Skip to main content
Dryad logo

Data from: Robust single-image tree diameter estimation with mobile phones

Citation

Holcomb, Amelia; Tong, Linzhe; Keshav, Srinivasan (2022), Data from: Robust single-image tree diameter estimation with mobile phones, Dryad, Dataset, https://doi.org/10.5061/dryad.vdncjsxxj

Abstract

Ground-based forest inventories are a key element of forest carbon monitoring, reporting, and verification schemes and a cornerstone of forest ecology research. Recent work using LiDAR-equipped mobile phones to automate parts of the forest inventory process assumes that tree trunks are well-spaced and visually unoccluded, or else requires manual intervention or offline processing to identify and measure tree trunks.

In this paper, we design an algorithm that exploits a low-cost smartphone LiDAR sensor to estimate trunk diameter automatically from a single image in complex and realistic field conditions. We implement our design and build it into an app on a Huawei P30 Pro smartphone, demonstrating that the algorithm has low enough computational cost to run on this commodity platform in near real-time.

We evaluate our app in three different forests across three seasons and find that in a corpus of 97 sample tree images, our app estimates trunk diameter with RMSE of 3.7 cm (R2 = .97; 8.0% mean error) compared to manual DBH measurement. It achieves a 100% tree detection rate while reducing surveyor time by up to a factor of 4.6.

Our work contributes to the search for a low-cost, low-expertise alternative to Terrestrial Laser Scanning that is nonetheless robust and efficient enough to compete with manual methods. We highlight the challenges that low-end mobile depth scanners face in occluded conditions and offer a lightweight, fully automatic approach for segmenting depth images and estimating trunk diameter despite these challenges. Our approach lowers the barriers to in situ forest measurement outside of an urban or plantation context, maintaining a tree detection and accuracy rate comparable to previous mobile phone methods even in complex forest conditions.

Methods

For details of data collection methods, see Materials & Methods, Section 3.4 in the corresponding paper.

The data was processed using code available at https://github.com/ameliaholcomb/trees

The output results can be reproduced in a file called results.csv (along with images demonstrating the processing steps) by running

$ python3 OfflineProcessing/v2_DepthAssistedSegmentation/RGBD.py /path/to/image/data /path/to/reference_widths.txt /path/to/output/directory

Usage Notes

The files with no extension are depth images, which are stored in the format
(x-coordinate of pixel, y-coordinate of pixel, depth measurement of pixel, confidence value)
as returned directly from the phone depth sensor.

The manually measured tree diameters for each plot are stored in reference_widths.txt, in the format
sample_number:diameter_in_m

The image file names are formatted as
Capture_Sample_{sample_number}_{capture_number}
The capture numbers are to ensure unique file naming. The sample numbers correspond to the sample numbers listed in reference_widths.txt

Funding

Canadian National Research Council

University of Cambridge