Contribution of texture and red-edge band for vegetated areas detection and identification

High resolution GIS data describing forests is an important knowledge, both for mapping and for environmental monitoring purposes. The extraction of such information out of imagery consists in a detection of woody areas followed by a thematic enrichment in forested areas, including a discrimination between evergreen, deciduous and mixt plantings. This paper attempts to automatize these photo-interpretation tasks. It particularly emphasizes on the determination of the most suitable input data to cope with these two classification problems. Two kinds of optical images have indeed been used: RapidEye data and 50cm ground resolution aerial ortho-images. Aerial data provided very high resolution information and texture indices, whereas RapidEye data brought additional radiometric information, and especially a red-edge channel. It has then been shown that texture information from aerial 50cm images was a key information for the woody area detection and that the red-edge band of RapidEye data appeared to be useful to discriminate between evergreen and deciduous plantings.