Moving obstacle detection using residual error of FOE estimation

The paper presents a simple moving obstacle detection method which used the residual error calculated in uses of focus of expansion (FOE) estimation. Method can be applied to many industrial tasks such as an intelligent machine surveillance system or an obstacle detection system for an autonomous vehicle system. First, the optical flow field is extracted from sequence of dynamic imaged captured by a fixed camera on a moving observer. Next, the FOE is estimated in local image regions. Its residual error is first calculated in the region. An image region corresponding to the block is added with the residual error. This process is repeated by sliding and adding for the local region, and changing the size of the local region. Finally, regions which have high residual error values are detected as candidate regions of moving obstacles. Experimental results using real outdoor scenes show the effectiveness of the proposed method.