On-line Kind’s Recognition of Auto Rack Girders Based on Combination of Fuzzy ART Neural Network with D-S Evidence Theory

摘要

Multi-character information fusion technology has been applied in sort and pattern recognition widely.For the question of artificial recognition on hundreds of camion rack girders being difficult,this paper introduces an on-line automatic inspecting method which synthesizes machine vision 、wavelet transform theory、Fuzzy ART neural network and D-S evidence theory on auto rack girders.Firstly,for the real-time gathered auto rack girders top images on assembly line,extract three character templates,and describe characters of images from the different aspects respectively.Top image of auto rack girders is partitioned to 16 sub-regions (4×4),account local entropy of 16 sub-regions respectively,which are used as a character template; in the same way,images are pre-processed by binary image,partitioned to 16(4×4) sub-images,accounting Normalized Moment of Inertia (NMI) of every sub-region separately,which is used as a character template; extract wavelet decomposition coefficient of images with 3-layer wavelet,energy values of 9 wavelet coefficient are used as a character template.Secondly,in order to gain basic confidence of recognition,three character templates data which are local entropy 、NMI and energy value of wavelet coefficient are used as inputs of Fuzzy ART neural network.Finally,according to composition rule of D-S evidence theory,gain total confidence of recognition.Experiments indicate,this method possessed advantage of more rapid 、more precise recognition and stronger anti-interference,and recognition rate meets demands of production.