Anais do XIX Simpósio Brasileiro de Sensoriamento Remoto

Forest Inventory and monitoring is important for coconut plantation owners as it helps them in tracking forest growth, fruit production rates and plantation vitality. From this aspect, automated Individual Tree Detection (ITD) is very helpful as it makes the aforementioned processes less time‐consuming, affordable, and efficient; however, applications of ITD is still at a latent stage in several emerging economies such as Brazil. Herein, we combined Light Detection and Ranging (lidar) and local maxima algorithm to automatically detect coconut tree tops from a plantation having plots of varying canopy cover densities. Our accuracy assessment results (average tree detection accuracy = 79.77%) shows that application of local maxima algorithm on lidar-derived canopy height models (CHM) - along with suitable filter window sizes and fixed window sizes, according to plantation density and within-plot tree distribution - can predict coconut trees with satisfiable accuracy (F-score > 0.85) and thereby assist the plantation sector’s monitoring practices.