Performance Analysis of Multi-step GA for Real-time 3D Model-Based Recognition for Underwater Vehicle

抄録

Nowadays, underwater vehicle has become essential tool such applications environmental survey, cable or pipeline tracking, unstructured undersea ocean exploration which is hazardous area. We have developed visual based underwater vehicle using dual-eyes cameras and three-dimensional marker. In this system,(position and orientation) pose between vehicle and target is estimated by means of 3D model-based recognition utilizing Multi-step GA. This paper analyses the performance of Multi-step GA for 3D recognition for underwater vehicle effectively based on the variations and modification in GA parameters. The recognition accuracy using Multi-step GA is evaluated by using offline full search. The indoor experimental results show that the proposed system can provide promising real-time 3D recognition accuracy.