3D reconstruction of uneven-aged forest in single tree scale using digital camera and SfM-MVS technique

ABSTRACT

This study highlights the potential of hand-held digital cameras and SFM-MVS to produce a 3D reconstruction of individual trees at the single tree level in a deciduous and uneven-aged forest (faculty of natural resource, Tarbiat Modares University-area with about 2400 m² and 35 trees). A high coefficient of the determinant (R²= 98%) was observed between DBH derived from both filed measurements and the SfM-MVS technique. The percentage of the coefficient of variation of the root mean square error (CV(RMSE)%) and Bias of the technique to estimate DBH of the trees were 5.66% and −0.09%, respectively. The technique demonstrated a good linear relationship between the estimated tree height via field measurement and SfM-MVS (R²= 89%), with a percentage of 14.84 for CV(RMSE)% and −11.27 for Bias. Furthermore, the technique’s performance in the 3D reconstruction of single trees showed an omission of 11.42%, commission error of 0%, and overall accuracy of 88.57%. The technique exhibited an accuracy of about −0.33 m for estimating trees’ location in the forest area. Finally, the findings suggested combining the SfM-MVS technique with a digital camera could be useful for 3D reconstruction of single trees and estimating tree attributes in uneven-aged and multi-structure forests.

KEYWORDS:

• Forest inventory
• tree attributes
• 3D point cloud
• image-matching
• digital stereo imagery

Disclosure statement

No potential conflict of interest was reported by the author(s).