Integration of Multispectral and C-Band SAR Data for Crop Classification

Integration of Multispectral and C-Band SAR Data for Crop Classification

Iannini, L.
Molijn, R.A.
Hanssen, R.F.

Civil Engineering and Geosciences

Geoscience & Remote Sensing

2013-12-31

The paper debates the impact of sensor configuration diversity on the crop classification performance. More specifically, the analysis accounts for multi-temporal and polarimetric C-Band SAR information used individually and in synergy with Multispectral imagery. The dataset used for the investigation comprises several multi-angle Radarsat-2 (RS2) fullpol acquisitions and RapidEye (RE) images both at fine resolution collected over the Indian Head (Canada) agricultural site area and spanning the summer crop growth cycle from May to September. A quasi-Maximum Likelihood (ML) classification approach applied at per-field level has been adopted to integrate the different data sources. The analysis provided evidence on the overall accuracy enhancement with respect to the individual sensor performances, with 4%-8% increase over a single RE image, a 40%-10% increase over a single 1-pol/full-pol image and 15%-0% increase over multitemporal 1-pol/full-pol RS2 series respectively. A more detailed crop analysis revealed that in particular canola and the cereals benefit from the integration, whereas lentil and flax can experience similar or worse performance when compared to the RE-based classification. Comments and suggestions for further development are presented.

crop classification
sensor integration
maximum likelihood
SAR polarimetry

http://resolver.tudelft.nl/uuid:14e44d11-4561-4211-b302-3a3cd3b33b26

https://doi.org/10.1117/12.2029330

SPIE

978-081949756-7

http://proceedings.spiedigitallibrary.org/proceeding.aspx?articleid=1757285

Proceedings of SPIE 8887: Remote Sensing for Agriculture, Ecosystems, and Hydrology XV, Dresden, Germany, 23-26 September 2013

Institutional Repository

conference paper

© 2013 SPIE