Full-automatic searching and extracting method for contact network facility point cloud

Abstract

The invention discloses a full-automatic searching and extracting method for contact network facility point clouds, which comprises the following steps: S1, collecting laser point cloud data along a target railway, and automatically searching and extracting contact network point clouds in the laser point cloud data based on a double-selection three-dimensional frame: determining a searching range through a rough selection three-dimensional frame, extracting track line points in the range according to the frame; tracking cutting and extraction are carried out in the track direction through the careful stereoscopic frame; and S2, performing semantic segmentation on the cut and extracted contact network point cloud based on deep learning, wherein the semantic segmentation comprises the following sub-steps: S2-1, marking point cloud data in a manual mode; s2-2, an MFFA model is constructed; s3, reconstructing a three-dimensional model of the overhead line system based on geometric features; and S4, geometric parameter detection based on the contact network model. The method is high in speed, high in extraction precision and suitable for topographic relief and curves; the accuracy rate and the calculation efficiency of the contact network semantic segmentation method are superior to those of similar algorithms.

Description

technical field [0001] The invention relates to the field of railway facility monitoring, in particular to a fully automatic search and extraction method for catenary facility point clouds. Background technique [0002] With the advantages of fast speed, high safety, good comfort and high cost performance, railway transportation has become a major mode of modern transportation. At present, railway transportation has gradually become one of the most important means of transportation between cities. Electrified railway is the main form of existing railway. Overhead Catenary System (OCS) refers to the electromechanical system that provides electric energy to the electric traction unit through the current collector. Such as figure 1 As shown, it usually consists of several parts such as contact suspension, support device, positioner, pillar and foundation. Wherein, the contact suspension includes contact wires, hanging strings, catenary cables and connecting parts, and the c...

AI Technical Summary

Problems solved by technology

[0006] (1) Relying more on manual pre-sample extraction and processing, and lack of automatic search and extraction methods for catenary facilities;

[0007] (2) In the deep neural network, from the perspective of feature extraction, many networks map features to high-dimensional space to obtain more and richer high-dimensional features, but they cannot effectively perform feature weight matrix according to feature importance. Distinguished, resulting in a large number of important feature information cannot be effectively transmitted and discarded

After the introduction of attention mechanism in a few network models, this situation has been improved, and the segmentation accuracy has been slightly improved, but the amount of calculation has increased sharply, and there is no effective attention mechanism to increase the weight coefficient of important features;

[0008] (3) The refinement structure plays an increasingly important role in the image depth semantic segmentation network. At present, there are few applications of the refinement structure in the preliminary segmentation results of point clouds and then fine segmentation, and the initial segmentation results cannot meet the application requirements. Segmentation accuracy;

[0009] (4) Lack of extraction and fusion of multi-scale features, insufficient fusion of shallow features and deep features, resulting in serious loss of effective feature expression

Images