Automatic Estimation of Tree Stem Attributes Using Terrestrial Laser Scanning in Central Indian Dry Deciduous Forests

Forest inventories are critical for effective management of forest resources. Recently, the use of terrestrial laser scanning (TLS) to automatically extract forest inventory parameters at tree level (e.g. tree location, diameter at breast height (DBH) and height) has gained significant importance. TLS using both single-scan and multi-scan techniques, not only helps in detailed and accurate measurements of tree objects but also helps increase the measurement frequency. In the current study, we develop an automated solution to extract forest inventory parameters at individual tree level from TLS data by using random sample consensus (RANSAC)-based circle fitting algorithm. The method was evaluated on both single- and multiscan data by characterizing four circular plots of radius 20 m in dry deciduous forests of Betul, Madhya Pradesh (India). Over all the plots, tree detection rates of 75% and 97% were obtained using single- and multi-scan TLS data respectively. Tree detection rates were significantly affected by increase in distance from the scanner, in single-scan approach when compared to multi-scan approach. Field based DBH measurements correlated well using both single (R² = 0.96) and multiple scans (R² = 0.99). The DBH estimates from multi-scan TLS data resulted in low ischolar_main-meansquare error (RMSE) of 2.2 cm compared to that of 4.1 cm using single-scan. Further, tree heights were extracted from TLS data and validated with selectively measured trees on field (R² = 0.98; N = 65). The RMSE of tree height was estimated to be 1.65 m. The current results show the potential use of TLS in automatically deriving forest inventory parameters with reliable accuracy at individual tree level.