Moving Object Detection in Dynamic Scene Using Spatial-Temporal Condition Information

This paper proposes a novel method to solve the non-linear classification problem of moving object detection in dynamic scene with spatial-temporal condition information.Our method converts the non-linear classification problem to a linear one by applying negative logarithm kernel transform algorithm.In order to get higher detection accuracy,a visual salient spatial-temporal domain is constructed to extract spatial-temporal condition information.Neighborhood weighted spatial-temporal condition information is employed for pixel classification.Combing with image block difference pre-detection,our proposed method can effectively detect moving object in dynamic scene in real time.