All objects in the frame of surveillance video have equal status, so that the objects which users interested in may not have more attention. A two-level user visual attention model are proposed for video surveillance scenes in this paper. Firstly, classify foreground and background objects, and then, according to user’s understanding of foreground and background, abstractly express the visual importance related to feature property, finally, calculate the visual importance of the object according to visual attention model. In video surveillance scene, the model aim to extract a user visual attention model learn from the human vision system. The validity of the model has been confirmed through a series of experiments.%视频监控图像帧中所有物体具有同等地位，造成用户感兴趣的物体并没有能够分配到更多的视觉注意力。文章针对视频监控场景，构造了一个两层的用户视觉注意力模型，首先，进行前景和背景的分类，其次，通过用户对前景和背景的理解，将物体的视觉重要性进行特征属性的抽象化表述，最后，根据注意力模型计算物体的重要性。模型旨在借鉴人类的视觉系统，获取一个基于视频监控场景的用户视觉注意力模型。实验表明，该模型获得了良好的效果。
展开▼
