The Fabric Recognition Based On Olfactory Neural Network

Fabric recognition technology has not yet developed to meet practical requirements, the technical problems are mainly in the diversity of the organizational structure and removing a number of complex factors's impact. The diversity of the organizational structure can not be changed , excluding a number of complex factors that influence is difficult to achieve, because a way for certain type of noise interference is effective while other methods would not necessarily have effects. Currently the international organization's ability to identify the fabric still remain in the three basic organizational structure, there is no effective method to identify the complex organization like complex changing organization . In this context propose to use olfactory neural network , which has good anti-noise performance , to identify fabric .In this paper, at first systematically study the olfactory neural network, and propose to improve the olfactory neural network through experimental data analysis .Then experimentally compare fabric's pretrea- tment methods, and suggest to improve the program, and use the improved 8-channel olfactory neural network to identify fabric.(1) Carry out in-depth experimental analysis of K0, KI and KII mathematical models and draw that the time that K0 model (PG layer) to achieve the stability in different circumstances incentive is convergent to a constant; analyse two kinds of KI model's detailed input-output features; and the input and output features of KII model (OB layer) in different input voltage or under different input time incentives is also analyzed in detail.(2) Carry out in-depth experimental analysis of the mathematical model of the KIII, confirm some of the shortcomings of the experimental methods of the past scholars by experimental data, improve training methods and handling of input values and improve the whole olfactory neural network recognition rate.(3) Compare the recognition effect of the pre-processing fabric tissue images through experiments and the experimental result in presenting a new method to extract the fabric point. The recognition rate of the twill weave greatly improves.(4) Study the recognition of complex changing organization, raise the interval sampling method within the minimum cycle, and the recognition rate is greatly improved than the previous methods. The above experiments were carried out under VC2005 platform simulation.