Machine Learning
| Hi-index | 0.00 |
Large area land cover mapping, involving large volumes of data, are becoming more prevalent in remote sensing applications. Thus there is a pressing need for increased automation in the land cover mapping process. The main objective of this research was to map land cover in the Small Sanjiang Plain where marsh distributed concentively combined Landsat TM imagery with ancillary geographical data and compare the performance of three machine learning algorithms (MLAs) including random foerst (RF), classification and regression tree (CART) and maximum liklihood classification (MLC). Comparisions were based on several criteria: overall accuracy, sensitivity to data set size and noise. Our results indicated that (1) Random Forest can achieve substantial improvements in accuracy over single classification trees and traditional MLC method, overall accuracy was 91.0%, kappa coefficient was 0.8943, with marsh class accuracy ranging from 77.4% to 90.0%; (2) Random forest was least sensitive to reduction in training sample size and it was most resistant to the presence of noise compared to CART and MLC. The comparison result revealed that random forest has potential to increase automation in large area land cover mapping while achieving reasonable map accuracy.