运用多时相直接分类法对土地利用进行遥感动态监测的研究

采用多时相直接分类法对贵州省 (原 )安顺市从1987～1995年间的土地利用变化状况进行了遥感动态监测。与以往方法相比 ,该方法不仅可以直接获得变化类型及其数量 ,而且可以避免分类后比较法所容易出现的逻辑性变化探测错误。对不同分类波段组合的试验表明 ,经过差值、比值处理的前3种波段组合 ,具有较差的分类效果 (总体精度只有30 %～40 % ) ,而那些能较好地保留原始信息的波段组合 ,则具有相对较好的分类效果 (总体精度超过70 % )。另外 ,分类效果还明显地受到所采用的辐射校正方法的影响 ,经过暗组亮组辐射校正的影像要明显比只经过简单辐射校正(直方图法 )的影像具有更好的分类效果 ,总体精度提高了16.6 %。该方法的应用结果表明 ,研究区在8年间经历了一些明显的土地利用变化 ,例如城镇扩展、耕地减少、菜地变迁等。

The research on land use change detection by using direct classification of stacked multitemporal TM images

The land use changes in Anshun City,Guizhou Province from 1987 to1995 were detected by using direct classification of stacked multitemporal TM images. Compared with other methods,such as postclassification comparison and pixel to pixel comparison methods,the change detection algorithm used herein can not only bypass the illogical error,but also obtain the “ from to” information directly.The results showed the difference or ratio bands had poor classification accuracy(overall accuracy equal approximately to 30％ ～ 40％ ),while those bands could well inherited the original ones achieved better accuracy(overall accuracy>70％ ).In addition,the different radiometric rectification methods could affect the accuracy of change detection to a large degree,the preliminary research indicated that the dark set bright set method might improve the accuracy by 16.6％ in comparison with the histogram algorithm.Also,our results indicated that the study area had experienced some notable land use changes during 8 years,such as urban expansion,cultivated land encroachment and spatial change of vegetable land.