ALOS PALSAR散射总功率的面向对象林火区提取

Extraction of the forest fire region based on the span of ALOS PALSAR by object-oriented analysis

利用合成孔径雷达(synthetic aperture Radar,SAR)开展林地过火区提取主要基于单个像元开展,而以像元集合为分析单元的面向对象技术在该领域的应用研究还较少。本研究基于ALOS PALSAR影像的散射总功率数据,采用分型网络演化的多尺度分割算法提取2009年发生在美国Alaska中东部的林火灾后的过火区,并将林火区提取结果与获取的燃烧强度辅助数据进行对比,验证算法应用的有效性。研究表明,相比之前基于像元的阈值分类方法,基于本文方法火后单一时相和林火前后2时相SAR散射总功率的林火区提取精度分别提高了12.7%和15.8%。有效证明了面向对象技术能够有效应用于SAR影像的信息提取,SAR卫星遥感在森林火灾监测方面具有较好的应用潜力。

At present,the extraction accuracy of the forest fire area by synthetic aperture Radar(SAR)is mainly limited to the analysis of single pixel.However,the application study of object-oriented technology based on pixel sets as the analysis unit is less in dealing with SAR images.In this paper,a multi-scale segmentation algorithm based on fractal net evolution approach(FNEA)was applied to the span of ALOS PALSAR images.Through the application research,the forest fire region,which happened in 2009 and was located in the Middle East of Alaska,USA,was extracted.The application validation of the algorithm was verified by comparing the experiment results with the auxiliary data of monitoring trends in burn severity(MTBS)data.The experiment results show that the classification accuracies of one-static span and two-static spans based on object-oriented analysis are improved by 12.7%and 15.8%respectively,compared with precious research.The researches show that object-oriented technology can be effectively applied to the information extraction form SAR image,and SAR technology has potential application in forest fire monitoring.