Science Service System

Summary of Proposal MTH2842

TitleCrop mapping with multi-temporal polarimetric TerraSAR-X data
Investigator shuhua, yan - Nanjing University, School of Geosciences
Team Member
Dr. yan, shuhua - Nanjing University, School of Geosciences
professor yan, jun - Nanjing University, School of Geosciences
Doc. yu, jianhong - Nanjing University, School of Geosciences
Doc. xu, ping - Nanjing University, School of Geosciences
master zhao, li - Nanjing University, School of Geosciences
Summary

Challenges exist in balancing national, regional and global food supplies with the increasing demand from a growing population. The world price of wheat, rice, corn, and soybean rose 226% from 2002 to 2008. The World Bank has quantified this impact and estimates that continued price rises since 2010 have resulted in a net increase in extreme poverty of about 44 million people. To meet these changing global flood requirements , the UN Food and FAO estimates that food production must double in the next 40 years. The development of sound policies and risk management strategies will aid in providing timely and accurate production estimates and forecast. Crop production estimates from TerraSAR-X data can make a contribution to monitoring food security at regional and global scales.


The radar response to the vegetation structure is polarization dependent, and crop classification accuracies could be improved when dual-polarization TerrraSAR-X data were employed. Analysis of PolSAR data demonstrate that polarimetric decomposition parameters provide overall crop classification accuracies equal to or better than accuracies achieved using linear dual-polarization data. Pixel-based classification approach has been developed with SAR. With this kinds of method, each pixel is individually assigned to a designated class and the resulting maps are often very noise due to high spatial variance in the landscape conditions. Moreover, the coherent nature of SARs result in noise in the data, reducing accuracies derived from pixel-based classification algorithms. An alternative approach, the object-based classification, first merges pixels into objects which are subsequently classified. This method has been successfully applied to high resolution optical imagery. For SAR imagery this approach would reduce radar speckle effects prior to classification.


In addition, object-based crop classification produce a polygon-based classification map which would more easily support trend analysis for crop monitoring. To study the object-based classification for SAR crop mapping and monitoring, TerraSAR-X polarimetric image was exploited in Tongling city in Anhui.

Back to list of proposals

DLR 2004-2016