机器人触觉图像的聚类处理

Cluster Analysis of Tactile Patterns in Robots

利用数值分类学中的聚类方法，对机器人夹持工件的触觉图像进行处理，可以有效识别工件及其数目、空间范围等．该方法对文中给出的两大类工件都是适用的；在聚类处理过程中还可灵活消除噪声；与利用距离矩阵实现的系统聚类法相比较，本文利用“两次扫描”实现的聚类法仅需考虑特殊点，运算量很小．

The cluster analysis developed from numerical taxology are used to deal with the tactile patterns of robots to identify the object numbers and scales. The schemes are suitable to process 2 kinds of objects with quite arbitrary forms, easy to cancel noises, and less computation than regular cluster analysis due to “double scanning”.