Mean shift segmentation applied to ADS40 data for automatic forest detection

National forest inventories (NFI) are essential for countrywide estimations of a wide range of forest functions. Our research aim is to derive measurable forest features out of airborne image data by using automatic computer-vision based methods. This paper focuses on tree layer detection of high resolution ADS40 data for automation. Preliminary experimental results of mean-shift segmentation method combined with curvature features from airborne laser scanning (ALS) for automatic tree layer detection are presented. Further research is needed to connect separate tree patches into forests according to specific forest definitions.