Multitemporal SAR images for monitoring cultivation systems using case-based reasoning

This paper demonstrates that multitemporal satellite SAR images are most suitable for monitoring the rapid changes of cultivation systems in a subtropical region. A new method is proposed by applying case-based reasoning (CBR) techniques to the classification of SAR images. Stratified sampling is carried out to collect the cases so that the variations of backscatters within a class can be appropriately captured. The use of discrete cases can conveniently represent the internal changes of a class under complicated situations, such as spatial changes in soil conditions and terrain features. These spatial variations are difficult to represent by using rules or mathematical equations. The proposed method has better classification performance than supervised classification methods in the study area. The case library is reusable for time-independent classification when the SAR images are acquired at the same time of the crop growth cycles for different years. The proposed method has been tested in the Pearl River Delta in South China.