有指向性的视觉注意计算机模型

A Computer Model of Directional Visual Attention

注意把有限的处理资源优先分配给那些需要精细加工的信息,能提高视觉信息加工中的检测能力和响应速度.基于生物视觉系统的生理结构特点,建立了模拟生物视觉注意系统的有指向性的视觉注意计算机模型.模型首先模拟生物视网膜的成像机制,将视场图像转化为视网膜图像;然后将最大梯度边缘检测和c-均值聚类等方法相结合,对视网膜图像中的目标进行编码,分别提取每个目标的颜色、中心以及边缘点集合等基本信息;最后用知识库中指向性目标的特征来指导注意焦点的转移.实验结果表明,利用此模型能较好地实现注意焦点的转移.

Attention assigns the limited processing resources to the information which needs processing finely. This feature can enhance the detection ability and response speed of the visual information processing. In the past few years, the allocation mechanism of attention has received a lot of attention and many approaches have been developed. The authors propose a computer model of directional visual attention, which can simulate biological visual attention systems based on their physiological structure. The proposed model has three layers： visual image inputting layer, object encoding layer and control layer. Visual image inputting layer transforms visual image into retina image through imaging mechanisms of biological retina and adiusts the radius of fovea to cover the attention object currently. Object encoding layer encodes objects in retina image by edge detection with maximum gradient and c-means algorithm firstly, then extracts the fundamental information about each object including color, central, set of edge pixels, etc. Control layer transfers the attention focus according to the characteristics of directional objects in knowledge base. In the control layer the space information is used, the object which is nearest from the focus of attention currently in space position has high priority in processing. Experimental results indicate that the attention focus can be transferred to directional objects with the proposed model.