Research And Application Of Hyper-spectral Image Spectrum Matching Technology Based On Knowledge

With the development of remote sensing technology, the resolution of remote sensing is also improving. At present it is applied widely in the areas of updating the current base geographic data, change of land using, land resources survey detecting. Hyper-spectral remote sensing coverage of the wavelength range of spectrum than the average width, number of spectral bands, high resolution, surface features can be compared to the spectral characteristics of fine curve, and may need to extract a specific band to highlight the characteristics of targets, so the research high-resolution remote sensing image matching techniques has important scientific significance and practical value.After decades of development, hyper-spectral remote sensing image data analysis and processing techniques have been huge advances in the traditional classification algorithm based on the formation of a series of development for hyper-spectral image classification algorithm. This article first outlines some spectral matching algorithm, and analyzes its advantages and disadvantages of conditions, due to the high spectral resolution hyper-spectral image, band many features, such as wavelength characteristics have led to wider recognition of the many difficulties, he traditional matching method in matching the high spectral image recognition because the image often caused huge amount of data calculation and processing complexity are greatly increased, leading to match the efficiency is greatly reduced, therefore, match the needs of high speed spectral images, effective, intelligent recognition processing method, this paper introduces the knowledge-based matching method meets these requirements.This paper introduces a knowledge-based hyper-spectral image matching methods, knowledge-based remote sensing image classification model is actually an expert system, knowledge-based remote sensing image classification process is a process of knowledge discovery, knowledge expression and knowledge interpretation, knowledge found dependent on the space that contains complete data, attribute data and spectral data. The proposed matching method based on knowledge of the spectrum is divided into four steps: define knowledge, entry rules, run the decision tree, classification process. the first we extract mineral absorption features of the spectrum under the expert knowledge of mineral, express these features in the form of rules, and then formulate the strategy for classification of these unknown samples using the rules, build and run the decision tree, obtain classification results and analyze them. Through practice to that of a knowledge-based methods for hyper-spectral data classification model is feasible and effective.