地性线提高卫星林火监测几何精度的方法

A Method of Improving Geometric Accuracy of Forest Fire Monitoring by Satellite Imagery Based on Terrain Features

利用卫星热点监测森林火灾,在我国已有近20年的应用历史,但提高火场或火点定位精度的一直没有大的突破,经过系统级及GCP控制校正的监测图像,往往还存在2-4个像元的误差,影响了开展地面核查和处置的工作效率。针对以上问题,提出了一种基于90m DEM（数字高程模型）提取地性线,并集成立体可视地形地貌图、江河及大型湖泊矢量地图,以此为地理参照数据控制校正MODIS林火监测图像的方法。实验结果表明,与原有的几何纠正技术相比,利用地性线控制精校正的监测图像,其火场的定位误差从2-4个像元提高到1个像元内,GCP（控制点）的均方误差仅为431m,该技术方法能极大地提高卫星林火监测的几何精度。

Taking advantage of hot spots to monitor forest fires has the application history of nearly 20 years in China. However,after the systematic and GCP correction, it still has a positioning error of 2-4 pixels,which affects the working efficiency seriously in ground-truthing. Aiming at this issue, a new method was proposed, which was based on 90 meter DEM terrain features and integrated stereoscopic visual topographic map and vector map of great rivers and large lakes,and can get the geography comparable data to control and correct the MODIS satellite images to monitor forest fires. The experimental results illustrate that the distance deviation error can be reduced from 2-4 pixels to within 1 pixel,and the mean square error of GCP is only 431 meters,which proves the effectiveness of the proposed method.