Abstract Views :132 |
PDF Views:1
Authors
Affiliations
1 College of Hydraulic and Electrical Engineering, Heilongjiang University, Harbin 150086, CN
2 College of Information and Computer Engineering, Northeast Forestry University, Harbin 150040, CN
3 School of Animal and Veterinary Sciences, The University of Adelaide, Roseworthy 5371, AU
Source
Nature Environment and Pollution Technology, Vol 15, No 1 (2016), Pagination: 243-248
Abstract
Land use cover (LUC) classification is one of the most important applications of optical remotely sensed data, while LUC mapping outcomes are used for global, local mapping, ecosystem assessment and environmental process monitoring. Hence, in this study, in order to evaluate the advantages and drawbacks of supervised classification schemes, the paper chose the optical image data of Landsat 8 OLI in Miyun county to test supervised classification and introduced Parallelepiped Method (PM), Minimum Distance (MD), Maximum Likelihood Classifier (MLC) and Support Vector Machines (SVMs) to improve classification accuracy of LUC mapping and to obtain the reliable LUC distribution. The four classified images reveal that the study area is dominated by considerable areas of forest land, with the overall accuracy found to be 87.89% (kappa = 0.8524) using SVMs, 85.26% (kappa = 0.8205) using MLC, 82.11% (kappa = 0.7813) using MD, and 74.74% (kappa = 0.6920) using PM. Based on the overall accuracy and kappa statistics, SVMs might be the first option in terms of classification accuracy without taking into account of the time costly and standard PC and laptops. MLC was the second accurate model classifiers from the classified image, which was always used to obtain LUC map information for economic potential in time and cost; and PM has shown the lowest overall classification accuracy with greater omission errors and commission errors.
Keywords
Land Use Cover, Image Classification, Support Vector Machines, Maximum Likelihood Classifier.
Full Text
