Review of Features Selection in Crop Classification Using Remote Sensing Data

Agriculture plays a key role in the national economy,and food production is important for the establishment of national and regional socio-economic development planning,to ensure national food security and social stability,and for guidance and control of macro cropping structure adjustment. Crop planting acreage is an important factor affecting food production and estimation of crop acreage using remote sensing data has become an important aspect of agriculture monitoring with remote sensing technique. Crop classification using remote sensing data is essential for improving crop acreage estimation accuracy. Features selection in crop classification of remote sensing data is an important step and effectively use of multiple features is especially significant for improving crop classification accuracy. Along with the facility in multi-source data acquiring,electromagnetic spectrum features,spatial features,temporal features and assistant data features are becoming more and more important in crop classification with remote sensing data. This paper briefly describes the characters and advantages of different features,including the multi-spectral feature,microwave scattering feature,multi-source data feature,hyperspectral data feature, spatial feature,temporal feature and assistant data feature. Furthermore,existing problems and developing trends in features selection for crop classification using remote sensing data are analyzed. It is indicated that weakness in theoretical research and limitation in combined application of features selection for crop classification of remote sensing data are two main problems. New features selection including biochemical component feature and canopy structure feature,integrated application of features of remote sensing data,sensitivity and uncertainty analysis of classification feature variables will be the three main issues of features selection for crop classification with remote sensing data in the future.