【摘要】 南方人工林生长迅速,轮伐期短。为探讨一种有效更新森林资源数据库的森林变化检测方法,快速检测短时期内森林采伐与更新的动态变化。以变化频繁快速,变化图斑多且小的广西人工林作为研究区,以2个时相的高分二号(GF-2)影像为数据源,利用多尺度分割和光谱差异分割2种方法对2期影像进行分割,通过对象的归一化差值植被指数(normalized difference vegetation index,NDVI)差值并基于分布函数确定阈值来提取变化区域与变化类型,实现森林变化快速检测。基于像元采用同样的方法进行处理,与面向对象NDVI差值法进行比较。结果表明面向对象NDVI差值法的总体精度达89. 76%,Kappa系数为0. 81,精度和提取效果优于基于像元NDVI差值法,更能刻画变化图斑的形状和边界,也能较准确地检测出微小变化的面积。该方法能适应南方人工林的变化特点,在实现快速检测变化的目的下,可用于森林资源数据库的更新。更多还原

【Abstract】 Plantation in southern China is growing rapidly,and rotation cutting period is short. To explore the forest change detection method used to update the forest resource database effectively and to monitor the dynamic changes in forest harvesting and renewal in a short period,the authors chose the plantation area of Shangsi County in Guangxi as the study area,where the plantation area changes frequently and rapidly and the change patterns are numerous and small. The GF-2 remote sensing images of two phases were used as data sources. Multi-scale segmentation and spectral difference segmentation were used to segment the two-phase images. The change areas and change types were extracted from the NDVI difference of the objects and the threshold value was determined based on the distribution function,so as to realize the rapid detection of forest change. In addition,the same method was adopted for pixel-based processing in comparison with object-oriented NDVI difference method. The results show that the overall accuracy of the object-oriented NDVI difference method is 87. 12%,and the Kappa coefficient is 0. 81. The accuracy and extraction effect are better than those of the pixel-based NDVI difference method,indicating that the object-oriented NDVI difference method can better depict the shape and boundary of the change spots and can also more accurately detect the small change area. This method can be adapted to detect the changing characteristics of plantation in south China and can also be used to update the forest resource database for the purpose of rapid change detection.更多还原