Forest Classification with High Resolution Satellite Images. Research into theUse of Spatial Information to Improve the Multi-Spectral Classification with Spot Images

摘要

Computerized classification of remote sensing imagery is addressed. Textureinformation is used by the human interpreter to distinguish different classes while a spectral classification program may not be able to recognize the difference. To overcome this problem, an approach incorporating texture information into the computerized classification process was tested over the north west coast of Tunisia and the Noordoost Polder in the Netherlands. No significant improvement was found in the forest classification test in Tunisia, possibly due to local circumstances, or by restrictions under which the research was carried out. The classification test of the Noordoost Polder showed substantially improved results.