摘要
为了更好地模拟人类视觉系统中的注意力选择,本文提出一种改进型机器人仿生认知神经网络.首先模拟人类视觉皮层结构,在已有模型基础上建立改进型仿生认知神经网络模型;增加位置层(Position Motor,PM)到感受野(Receptive Field,RF)的自上而下(top-down)的视觉注意,同时下颞叶(Inferior Temporal,IT)不再接收全局视觉信息,而改为接收带有自下而上(bottom-up)视觉注意的局部信息,不仅降低数据处理的复杂度,也更加符合人类格式塔心理;最后利用该模型实现机器人复杂背景下目标识别与跟踪.实验结果证明该方法在有效减少数据冗余、缩短处理时间的同时,还可有效提高机器人视觉系统对目标的识别准确率.
To better simulate the attention selection in human visual system,an improved bionic cognitive neural network for robot is proposed. Firstly,to simulate human visual cortex structure,an improved bionic cognitive neural network is established on the basis of the existing models; it adds top-down visual attention from position motor( Position M otor,PM)to receptive field( Receptive Field,RF),and meanwhile,inferior temporal( Inferior Temporal,IT) no longer receives global visual information and turns to receive local information with bottom-up visual attention,not only reducing the complexity of data processing,but also keeping with human Gestalt psychology. Finally,the model is utilized to realize the robot target recognition and tracking in complex background. Experimental results showthat the method can reduce data redundancy and processing time,and also effectively improve the target recognition accuracy in the robot vision system.
出处
《电子学报》
EI
CAS
CSCD
北大核心
2015年第6期1084-1089,共6页
Acta Electronica Sinica
基金
教育部重大创新工程培育资金资助项目(No.708045)
国家自然科学基金项目(No.61325018)