Detection and Classification of trees using Deep Learning

Abstract

A Common problem in forest applications is that of finding and counting trees. Currently this is frequently done using land surveys that are expensive and time-consuming. Therefore, in this study, we aim to be able to detect and count individual tree tops automatically and efficiently by using drone-acquired forest images. In order to achieve this goal, we use computer vision and deep learning techniques to automatically detect individual trees in RGB image mosaics and Digital Elevation Models. Previous similar studies have been carried out in flat, plantation forests. In this respect our data is particularly challenging as we captured it in a Japanese natural mixed forest set in hilly terrain. We explore the use of Deep Learning networks to predict tree density in small regions of the forest and couple it with different image clustering algorithms to separate individual trees.