论文部分内容阅读
植被检测是利用近景影像进行滑坡变形监测应用的一项关键步骤,滑坡近景影像中植被信息的特殊性给自动化匹配带来困难。针对经过高精度影像配准和点云滤波等处理后仍然存在形变区域检测误差,通过叠加滑坡体形变前后的数字表面模型和植被检测结果,实现对误检测形变区域的定位和分析。实验结果表明,形变区域检测误差主要来自植被剔除残余、点云滤波或点云模型不精确和数字高程模型采样误差等。
Vegetation detection is a key step in landslide deformation monitoring using close-range images. The particularity of vegetation information in near-landslide images makes it difficult to automate matching. Aiming at the error of detection of deformation area after high-precision image registration and point cloud filtering, the digital surface model and vegetation detection results before and after landslide deformation are superimposed to locate and analyze the false detection deformation area. The experimental results show that the detection errors in the deformation area are mainly caused by the vegetation removals, inaccuracies of point cloud filtering or point cloud model and sampling errors of the digital elevation model.