A new automatic obstacle detection method based on selective updating of Gaussian mixture model

Abstract

Obstacle detection is a hot topic in intelligent visual surveillance system. This paper proposed an automatic obstacle detection method applying to traffic surveillance, which can be used to prevent the traffic accident. In our framework, the images are captured by the traffic surveillance. The GMM (Gaussian Mixture Model) is taken as a short-term background, and foreground objects are extracted by the algorithm SUOG (Selective Updating of GMM). At last, a detection method related object speed and FROI (Flushed Region of Interest) algorithm is proposed. FROI algorithm is based on the concept of connected domain and used to eliminate noises outside road and improve real-time capability. Experiments demonstrate that the proposed obstacle detection method can detect the obstacle effectively and accurately, it can fulfill the requirement of practical application.