Spatial tree pattern analysis from maxima of smoothed aerial photographs

An aerial photograph of a forest is digitized and smoothed, e.g. by a spher ically symmetric kernel at a suitably chosen bandwidth or by matching of te mplates derived from a light reflection model. The maxima of the smoothed i mage yields estimates of individual tree positions and thus allows statisti cal analysis of the spatial neighbour patterns, which are important for pre dicting future growth and timber quality of the trees. Theoretically, the statistical problem is estimation of one marked point pr ocess from observation of another such process. In our application the proc ess to be estimated is the true tree position process with marks consisting of stem height, stem diameter and crown shape. The observed point process corresponds to local maxima of the smoothed image with marks determined by the local properties of the image close to these maxima.