An object-based information extraction technology for dominant tree species group types
Tian, T.; Fan, W.-y.; Lu, W.; Xiao, X.
Ying Yong Sheng Tai Xue Bao 26(6): 1665-1672
2015
ISSN/ISBN: 1001-9332 PMID: 26572016 Document Number: 681228
Information extraction for dominant tree group types is difficult in remote sensing image classification, howevers, the object-oriented classification method using high spatial resolution remote sensing data is a new method to realize the accurate type information extraction. In this paper, taking the Jiangle Forest Farm in Fujian Province as the research area, based on the Quickbird image data in 2013, the object-oriented method was adopted to identify the farmland, shrub-herbaceous plant, young afforested land, Pinus massoniana, Cunninghamia lanceolata and broad-leave tree types. Three types of classification factors including spectral, texture, and different vegetation indices were used to establish a class hierarchy. According to the different levels, membership functions and the decision tree classification rules were adopted. The results showed that the method based on the object-oriented method by using texture, spectrum and the vegetation indices achieved the classification accuracy of 91.3%, which was increased by 5.7% compared with that by only using the texture and spectrum.