Derivation of a 3-D forest canopy model and detection of tree crowns from high-resolution aerial photos

This paper presents our proposal of a methodology procedure aimed at derivation 3-D forest canopy model from aerial multi-spectral stereo photos with a high resolution. Digital model is derived from photos using methods of digital photogrammetry. Furthermore, we used this model to determine the tree tops. For this purpose we created a system of criteria based on mathematical and logical relations. Tree tops represent the number of trees in the forest stand. We examined the accuracy of this classification on 5 areas. Errors of automated classification compared to terrestrial measurements ranged from underestimation by 41% to overestimation by 40%. It would require to incorporate a spectral information into the process of tree tops determination in order to improve the classification.