Tree diameter at breast height measurement based on computer vision

Abstract

Diameter at breast height (DBH) is a crucial metric in forestry, serving as a key input for estimating timber volumes and biomass, assessing forest health, and aiding in biodiversity and climate change studies. However, traditional measurement methods practiced today are time-consuming and labour-intensive, while many advanced methods introduced in recent years require high upfront costs, limiting wide adoption by small-scale institutions and projects. This research paper aims to explore innovative approaches to DBH measurement that balance accuracy with cost-effectiveness, ultimately contributing to the broader goals of sustainability and environmental protection. In this paper, the authors propose an automated DBH measurement method, extracting the value from smartphone RGB images through the utilization of computer vision techniques and mathematical algorithms. By incorporating tree distance data in Phase 3 of the study, the proposed method achieved accuracy comparable to manual tape measurements while significantly reducing the time and resources required for fieldwork. Specifically, 74 out of 143 trees (51.7%) had an estimated DBH that fell within 1 cm of the actual measurements, resulting in an absolute mean error (MAE) of 1.10 cm, root mean square error (RMSE) of 1.80 cm, and relative root mean square error (RRMSE) of 6.0%. Thus, this hybrid approach offers a promising solution for forestry applications, enhancing both the efficiency and accessibility of DBH data collection.