The development of automatic methods for handling SAR images over land

Aspects of image handling relevant to a SAR producing images over land, particularly for monitoring agriculture and forestry, are discussed. Identification of ground cover type (classification); change detection; and change interpretation are considered. Segment-growing methods, algorithm performance, and thin line detection as part of the segmentation process are treated. After segmentation, shape and texture of segments can be measured. Work on the former and a comparison of man and machine perceptions of texture are described. Registration requires feature extraction, and results indicate that automatic correlation methods appear viable for speckled images. Calibration, classification, and interpretation of change require more experimental data, and merging of data from sources apart from the SAR.