Real-time object segmentation for visual object detection in dynamic scenes

This paper presents a real-time object segmentation approach for visual object detection in dynamic scenes. This object segmentation approach is based on a novel general object feature which is defined subtly combining multiple low-level features and the uniqueness of the target object. Then the object segmentation approach is applied to detect vehicle and lane marking in dynamic scenes. Experiment results with test dataset extracted from real traffic scenes on highways and urban roads show that the approach proposed in this paper can achieve a high detection rate with an extreme low time cost.