动态场景中视觉元素的注意序列表达及目标识别

摘要

为有效表达动态场景中的目标，实现对该目标的实时识别和判断，提出用注意序列来表达动态场景中视觉元素的方法。利用一组视觉元素构造观测向量，组成注意序列，注意序列的每一个观测向量是对视网膜中央凹滤波器响应的编码，有效表征视频图像的特征集，可以实现移动机器人对目标的识别。实验结果表明，注意序列可以充分表达动态场景，通过比较当前的观测向量和经过训练的平均观测向量，在数次扫视后可以快速实时地实现移动机器人对目标的有效识别和判断。%To express the targets in dynamic scenes efficiently and realize the real time object recognition and judgment,a method of expressing the visual elements in dynamic scenes using the attentional sequence was presented.A set of visual elements was used to construct observation vector,then attentional sequence was formed,in which each observation vector was the code of the response to the foveal filter.On this basis,the feature set of video image was represented effectively and the mobile robot recog-nizing the objects was realized.The experimental results show that the dynamic scene can be fully expressed using attentional se-quence,and through comparing the current observation vector and a mean observation vector after training,mobile robot recog-nizing and judging objects in real time can be realized in a number of saccades.