摘要
在陕北黄土丘陵沟壑区,采用单一传感器的遥感影像提取土地利用信息,存在着识别的土地利用类别少、某些类别混分现象较严重、分类结果的精度较低等问题。以TM多光谱数据和SPOT全色光谱数据的融合为例,提出了适宜于该地区的两种影像融合方法:主成分变换法和乘积运算法,并从影像的光谱质量、纹理信息和目视效果等方面对其进行了对比与评价。结果显示,主成分变换法为较理想的融合方法。以陕北无定河流域为实验样区的土地利用自动分类结果表明,该方法的应用使土地利用各类别的提取精度都有不同程度的提高;水体、水田和城镇用地等面积较小的类别分类正确率提高达到10%以上;坡耕地与林草地的混分明显减少,分类精度均提高了5%以上;分类总精度从82.0%提高到89.2%,取得了良好的分类效果。此研究对于遥感影像融合技术的评价与应用进行了有益的探索,同时为该地区的土地利用动态监测提供了关键技术。
There are many limitations in extracting land use categories based on single remotely sensed data in the Loess Hill and Gully Area of Northern Shaanxi. Few categories can be recognized; some categories mixture in the classification so that the accuracy is spoiled. Taking the fusion of riM multi-spectral data and SPOT pan data for a case study, the paper discussed two fusion methods suitable for the region： Principal Component Merge, and Multiplicative. Furthermore, the two methods were compared in terms of spectral quality, spatial texture and visual effects. Experimental results show that PC Merge is the best among the all methods related. The classification result at a watershed test area proves that the total accuracy increased from 82.0% to 89.2%, especially in extracting city and town, paddy field and water area where the classification accuracy increased over 10%, the mixture of sloping field and forest （grassland） in classification decreased remarkably and the accuracy of the two categories increased over 5%, respectively. This research plays an effective promotion in evaluating and applying image fusion method of various kinds and is of critical significance in land use dynamic detection in the region.
出处
《中国农学通报》
CSCD
2006年第1期361-365,共5页
Chinese Agricultural Science Bulletin
基金
中国科学院知识创新重要方向项目"黄土高原水土保持的区域环境效应研究"(KZCX3-SW-421)。
关键词
影像融合
土地利用动态监测
分类
陕北黄土丘陵沟壑区
Image fusion, Land use dynamic detection, Classification, Loess hill and gully area of Northern Shaanxi
