基于自动搜索和光谱匹配技术的训练样本纯化算法

提出了一种基于局部自动搜索和光谱匹配技术的监督分类训练样本的纯化方法。该方法首先利用遥感影像中像元的灰度信息在图像上局部范围内自动搜索和选择最佳样区位置，然后利用光谱匹配的思想对寻找到的最佳样区在光谱空间上进一步纯化。实验结果证明，通过手工选择样区的辅助，该算法能够自动有效地搜寻到最佳样区的位置，并对最佳样区进行纯化处理。原始遥感图像经过本文的样区纯化算法处理后，无论是目视判读效果，还是分类后混淆矩阵的统计及分类精度，均优于纯化处理前的分类结果，具有一定的实用价值。

Purified Algorithm for Training Samples Based on Automatic Searching and Spectral Matching Technique

A purified algorithm for training samples based on local automatic searching and spectral matching technique is presented.In the first step,the optimal training sites are searched and selected automatically by taking advantage of local spatial information in remote sensed images.The selected training sites are then purified further in spectral space using spectral matching technique.The proposed algorithm is capable of purifying training samples with spatial and spectral information.Experimental results are given to show that supervised classification with the proposed purification algorithm has superiority capability over the traditional supervised classification without purifying training sites on visual judgment and accuracy assessment.