Kharif dryland crop identification based on synthetic aperture radar in the North China Plain

During the key growth period of kharif dry-land crops in the north of China, due to the big impact of cloudy or rainy weather, it´s impossible to acquire optical remote sensing data in a timely and efficient manner. Therefore, it´s very necessary to use radar remote sensing to identify kharif dry-land crops. With Shenzhou City of Hubei Province as the study area, this paper has selected 6 sessions of Radarsat-2 fully polarimetric images which cover the area from June 3rd to Oct 1st in 2014 as the data source. Through the analysis of the backward-scattering characteristic of various ground objects, we found that cross-polarization channels had a better performance than like-polarization in identifying dry-land crops. Also, we put forward the optimum polarization and phase for identifying dry-land crops with the support vector machine (SVM) classification accuracy and Jeffries-Matusita (J-M) distance as the standard. We conducted identification of 5 major ground objects in the study area with the decision tree classifier (DTC) and the SVM method. The result suggests that radar data can be effectively applied in identifying dry-land crops, SVM is superior to DTC in identifying dry-land crops and SVM has a distinct advantage in identifying small areas and controlling speckle noise.