Improved Artificial Immune System And Its Application On The Process Of Hyperspectral Data

Recent years, with the developing of sensor technology, hyperspectral image has got unprecedented development. The quantity of the bands of the hyperspectral image data is very large, and the correlation of the bands is quite strong, too. On one hand, the hugeness of the data brings difficulties to not only data storage, but also data processing. On the other hands, it holds back the applying of the hyperspectral image data in some degrees. In the progress of remote data analyzing, the quantity precision affects the measure complexity and the bands number very much. The hyperspectral image data contain lots of bands, between which there are lots of pertinences, too. So it produces large data redundancies. It is emergent to design new method to process the hyperspectral data.This Paper describes some AIS (Artificial Immune System) based models and methods during the process of Hyperion data. AIS is an intelligence system inspired from the biological immune theory. The principles of biological immune theory are inspired more and more to design artificial immune model and algorithms. AIS has been introduced into many subjects which include data mining, dynamic protection, pattern recognition, self control, artificial intelligence, information security, multimodal function's optimization and so on. Traditional AIS can not be applied to the processing of hyperspectral image. To satisfy the processing of hyperspectral image, we need consider the characteristic of hyperspectral image firstly, and then improve the original algorithms even rebuild the models. This paper does a series of works on this.This paper designs a progress to inspire the sensitive bands of corn spectral data, and improve the method by introducing the AIS theory. It provides experiences for the research of corn intimidation on the selected fields.Firstly, this paper does lots of researches on the history and status quo of AIS, and learns experiences by collect others' researches of AIS on their major. Especially, it costs a lot that make use of AIS to solve multi-objective optimization problem.Secondly, this paper concludes and analyses the existing approaches on dimensional reduction of Hyperion data, which is abstracted to a mathematical problem. OIF index is selected as the evaluation criteria. HDRM model is built by inspiring AIS theory. An experimental image is used to test the model.On the end, this paper introduces a method to selecting sensitive points of the corn spectral data. The clonal selection algorithm of AIS is used in the experiment. The experiment use SAM(spectral angle mapper) as the evaluation criteria. Under the support of measurement data from Changchun and Beijing, four important points are extracted from corn spectral sets from three experimental plots, and the result indicates that the precision of discrimination can reach 80%.