Research On Hyper-Spectrum Image Classification Of Mineral Based On ANN

Spectral imaging technology, appeared in the early 1980s, has achieved significant advances with the development of high-performance computing technology and spectral system. At present, hyper spectral image is a high resolution remote sensing images, which in further provides more surface information. Its applications cover the area of mineral resources investigation, environmental and ecological monitoring, marine atmosphere research and so on, which has been considered as one of the most cutting-edge disciplines.This paper focuses on the application of hyper spectral mineral recognition based on artificial neural network (ANN). It covers spectrum theory, spectral properties of minerals, and artificial neural network theory. It also designs and implements a Back-Propagation Multilayer Feed Forward Network for mineral recognition.The data information of hyper spectral is huge and disturbance, so the traditional methods of classification are not available for practical application. However, artificial neural network with the characters of parallel processing, fuzzy recognition and nonlinear mapping will make itself suitable for hyper spectral recognition. But there also come the problems. The input eigenvector with high dimension and the huge sample space may make long time for network learning or even cause failure of training. That is why we should adjust the parameter of the network for optimization. The principle and design of the artificial neural network model will be described in detail in this paper. The main factors of the artificial neural network will also be discussed. Practices show that the recognition based on artificial neural network is effective and feasible.