Mapping rice planting area from Landsat 8 imagery using one-class support vector machine

Timely and accurate estimation of rice planting area would greatly optimize our prediction of rice production, which provides invaluable information for government in formulating policies with regard to national food security. Previous studies have shown great potential of optical remote sensing as an effective way to map rice planting area. Commonly used classification techniques, which mainly focus on multi-class classification, have been successfully applied in many cases. However, multi-class classifiers require all classes that occur in a study area to be labeled and sampled, which also means redundant training sets, high time and labor cost. In this study, we propose to use one-class support vector machine (OCSVM) as a classifier for identifying the rice planting area in Jiangsu, China with Landsat Optical Land Imager (OLI) imagery. An evaluation of the rice planting area shows an overall accuracy of 92% compared to validation dataset. The mapping approach has the potential for efficient and accurate mapping of rice planting area with Landsat imagery at the regional level.