Recognition of Fabric Structures Based on Improved Olfactory Neural Network

This paper presents an improved KIII stimulation model based on olfactory neural network (ONN) system. Taking the output responses of M1 nodes as benchmarks, the model chooses 8-channel KIII network. Analyses of both the methods for taking values of M1 nodes and the cross-connect weights among M1 nodes in different channels have provided us with an approach for taking response values of M1 nodes in a back-to-front way, and a quantitative method called first multiplication then addition, to the input impetus. The experimental results show that this model has greatly improved the recognition rate of the plain weaves, twill weaves, stain weaves and complex weaves, compared to former methods based on Neural Network. In addition, the average recognition speed of the 8-channel KIII network model is much quicker than 64-channel model.