摘要
与视觉相比,机器人触觉图象具有分辨率低、噪声大、信息不完整等特点,若采用常规分类器,识别率往往不高.本文基于 Hopfield 联想记忆神经网络,构造了一种适合于识别信息不完整、噪声较大的图象的分类器,并成功地应用于智能机器人触觉图象的识别.通过实验,我们对这种分类器的性能与三种常规分类器——最小距离、最近邻和贝叶斯分类器作了比较.
Compased with vision,robotic tactile image are usually of low resolution,noisy and incomplete,so theclassification accuracy is often not satisfied when using conventional classifiers.In this paper,we present aclassifier based on the hopfield associative memory(HAM),and applied it successfully to the recognition oftactile images.The comparison between the associative memory classifier and three conventional classifiers isshown in the exnerimente.
出处
《机器人》
EI
CSCD
北大核心
1993年第6期48-51,共4页
Robot
基金
国家自然科学基金项目:机器人触觉智能系统(顾学真等)
模式识别国家重点实验室课题基金
