Assessing Performance of Granulometric Method for Individual Tree Counting (ITC)

要旨

The study aims to suggest a new approach for individual tree counting (ITC) using satellite data with advanced geoinformatics techniques to speed up forest inventory processes. Delineation of individual trees in the Deciduous Forest is challenging due to varying tree canopy size and complex mosaic of the crowns. Existing techniques (eg., Valley Following and Watershed Segmentation) are of complex nature on the part of implementation of these techniques and their shortcomings. Hence, it was decided to adopt a new, simple and user friendly technique. The theory of Granulometry along with morphological techniques was assessed to give better results. The study area is part of Forest Research Institute, Dehradun, in the state of Uttaranchal, India. The data used IKONOS panchromatic image. In areas having dense tree crowns Laplacian of Gaussian (LoG) filter was applied which produced good results. All the images obtained from method were then converted to binary images. This was done to facilitate "Particle Analysis" which was carried out for actual counting of the identified crowns. It was found that delineation carried out was fair to provide the count of the actual tree crowns but lacking in precise delineation. In mixed species separated plantation test image the 95% was accuracy obtained. While in single species clumped trees, 82.2% accuracy was obtained.