Tree classification using aerial imagery

Abstract

Type of a tree or species of a tree is its most extraneous property. Remote sensing has served as great solution to the problems related to manually acquiring data for tree species classification. Most of the studies in this area of research use Lidar & satellite imagery or combination of data sources. But these techniques have some disadvantages like Lidar is not cost infective in small in rainfall. Hyper-spectral data is not feasible and Multi-spectral data are prone to factors such as viewing angle, sun angle, day / year time (seasons).Some studies have used optical imagery with the combination of other data sources. The optical imagery that have been used does not provide aerial view but side view of the tree which is not feasible to acquire for large areas. Therefore, there is a need for the solution which is effective, economical and scalable. A little research has been done on the classification, with over head view, using low cost commercial drones with optical sensors. In this research optical aerial imagery has been used for tree species classification. Combination of statistical and spatial features is used as in put to the two classifiers artificial neural network (ANN) and support vector machine (SVM).The results show that 97.4% and 96.26% accuracy, respectively, has been achieved.