Individual Tree Top Position Estimation by Template Voting

Templates derived from an optical tree crown model have been used with good results in the estimation of individual spruce tree top positions in aerial photographs. The method can nd tree tops from a wide range of viewing angles including extreme oo-nadir views, but a requirement has been that the number of trees was estimated in advance and that various parameters were well calibrated. A development in the template matching technique is presented, in which the variations in backdrop for imaged trees is handled by letting multiple template matchings cast votess to determine the most likely positions of tree tops. The method requires less accurate parameter calibration and allows clear separation of high qualityy tree top candidates from those less likely to correspond to actual trees, without the need for an a priori stem count estimate. Performance is evaluated on images of a thinning experiment in Norway Spruce. The voting method seems able to compete with the simple template matching with respect to the quality of the tree top candidates found.