基于优化FGFCM方法的滑坡遥感影像自动提取  

Automatic Extraction of Landslide from Remote Sensing Image Based on Optimized FGFCM Method

作　　者：谭秋焰 吴彩燕[1,2,3] 贾菊桃 朱新婷 廖军 TAN Qiuyan;WU Caiyan;JIA Jutao;ZHU Xinting;LIAO Jun(School of Environment and Resource,Southwest University of Science and Technology,Mianyang 621010,Sichuan,China;Tianfu Institute of Research and Innovation,Southwest University of Science and Technology,Chengdu 610299,Sichuan,China;Mianyang S&T City Division,the National Remote Sensing Center of China,Mianyang 621010,Sichuan,China;Sichuan Academy of Safety Science and Technology,Chengdu 610000,Sichuan,China)

机构地区：[1]西南科技大学环境与资源学院,四川绵阳621010 [2]西南科大四川天府新区创新研究院,成都610299 [3]国家遥感中心绵阳科技城分部,四川绵阳621010 [4]四川省安全科学技术研究院,成都610000

出　　处：《西南科技大学学报》2023年第1期54-60,共7页Journal of Southwest University of Science and Technology

基　　金：国家自然科学基金项目(41301587);四川省科技厅项目(2020YFS0389);第三次新疆综合科学考察项目(2021xjkk1404)。

摘　　要：为实现滑坡遥感影像的自动提取,以贵州省水城县为研究区,运用GF-1遥感影像数据,剔除明显非滑坡地物后,应用云变换计算样本区影像像元亮度值,统计得到聚类中心值和聚类个数,结合快速广义模糊C-均值聚类算法进行像元聚类,提取出128处准确滑坡,kappa系数达0.752,总体精度达0.880。该方法可自动提取出滑坡范围,减少工作量和主观影响,提高了基于遥感影像进行滑坡识别的效率和精度。In order to realize the automatic extraction of landslide from remote sensing images,Shuicheng County,Guizhou Province,was taken as the research area.Using the GF-1 remote sensing image data and removing the obvious non-landslide features,the image digital number in the sample area was calculated by cloud trasforation,and the cluster center value and cluster number were obtained statistically.And combined with the fast generalized fuzzy C-means clustering algorithm for pixel clustering,128 accurate landslides were extracted.The kappa coefficient was 0.752,and the overall accuracy was 0.880.This method can automatically extract the landslide range,reduce the workload and subjective influence,and improve the efficiency and accuracy of landslide recognition based on remote sensing images.

关 键 词：滑坡 遥感影像 快速广义模糊C-均值聚类算法 云变换 

分 类 号：P237[天文地球—摄影测量与遥感] P642.22[天文地球—测绘科学与技术]