Toddy Palm Trees Classification and Counting Using Drone Video: Retuning Hyperparameter Mask-RCNN

Abstract

Toddy Palm Tree, Borassusflabellifer, is one of the famous palm trees family in Myanmar, India and South East Asia. It is most important economic and symbolic tree in Central Myanmar. This paper presented palm tree classification on Toddy palm, Coconut palm, Palm oil and counting using combines a contribute method remote sensing drone video and deep learning architecture known as mask R-CNN with retuning hyperparameter strategy. The aerial images, are organized by Drone in Upper and Delta Coastal area of Myanmar, extracted features bounding boxes and classified. As the system prepared Myanmar Palm trees dataset with over 12,000 images, examined the performance with retuning hyper-parameter by using Bayesian optimization algorithm in predefined learning rate and momentum. The research concluded that tuning learning will improve the performance of classification for local palm tree segmentation task. The result show that the system research can accurately define the Toddy palm with better accuracy and counting perfectly.