Crop Classification Based on Image Segmentation and Phenological Similarity Using SAR Imagery

Abstract

Crop yield is a key factor in agricultural production management and agricultural policy formulation, which affects the stability of the country and society. Therefore, accessing the spatial distribution of crops in real time is very important. In this study, multi-temporal RADARSAT-2 fine beam quad-polarized SAR data was obtained, and a crop classification method based on similarity analysis of phenological features and image segmentation technology was proposed. The main idea of this method is to construct a standard phenological feature sequence (SPFS) based on backscatter coefficient and Cloude decomposition parameters for each crop by using the training data, and then the similarity coefficient between the segmented test data and the phenological sequence of each crop is calculated to judge the type of ground objects. The results show that the overall classification accuracy based on block scale reached to 76.06%, which indicates that the phenological features and image segmentation is beneficial for accuracy improving in crop classification. And the multi-temporal SAR data own great potential in agricultural monitoring.