基于数据融合的土地遥感分类技术研究

以吐鲁番主要农区为试验区，运用多源遥感数据融合图像进行土地利用现状调查及葡萄种植面积遥感调查。运用IHS变换方法进行了研究区多光谱与全色高分辨率图像的融合工作，生成了质量较好的图像。用于土地分类及葡萄面积提取。在对研究区土地利用类型的分类中，本文采用的分类方法是结合目视判读的经验，在无监分类法和有监分类法的基础上，辅以地面数据，在遥感图像处理常规分类方法的范畴内，采取最大程度的有利于提高分类精度的措施，得到了全试验区相对最高的计算机自动分类精度65％；然后采取分区分类或人工解译改进分类结果，可使研究区的土地利用分类精度超过90％。因此，此方法在土地利用现状调查及作物面积监测中是可行的。

Research on Technologies of land Classification Based on Multi-source Remote Sensing Data Fusion

In this paper,the part of Tulufan region is the experimental area.The classification method based on fused images and an application in surveying land-use types are discussed. IHS transformation is used to integrate TM multispectral image and high spatial resolution panchromatic image.Its signal clarity was improved. we use the better classification method to classify the land-use type.This method was produced by integrating unsupervised classification and supervised classification procedures before image interpretation.By taking all measures that may make the accuracy higher within the framework of conventional image classication methods, a result for the upper limit ,65%. Then classification in every small region or the human-computer interaction interpretation is used to improve classification result.The last accuracy of fused remote sensing image can be up to 90%.Therefore,the result of this study turned out that it is feasible to apply this classification method in investigating land-use types.