| Home  | About ScienceAsia  | Publication charge  | Advertise with us  | Subscription for printed version  | Contact us  
Editorial Board
Journal Policy
Instructions for Authors
Online submission
Author Login
Reviewer Login
Volume 49 Number 1
Volume 48 Number 6
Volume 48 Number 5
Volume 48 Number 4
Volume 48 Number 3
Volume 48 Number 2
Earlier issues
Volume 42 Number 3 Volume 42 Number 4 Volume 42 Number 5

previous article next article

Research articles

ScienceAsia 42 (2016): 283-289 |doi: 10.2306/scienceasia1513-1874.2016.42.283

Integration of multi-classifiers in object-based methods for forest classification in the Loess plateau, China

Pengxiang Zhaoa, Jun Zhaoa,*, Jianghua Wub, Yanzheng Yanga,c, Wei Xuea, Yichen Houa

ABSTRACT:     The object-oriented method with three integrated different classifiers was applied to classify satellite images of the forest in the Loess Plateau in China. After image segmentation, feature selection, and training sample selection, three classifiers–-the support vector machine (SVM), k-nearest neighbour algorithm, and classification and regression tree–-were used for forest classification using SPOT images as the data source. Results indicated that the object-oriented method with the three classifiers effectively extracted Chinese pine forestland, Betula forestland, oak forestland, shrubland, wasteland, farmland, and roads in the study area. The main segmentation parameters of scale, colour, and shape performed best when their values were set to 100, 0.9, 0.1 in forestland and 60, 0.5, 0.5 in non-forestland, respectively. In addition, SVM was the best classifier applied to the forest classification with an overall accuracy of 78% and a kappa coefficient of 0.737. This study provides a fast and flexible approach to forest classification and lays the foundation for forest management and forest resource surveys.

Download PDF

24 Downloads 1206 Views

a College of Forestry, Northwest A&F University, Taicheng 3 Road, Yangling 712100 China
b Sustainable Resource Management, Grenfell Campus, Memorial University of Newfoundland, Corner Brook, NL, A2H6P9 Canada
c Centre for Earth System Science, Tsinghua University, 100084, Beijing, China

* Corresponding author, E-mail: zhaojun148@126.com

Received 22 Nov 2014, Accepted 0 0000