Individual Olive Tree Detection in RGB Images

In this paper, an automatic method for detecting and counting olive trees in RGB images acquired by an unmanned aerial vehicle (UAV) is developed. Our approach is based on implementation of RetinaNet model and DeepForest Phyton package. For improvement of pretrained model via transfer learning, five olive groves were mapped using UAV, trees were manually labeled, and new image dataset was created. Several models were built, each being trained and evaluated on different set of images from selected olive groves. Experimental results obtained on a UAV image acquired over test olive groves are reported and discussed. Detection results showed high reliability of proposed approach and great improvement in performance compared to pretrained model.