SAR Applications in Agriculture: Multiband Correlation and Segmentation

This paper presents the results of an experiment on the application of SAR data for crop classification at an agricultural site in Quebec. The main aspects discussed include the correlation of X and C band data for 14 vegetation types and the comparison of a pixel classification and a classification using segmentation results. The segmentation methodology uses a new algorithm (Ait Belaid et al. 1989), which incorporates cartographic information into the segmentation process. The correlation analyses show that multi-frequency data (X and C band) add significant information for vegetation analyses. Classification results show that the overall classification accuracies obtained with a normal pixel classification of either X or C band data were generally low (less than 45%). However, the combination of X and C bands gives better overall classification accuracies (up to 53%). Improvements in classification accuracies for individual crops of 10 to 30% are found using a segment classifier, which employs the segm...