Overhead Wire 3D Recognition from Multi-Viewpoint Line Segments
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for overhead wire recognition using digital cameras struggle with low accuracy 3D point clouds generated from multi-viewpoint images, making it difficult to accurately identify and model overhead wires due to errors and challenges in feature matching.
Innovation Solution
An overhead wire recognition device and method that utilizes an input unit for multi-viewpoint images, an overhead wire candidate extraction unit to select and extract line segments, and an overhead wire model estimation unit to project and estimate a 3D model of the wire based on camera positions and attitudes, allowing for accurate recognition without the need for expensive laser scanners.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If SfM and MVS are used to generate 3D point clouds from multi-viewpoint images, then the cost of equipment is reduced, but the measurement precision deteriorates due to low accuracy and errors in feature matching
Solution Approach 1:
The patent segments the overhead wire recognition process into distinct stages: extracting line segments from 2D images, projecting them to 3D space, and fitting catenary curves. This segmentation allows each stage to be optimized independently, maintaining accuracy while using affordable cameras instead of expensive laser scanners.
Solution Approach 2:
The patent changes the approach from working with raw 3D point clouds to working with 2D line segment parameters extracted from images. By parameterizing the overhead wire as a catenary curve and using projection geometry, the system achieves accurate 3D modeling without relying on high-precision 3D point cloud data.
2Adaptability or versatility
If multi-viewpoint images are used for overhead wire recognition, then the versatility of the system is improved, but the difficulty of detecting and measuring increases due to challenges in feature matching
Solution Approach 1:
The patent extracts only the essential features needed for overhead wire recognition: line segments from 2D images. By taking out and focusing on these key geometric features rather than attempting to match all image features, the system simplifies the detection process while maintaining versatility across different imaging conditions.
Solution Approach 2:
The patent uses 2D line segment projections as simplified copies of the 3D overhead wire structure. Instead of directly measuring and matching complex 3D features, the system works with 2D projections that can be easily extracted and matched, then reconstructs the 3D model through mathematical projection and curve fitting.
Data Source
AI summary
An overhead wire recognition device and method are capable of overhead wire recognition based on plural images taken by an imaging device such as a digital camera. The overhead wire recognition device includes an input unit to which multi-viewpoint images taken by a camera from different viewpoints and camera positions/attitudes are inputted, an overhead wire candidate extraction unit which selects, from the multi-viewpoint input images inputted to the input unit, an overhead wire region including an overhead wire and extracts, from the overhead wire region, a line segment as a candidate of the overhead wire, and an overhead wire model estimation unit which projects the candidate line segment in an imaginary 3D space, determines, based on the candidate line segment, the camera positions/attitudes, and the multi-viewpoint images, a plane where the overhead wire is present in the 3D space, and estimates a 3D model of the overhead wire.