Mangrove individual tree detection based on the uncrewed aerial vehicle multispectral imagery

Abstract

Mangroves are one of the most threatened and vulnerable ecosystems in the world. They dominate the coastal wetlands of tropical and subtropical regions. These forest areas are frequently difficult to access, typical methods of forest inventory are difficult undertakings. Thus, the acquisition of forest attributes through remote sensing, such as the validated individual tree detection (ITD) aim can support management and conservation proposals. This study comprehensively validated ITD from high-resolution multispectral uncrewed aerial vehicle (UAV) surveys, forest inventory data, multiresolution segmentation, mean shift segmentation and gaussian model the aim of quantifying the canopies in different areas of mangrove. The validations methods were based on the number plot size, the polygon canopy, and point canopy obtained according to simple sampling plan statistics. The results demonstrated that when using plot size inside the flight area to calibrate the ITD methods, the gaussian model and the segmentation methods can reach an acceptable quality limit (AQL) of 15%. On the other hand, this study suggests as alternative to ITD a tree like objects (TLO) method for flying through dense forests without forest inventory with AQL of 40%. This study is critical for conservation and management efforts because it applies the reliability of multispectral UAV surveys to assessing tree abundance in mangrove forests and can be applied to quantify biomass and blue carbon in dense forests.