Studies On Video Salient Region Detection Based On Visual Attention

Vision-based salient object detection technology is an important research topic in the field of computer vision. It is a critical technology of following applications such as intelligent video analysis. And it is wide prospect as well. Most methods of vision salient object detection are only applied to static background such as surveillance video. However, there are many limitations in video salient object detection with dynamic background.This paper applies the visual attention mechanism to video salient object detection employing the visual attention model to detect salient region of vehicle video which is based on region motion contrast. Furthermore, we detect unusual motion in the detected salient regions.The improved algorithm of video salient region detection is fit for object detection of dynamic video. Unusual motion detection based on vehicle video is reasonable as well. And we get good search results in a series of experiments. The key research work is shown as follows:Firstly, the vision attention mechanism and the technology of video moving object detection are reviewed, and the difficulties of video object detection are analyzed. We classify the methods of video object detection based on visual attention mechanism and introduce technologies of vehicle video object detection.Secondly, the paper details video salient region detection model based on region motion contrast. We analyze its weakness and present a new algorithm of video salient region detection base on the characteristic of vehicle video. The improved algorithm has considered wider sampling points and better choice of moving region. Experimental results show that the new algorithm can detect vehicle video salient region effectively.Then, based on the analysis of the characteristics of vehicle video and from the aspect of risk detection, this paper presents a definition of unusual motion in vehicle video and gives an unusual motion detection algorithm of vehicle video. And we do the experiments based on the result of Chapter 3. The experiments results show that the algorithm is effective.Finally, we design and implement a vehicle video unusual motion detection system based on visual attention mechanism. Then, system evaluation is made which shows that the algorithm is applicable to different scenarios with better accuracy.