Evaluation of Ground Plane Detection for Estimating Breast Height in Stereo Images

Diameter at breast height is a standard method for characterizing trees and ultimately forests over large geographical scales. Expressing the cross-section of a tree stem with a diameter measurement has inherent errors, as stems are not perfect conical frustums. Diameter suffices as an approximation, since methods for the exact cross-sectional area of a stem are infeasible in practice. There are other primary sources of error in diameter at breast height measurements: diameter measurement error and breast height measurement error. This paper addresses the latter in the context of an automated measurement system using 3D photography. We consider breast height estimation a fundamental component in such systems, as errors in measurement height can precipitate significant errors in stand-level attribute estimation as diameter changes along the height of the stem because of taper. We evaluate the efficacy of breast height estimation using stereo photogrammetry by fitting a planar surface to the ground and elevating the plane to breast height. Results from a validation test on 560 observations show that breast height can be estimated with a root mean squared error of 10.20 cm within 10 m from the camera and 13.36 cm within 20 m from the camera on flat terrain in varying levels of stem density. Study Implications: Increasing the level of automation in forest operations is one approach to address workforce attrition and safety issues in the industry. Automated tree measurement and mapping systems present an opportunity to reduce operational costs by providing machine operators with necessary information to carry out silvicultural prescriptions and eliminate the need for individual tree marking in certain situations. In order to maintain trust between landowners and harvesting contractors, it is imperative that automatic measurement systems are rigorously tested to determine that measurement accuracy is within acceptable limits. Breast height measurement error is often overlooked as a source of error in estimating diameter at breast height. However, accurate breast height estimation is a necessary prerequisite for reliable, consistent, and precise diameter at breast height measurement. The capacity of machine vision systems to automate forest inventory and monitoring tasks is increasing and can help managers achieve sustainable forestry objectives on more acres in less.