基于二维小波变换的遥感分类研究

对于遥感影像的植被特征直接进行计算机智能提取、分类,不同的理论下有着不同的方法。但是由于受到各种因素的影响,传统方法的提取精度还不是很高。在对研究区内的试验区图像运用小波理论,对图像进行二维小波变换处理并提取NDVI的基础上,再进行分类,将得到的结果与直接进行计算机自动分类结果相比较,得到如下结论:①经过小波变换后的图像,其相同地物内部的土壤亮度噪声得到削弱,不同地物的边缘差异得到增强,有利于进行图像的计算机智能分类;②对图像进行小波变换后试验区植被特征的分类精度明显高于没有经过小波变化的图像分类精度;③由于小波函数的多样性,小波系数的不确定性,二维小波变换用于遥感图像的植被特征分类技术还不是很成熟,这种技术还有待进一步改进。

A Study on the TM Classification of Vegetation Feature Based on Two-Dimensional Wavelet Transformation

Directly intellective selection and classification of the vegetation feature in the TM have different methods of dealing with remote sensing. But the presicion of selection by traditional methods is not high because they are influenced by many factors. The wavelet is a new method of image manipulation which is developing in two or three years. In this paper, the image is transformed by the two-dimensional Wavelet and transformed by NDVI, finally classicified by unsupervised. The result is compared to ones which is not transformed by wavelet. The four conclusions are given. Firstly, the image which is transformed by wavelet obviously strengths the edge discrepancy of different sorts of vegetation. Secondly, the presicion of classification of the transformed image is higher than that of raw image. Thirdly, after image transformed by wavelet, the results of classification relatively accord with the facts. It can be spread extently and have a value of application. Lastly, even though the method which the image is transformed by wavelet and classified offers a new approach of selecting image feature, but it is not maturely because the fuctions and coefficient of wavelet are not confirmed. So the method will be more improved.