• 研究论文 •

### “基于片层-面向类”的竹林信息提取算法与应用分析

1. （1．福建农林大学3S技术应用研究所，福建 福州 350002；2．福建农林大学林学院，福建 福州 350002；3．三明学院，福建 三明 365000）
• 收稿日期:2011-04-06 修回日期:1900-01-01 出版日期:2012-01-25 发布日期:2012-01-25
• 通讯作者: 刘健

### Algorithm and Application Analysis of FilmBased & ClassOriented for Bamboo Forest Information Extraction

YU Kunyong1,2, XU Zhanghua1，2, LIU Jian1,2,3,MIAO Lijuan1,2

1. (1.〖JP2〗Institute of Geomatics Application,Fujian Agriculture and Forestry University,Fuzhou 350002,China；〖JP〗2.College of Forestry, Fujian Agriculture and Forestry University, Fuzhou 350002, China；3.Sanming University, Sanming 365000, China)
• Received:2011-04-06 Revised:1900-01-01 Online:2012-01-25 Published:2012-01-25

Abstract:

As the continued excavation for economic benefits of bamboo resources, it has been a problem which is urgent need to solve to know the distribution and dynamic information of bamboo forest timely and accurately. However, the traditional field survey cannot meet this demand. The development of Geomatics technology has provided new approaches to seize the bamboo forest distribution information, but restricted by many factors, the extraction of bamboo information is not effective for quite a long time. On the basis of former studies, the paper took the highresolution of ALOS image as the data source and focuses on bamboo forest information extraction in Shunchang County, which is hilly area in Southern China, with the algorithm of Filmbased & Classoriented which has been put forward by us, in order to test the effectiveness of the method. In addition to using the traditional classification method of pixelbased, it introduced the objectoriented method that has been fashional in recent years, in order to compare them. What's more, it processed the nonfilm area further for higher bamboo forest information extraction accuracies. Verifying with the surveying points showed that the extraction accuracies with pixelbased and objectoriented were 84.42% and 86.46% respectively with Kappa of 0.677 3 and 0.723 4. Four results could be indicated. Firstly, both these two methods showed that the algorithm of Filmbased & Classoriented was well suitable for the bamboo forest information extraction, and could supply a significant technical reference to the image interpretation of hilly area. Secondly, from the accuracy, the method of objectoriented was a little better than the pixelbased one. Nevertheless, from the visual effect of view, there were spiced salt shapes in the figure of pixelbased, while that of objectoriented had better continuity. Although the results of these two methods were little different, the objectoriented one may be more suitable for the highresolution images such as ALOS image. Thirdly, it is hard to interpret and extract the bamboo forest information with the help of spectral features from remote sensed images single, so we have to bring in the texture information to raise the extraction accuracy. In addition to these, the terrain factors including elevation, slope and aspect are needed to regard carefully. We could infer that the bamboo forest information extraction accuracy will be raised more if the terrain combination conditions were taken to account.