Dryland Crop Classification And Acreage Estimation Based On Microwave Remote Sensing

Timely and accurately crop type identification and area monitoring are the basis for agriculturalmonitoring. Because of the ability of all-day and all-weather to monitor land surface, SAR remotesensing is widely used in the field of crop remote sensing monitoring. There have been many domesticand international studies focus on the rice acreage estimation and growth monitoring by SAR images.The studies on dryland crop monitoring by SAR images are not so many. In the operational crop remotesensing monitoring system of the Ministry Department of Agriculture, it is difficult to obtain sufficientquantities and effective optical remote sensing data because of adverse cloudy and rainy weather duringthe critical period of dryland crop growth even in the north China. In order to support for the researchesand applications, in this paper, study area was set up in the North China Plain, to study the dryland cropidentification and crop acreage monitoring techniques using SAR remote sensing data.In this paper, multi-temporal and multi-polarization SAR data was used to classify dryland croptype and estimate acreage in Zaoqiang, Hengshui city of Hebei Province.. Firstly, the crop parameterdata were collected, like biomass, water content, leaf area index, plant height on3crop phonologicalstages in2012, which were synchronous with SAR imagery acquisition. Then, the statisticalcharacteristic and temporal changes of these parameters were analyzed. Secondly, based onRADARSAT-2data, the author analyzed the response characteristics between crops and radarbackscattering coefficient and also compared the differences of response characteristics of corn andcotton. Finally, the crop identification methods were studied based on RADARSAT-2remote sensingdata, and the integration of the optical remote sensing data from ZY-3and SAR remote sensing data toenhance crop monitoring accuracy was also studied.Based on the research and experiments in this study, the main conclusions are as follows:(1) Based on the study of correlation analysis of crop parameters and backward scatteringcoefficient, it can be concluded that the backward scattering coefficient was sensitive to crop height andcrop water content. In different crop growth stages, the two crops of corn and cotton have differentresponse mechanism to backward scattering coefficient. This provides the theoretical basis for usingmulti-temporal and multi-polarization SAR data to identify different crops in the study region.(2) As to the classification effect of different SAR polarizations, the crop classification by VVpolarization data is higher than that by HH polarization, with either of them is better than that for thecross-polarizations. As to the classification effect of different temporal data, the accuracy for the data onthe September24image is best. Crop classification accuracy will increase with the increasing in thenumber of data sets of different SAR polarizations and temporal phases. The crop classificationperformances for the HV and VH cross polarization data are similar., If both of them were included wayin the classification, the classification accuracy will decrease. The classification accuracy for theminimum distance classification method is better than that for support vector machine (SVM) method.In this study, the overall crop classification accuracy is85%, using this minimum distance classification method to classify the dryland crops based on multi-temporal and multi-polarization RADARSAT-2data.(3)Due to the negative influence of building to classification accuracy, in this paper, the buildingarea was extracted from ZY-3remote sensing data as the mask in SAR data. Without the negativeinfluence from buildings, the crop classification accuracy for SAR data is93%, as this result is close tothat from ZY-3optical remote sensing data, where it is94%.