A comparison of support vector machines and manual change detection for land-cover map updating in Massachusetts, USA

Abstract

The remote sensing community has recently adopted land-cover map updating methodologies using spectral image differencing, change masking and concatenation procedures to monitor land change accurately and consistently. Unfortunately, map updating requires costly, time-consuming manual image interpretation to achieve accurate spectral threshold placement for land-change masking. The purpose of this study is to minimize time and costs associated with manual image interpretation of change thresholds by developing a new, semi-automated method using support vector machines (SVM). The results of this study show that the SVM change detection method produced more accurate results and required considerably less time and user effort than the manual change detection method, and is thus an effective alternative to manual methods of land-cover map updating.

Acknowledgements

This study is based on work supported by the National Science Foundation (NSF) under grant No. SES-0849985 (REU Site) and by the Clark University O'Connor ’78 Endowment.