Shape of object recognition based on RS-ANN for mobile robot

摘要

When studying mobile robot to recognize shape of object in dynamic surroundings, we proposed a hybrid recognition algorithm based on the combination of rough set theory and BP neural network. RS has the capability for intelligent data analysis, and BP network can approach most problems accurately and exactly, the algorithm put respective advantages of two theories to use. Firstly, information table which was formed by training sample set was reduced by RS in order to find minimal decision regulations, and then the regulations confirmed the structure of ANN and recognized the shape by BP neural network. At the same time, the reduction of RS enhanced the efficiency of training sample set, and simplified the scale of neural network. Experimental results showed that the algorithm here had the better performance in exactness and speediness when compared with the only BP network.