一种构造系数的自相关函数特征提取算法

A Feature Extraction Algorithm Based on Constructed Weighted Coefficient and Autocorrelation Function

由于遥感影像具有数据量大、维数高和不确定性等特点,遥感影像的分类已经远远超出了人的分析和解译能力,为了达到理想的分类效果,提取深层次空间结构信息的需求越来越强烈。根据各类样本的均值和方差构造加权系数,对样本的自相关函数进行加权,提出1种新的自相关函数特征提取算法,以改善样本不足造成的分类精度较低问题;采用支持向量机方法,对新的样本数据进行训练与分类性能研究。实验结果表明分类精度提高,在一定程度上能够反映遥感影像的深层次空间结构信息,验证了此算法的有效性与可行性。

Remote sensing image features huge data, high dimension and uncertainty. And remote sensing image classification has gone beyond our analysis and interpretation ability. To reach ideal classification results, demand of deep spatial feature extraction is extremely urgent. Based on the idea of SVM, a new approach based on autocorrelation feature extraction and constructed weighted coefficient has been proposed in this paper. New sample is created by combining autocorrelation function feature and sample feature. This approach analyzes classification result based on new sample. Experiment results show that the classification accuracy is increased and spatial feature of remote sensing image can be reflected to some extent. This verifies the effectiveness and feasibility of this approach.