融合光谱与地形因子的土地利用/覆被变化检测

A Land Use/Land Cover Change Detection Method Integrated Spectrum and Terrain Factors

针对仅利用遥感影像单一光谱指数提取土地利用/覆盖变化信息精度较低的问题,提出融合光谱与地形因子的土地利用/覆盖综合变化检测方法,用于提高总体变化检测精度。首先利用互补的4个光谱检测指标获得尽可能消除季相差异误差的初始变化检测结果,然后根据坡度、坡向剔除因山体阴影造成的虚检部分。以江西省西南地区Landsat TM影像为例,对比了融合光谱和地形因子的综合变化检测方法与仅利用光谱信息的分类后变化检测方法的结果,表明前者的变化检测精度比后者提高了13.67%,且虚检率降低了15.56%,这说明融合光谱和地形因子的综合变化检测方法能够显著提高遥感变化检测精度。

At present,the method of using the difference or ratio of the spectral band of different phase images to gain change information has been widely used.But the detection result which is only detected by single band or single index tends to have a lot of commission errors and omission errors,and the precision is low.Previous studies show that the spectral differences of vegetation between summer and winter lead to more commission errors and mountain shadow can have a large impact on detection result in mountain region.Therefore,this paper puts forward the comprehensive Land Use/Land Cover change detection method to improve the overall change detection accuracy which integrates spectrum and terrain factors.This method uses four complementary spectral indices to capture more comprehensive change information,including both change spatial location and change direction（biomass increase and biomass decrease）.At the same time,the method uses slope and aspect to weed out commission errors caused by terrain.This research,using Landsat TM images in southwest of Jiangxi province as an example,compares the change detection results between the comprehensive Land Use/Land Cover change detection method integrated spectrum and terrain factors and classification change detection method.The result shows that this method improves precision by 13.67%than maximum likelihood classification and reduces the commission rate by 15.56%,which fully indicates that the method this paper proposed can improve the accuracy of remote sensing change detection significantly.