Consistent target tracking via multiple underwater cameras

摘要

Multi-targets tracking has shown great prospects in ocean investigations. In this paper, we apply the adaptive background mixture models to extracts the moving target from the image sequences, and propose a new method which is made use of the EKF and SURF-RANSAC to realize multi-targets tracking via multiple underwater cameras. In this method the centroid coordinate homographic mapping and the Speeded Up Robust Features (SURF) technique have been taken to capture and match the same target from the multiple cameras. Simulation experiments have shown that the proposed method could perform robustly and stably to perform the multi-targets tracking.