preprocessed image and the generated sparse point cloud, and label the interest region containing the key components; and the disease identification module is configured to identify and obtain a disease detection result based on the region of interest by utilizing a pre-trained detection model.

Description

In-service power transmission tower disease identification method and system based on multi-mode perception Technical Field The invention belongs to the field of modeling of power transmission towers, and particularly relates to an in-service power transmission tower disease identification method and system based on multi-mode perception. Background The statements in this section merely provide background information related to the present disclosure and may not necessarily constitute prior art. The power transmission tower is used as a key infrastructure of a power system, and intelligent inspection, defect detection and digital operation and maintenance can be realized by three-dimensional reconstruction of the power transmission tower. At present, three-dimensional reconstruction technology has formed various technical paths in the fields of buildings, bridges, cities and the like. Existing three-dimensional reconstruction techniques are roughly divided into two types, unmanned aerial vehicle-based multi-view photogrammetry and laser radar-based measurement. The method is characterized in that the method comprises the steps of acquiring images through vertical and oblique view angles during urban reconstruction, training a local reconstruction model by utilizing a three-dimensional grid model and splicing the local reconstruction model into a global model, and realizing complete coverage of the surface and the hole site of the pier through a design route by adopting a full scene coverage strategy during pier measurement. The laser radar measurement technology is applied to reconstruction of a railway station network frame, a double-laser radar and ultrasonic obstacle avoidance unmanned aerial vehicle system is adopted, and a three-dimensional model is built through panorama coverage and a point cloud generation algorithm. Aiming at three-dimensional reconstruction of a power transmission tower, the prior art mainly has the following problems: (1) The problems of ambiguity in texture matching and insufficient scene adaptation can occur when the transmission tower is measured based on unmanned aerial vehicle multi-view photography. The system is lack of full-flow adaptation of the power industry standard, is not in butt joint with a power inspection system in the aspects of acquisition parameters, model precision and the like, and cannot directly support operation and maintenance functions such as space measurement, defect identification and the like; (2) The point cloud model generated based on laser radar measurement is used for solving the problems of texture information loss and poor flexibility when used for a power transmission tower. The point cloud data generated by the laser radar is easy to generate noise at the positions of the micro parts such as bolts, hardware fittings and the like, the model precision is influenced, and the operation and maintenance requirements are difficult to directly service; (3) Whether unmanned aerial vehicle multi-view photogrammetry or laser radar measurement is adopted, only a visual model of the power transmission tower can be built, a technical fault from visual reconstruction to structural mechanics analysis exists, and finite element analysis and rapid integrated modeling cannot be directly supported. Disclosure of Invention In order to solve the problems, the invention provides a method and a system for identifying in-service transmission tower diseases based on multi-modal sensing, which realize the deep fusion of multi-modal data, the method can not only play the advantage of the texture definition of the image, but also obtain the integrated target of high-precision three-dimensional reconstruction of the power transmission tower and precise disease identification and positioning by utilizing the geometric high-precision characteristic of the laser. According to some embodiments, the present invention employs the following technical solutions: an in-service power transmission tower disease identification method based on multi-mode sensing comprises the following steps: Acquiring an image, laser point clouds and positioning data obtained by surrounding shooting of a power transmission tower body and the periphery at multiple angles; preprocessing the acquired data, and deleting invalid data; Performing aerial triangulation based on the preprocessed laser point cloud data and generating sparse point cloud by using the preprocessed positioning data as beam adjustment constraint; constructing a three-dimensional reconstruction model of the tower body based on the preprocessed image and the generated sparse point cloud, and labeling an interesting area containing key components; and identifying and obtaining a disease detection result based on the region of interest by utilizing a pre-trained detection model. The process of acquiring the images, the laser point clouds and the positioning data obtained by the multi-angle surrounding shooting of the power transmiss