改进ICA算法及其在作物光谱分类中的应用

Improved Independent Component Analysis Algorithm and Its Application on Crop Spectral Classification

针对传统的固定点算法对分离矩阵初始值敏感的问题,提出一种改进的独立分量分析(ICA)算法,通过在传统的算法核心迭代过程中加入搜索因子,降低算法对矩阵初始值的依赖,提高处理效率。将ICA算法应用于作物精细光谱的分类,分别利用传统固定点算法和改进的固定点算法对混合光谱进行信息提取与分离。实验证明,改进的ICA算法在与传统算法作物光谱分类效果相当的情况下,迭代次数减少26%,提高了独立分量的分离效率,是一种有效的作物光谱分类方法。

Aiming at the issue of the traditional fixed-point algorithm＇s sensitivity on the initial value of the separation matrix,an improved Independent Component Analysis（ICA） algorithm is proposed.The algorithm has a low dependence on the initial value by referring the search factor in the core iterative process of the traditional algorithm,and the method improves the efficiency of the algorithm.The mixed spectral information is separated by using traditional fixed-point algorithm and improved fixed-point algorithm.Experiments show that the improved ICA algorithm decreases iterations by 26% with the considerable spectral classification effect to the traditional algorithm,and improves the separation efficiency of independent component,thus it is an effective crop spectral classification method.