Investigation Of Hyperspectral Remote Sensing Images Classification Using Multiple Classifiers Combination

The advent of hyperspectral was one of the most significant breakthroughs in remote sensing. Hyperspectral remote sensing has higher spectral resolution as the same time retain higher spatial resolution, so its capability of distinguishing the different and describing the same ground objects in details enhanced greatly. Hyperspectral imagery generally consists of dozens or hundreds of narrow, also contiguous, spectral bands, which accounts for the computational problem and the phenomenon where the response of bands tends to be highly correlated. This paper aim to enhance the performance of land cover classification, which is one of the most prominent applications of hyperspectral remote sensing. Relative studies demonstrate that classification accuracy depends on dimensionality, the continuity of the image space and the performance of classifier, the main conclusions are as follows:1. In order to solve the disaster of dimension in the hyperspectral data, taking into account high information and low correlation of bands ,use an adaptive band selection (ABS) of the reduced- dimension method.2. Research and realization of the ECHO algorithm. The algorithm makes full use of the continuity of the objects in the space , first cut hyperspectral image into blocks, and then classified, are usually improved the defect of in the sense of classified information only to consider the spectrum.3. Research and application of a combination classification of high spectral classification. It has been demonstrated that different classifiers potentially offer complementary information about the pattern to be classified. This implies that in hyperspectral classification, a reasonable ensemble of classifiers can be beneficial to obtaining more accurate classification results. Based on analyzing the traditional combining schemes, a novel combing scheme, which is named hybrid combining scheme, is put forward. The proposed combining scheme possesses the virtues of both serial combination and parallel combination, so it is very efficient and effective in hyperspectral classification. Meanwhile the hybrid combination is very extensive due to the fact that it is based on the output of measurement level. Experiments show that the combination classifiers in the application of hyperspectral data classification with the classification of superior performance can be better classification results, and has a strong adaptability.