改进型遗传算法和支持向量机的波段选择研究

Research of Band Selection Based on Improved Genetic Algorithm and Support Vector Machine to Select

回顾了传统高光谱图像的特征选择算法所存在的问题,将支持向量机分类器和自适应小生境遗传算法引入到高光谱图像的特征选择上,提出了一种高光谱图像特征选择算法,该算法充分考虑了高光谱图像的波段相关性,同时克服了传统遗传算法的早熟问题和支持向量机的参数选择问题,以微小的代价,大大地压缩了光谱矢量的特征维数,达到令人满意的分类效果。

This paper analyses the problems of the traditional Hyperspectral Image Feature Selection Algorithm. Using support vector machine classifier and adaptive niche genetic algorithm into the characteristics of hyperspectral image selection, it propose a new hyperspectral image feature selection algorithm. The algorithm takes into account the band hyperspectral image correlation, while overcoming the early problems that traditional genetic algorithm caused, the cost of the parameters of support vector machine selection. The method greatly compresses spectral characteristics of the vector dimension, and achieves satisfactory results.