A Background Modeling Approach Based on Visual Background Extractor
Can Liu, Lin Qi, Guichi Liu · 2015
Moving objects detection is an important research content in computer vision field， and it plays a very important role in many vision applications such as smart video surveillance, intelligent transportation, and human-computer interaction, traffic control, activity recognition, object tracking and behavior understanding.The visual background extractor (ViBe) is a classical approach, however, when a moving objects in the initialized frame, it will take a lot time to fuse the gost and when there braches or surface of the water in the frame, it can make more false detection.In this paper, an algorithm based on ViBe is proposed.Though frame subtraction we acquire a real background, moreover we initial the model.On the other hand, we use 3-5 frames initial the model.The improved approaches solve the problems of gost or blinking pixels.Experiment show that, the algorithm can improve the effective of detection in the specific scene with similar computer load.