"Artificial intelligence-assisted tree diameter-at-breast-height measurement" (2026)
/Zhiheng Yin, Songlin Fei, and Song Zhang, “Artificial intelligence-assisted tree diameter-at-breast-height measurement,” Measurement Science and Technology, 37 465902 (2026)
Abstract:
Accurate diameter-at-breast-height (DBH) measurement is essential for forest inventory and ecological monitoring. This work evaluates two artificial intelligence-assisted smartphone frameworks: a text-prompt Segment Anything Model 3.1 (SAM3.1) that segments trunks from red–green–blue images, and a depth-guided MobileSAMv2 model that uses depth cues to generate geometry-based prompts. The resulting masks are integrated into a geometric pipeline for accurate and interpretable DBH estimation. Both methods were tested on 202 planted and 285 natural trees. In planted forests, both achieved high accuracy with no failures and a root mean square error (RMSE) of 1.0 cm. In natural forests, SAM3.1 remained robust (1.5 cm RMSE), while MobileSAMv2 failed in two cases and showed lower accuracy (1.9 cm excluding failures). Using all 487 images, MobileSAMv2 model inference averaged 1.3 s per image on an iPhone 15 Pro. SAM3.1 segmentation averaged 37.2 s per image on an Intel Core i7-1260P CPU, compared with 3.1 s on an NVIDIA Tesla T4 GPU and 0.8 s on an NVIDIA A100-SXM4-40GB GPU through Google Colab. Overall, MobileSAMv2 supports direct on-device deployment, while SAM3.1 offers greater robustness in complex environments and benefits substantially from GPU acceleration.