Enhanced Treetop Detection in Savanna Woodland: A New Method Using Variable Window Sizes and Marker-Controlled Watershed Segmentation

This study presents a new method for detecting individual treetops from lidar data and applies marker-controlled watershed segmentation to isolate individual trees in savanna woodland. The treetops were detected by searching local maxima in a canopy maxima model (CMM) with variable window sizes. Unlike previous methods, the variable window sizes were determined by the lower limit of the prediction intervals of the regression curve between crown size and tree height. The canopy maxima model was created to reduce the commission errors of treetop detection. Treetops were also detected based on the fact that they are typically located around the center of crowns. The tree delineation accuracy was evaluated using a five-fold cross-validation method. Results showed that the absolute accuracy of tree isolation was 64.1%, significantly higher than the accuracy of the method that only searched local maxima within window sizes determined by the regression curve (37.0%).

Additionally, the proposed method was compared with two existing methods: a manual delineation method and a watershed segmentation method using a fixed threshold. The proposed method outperformed both methods, including the manual delineation method, with an absolute accuracy of 64.1% compared to 58.5% and 60.1%, respectively.

The study also demonstrated the effectiveness of the proposed method in detecting individual trees in savanna woodlands, characterized by scattered trees and open canopy cover. The method could be further enhanced by incorporating additional information, such as spectral and textural features, and by applying machine learning algorithms to improve the accuracy of tree detection.

Overall, the study presents a new and promising method for detecting individual trees from lidar data, which could have important implications for forest monitoring and management.