Traditional segmentation approaches do not generally work well with high resolution satellite imagery on account of the spatial complexity of the data and the problem of spectrally 'mixed' pixels at the borders of useful segments. In many areas of Europe the proportion of pixels which can not be treated easily in segmentation is high because of small field sizes. In this paper the failings of segmentation are examined and recommendations are made for redefining it and peradapted specifically to application requirements. 2,forming it as a heuristic image partitioning procedure 3.

No abstract

No abstract

A new segmentation method is applied to different types and combinations of optical data and radar data over an agricultural area. The images, which were of different resolution, have been coregistered in order to make a qualitative comparison of the segmented data possible. Comparison of the segmentation results on optical and radar images is done using ground truth for certain crops.

No abstract