Motion Stereo including Tracking Stability Analysis for 3-D Cable Reconstruction

Abstract

We propose a novel approach to the three-dimensional reconstruction of flexible cables for applications in factory automation, such as cable handling and connecter insertion using robotic arms avoiding conflicts among multiple cables. The approach is based on motion stereo with a single vision sensor. To solve the stereo correspondence problem efficiently and effectively, feature-point projection with slit beams and a feature tracking algorithm based on a robust image-matching method is applied. In addition, we define the tracking stability of the feature points in order to reject defective stereo correspondences. Experiments using these methods demonstrate that an arched cable shape can be reconstructed to an accuracy of 1.5%.