Enhancing forest inventory via a videogrammetry approach for robust 3D reconstruction: A study using Insta 360 Pro 2

Abstract

In this study, we explore the application of videogrammetry for 3D reconstruction in complex forest environments, aiming to enhance forest inventory measurement methods. Traditional techniques are often labor-intensive and lack scalability in dense or challenging terrain. We assess the efficacy of videogrammetry for generating 3D point clouds in complex forest environments, focusing on an Insta 360 Pro 2 setup with six fish-eye cameras. Harnessing this lightweight and user-friendly technology, we aim to elevate the process of data collection while delivering realistic visual representations of forest areas. Our approach enables the estimation of key forest characteristics, such as tree distribution and Diameter at Breast Height (DBH). The average errors for tree position and DBH measurements range from 5.2 cm to 18.8 cm and from 0.9 cm to 1.9 cm, respectively. The reconstructed 3D tree information is structurally similar to data obtained with Terrestrial Laser Scanning (TLS), with normally distributed Multiscale Model-to-Model Cloud Comparison (M3C2) errors with a mean of 0 cm and a standard deviation of 15 cm to 22 cm. Our method reduces the need for manual data collection, thus supporting effective forest management and planning.

Code and data availability

The authors publicly deposited the videogrammetric point clouds generated by their pipeline (with walkthrough demos and TLS comparison videos) on Zenodo, directly reproducing this paper's 3D reconstruction measurements.