论文部分内容阅读
森林训练样本自动提取算法(TDA)已在Landsat图像分析中得到了成功应用,笔者以广西苍梧县广平镇为研究区,采用2007年ALOS、2011年Rapid Eye遥感图像,试验该算法在高分辨率图像中的应用。研究首先根据图像光谱特性自动识别出纯净森林训练样本,然后依据归一化的整合森林指数图像提取两期森林/非森林分类结果并以此进行林地变化检测,经过精度分析结果表明,面积总误差为-2.6%,空间位置精度为87.7%,说明该算法可有效地从高分辨率遥感图像提取出纯净的森林训练样本,为森林/非森林分类以及变化检测提供基础数据。
The automatic training algorithm for forest training samples (TDA) has been successfully applied in Landsat image analysis. Taking Guangping Town, Cangwu County, Guangxi as a study area, the author uses the 2007 Rapid Eye remote sensing imagery of ALOS and 2011 to test the algorithm in high Resolution Image in the application. The study first automatically identifies the pure forest training samples according to the spectral characteristics of the image, and then extracts two forest / non-forest classification results according to the normalized integrated forest index image to detect the forestland change. The precision analysis results show that the total area error Is -2.6%, and the spatial position accuracy is 87.7%, which shows that this algorithm can effectively extract pure forest training samples from high resolution remote sensing images and provide basic data for forest / non-forest classification and change detection.