Non-contact weight estimation of piglets using RGB images and explicit geometric priors via cross-attention fusion
Smart Agricultural Technology · 2026-12 · 15 · 102691
DOI: 10.1016/j.atech.2026.102691 ↗Kurz gefasst
Fachartikel in der Zeitschrift Smart Agricultural Technology, 2026-12. Verfasst von Xiaodie Zhong und 6 weiteren Autor:innen. Die Arbeit nennt 41 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY-NC-ND 4.0).
Abstract (Originalsprache)
Body weight is a key phenotypic indicator of piglet growth, development, and health status. Conventional manual weighing can induce stress, whereas depth-camera-based non-contact methods remain costly for large-scale deployment. Ordinary two-dimensional RGB images are inexpensive and easy to deploy, but they provide limited information about local structural scale. To address these limitations, this study proposes a piglet weight-estimation model that combines explicit geometric priors with a Cross-Attention mechanism. First, a YOLO11-based cascade image-processing pipeline extracts individual masks by instance segmentation, and a body-axis straightness constraint is introduced for automated posture filtering. Second, a compact set of explicit geometric priors, termed Geom-4, is selected from projected area, body length, shoulder width, and hip width to provide interpretable scale cues for two-dimensional contour representations. On this basis, an image–geometry Cross-Attention mechanism is designed to support dynamic interaction between explicit geometric priors and image features, thereby strengthening the representation of overall scale and local structure in two-dimensional contours. In real pig-house environments, the proposed method achieved a mean absolute error (MAE) of 0.243 kg and a coefficient of determination ( R 2 ) of 0.933, outperforming image-only regression and conventional feature-concatenation strategies. These results indicate that cross-attention fusion of explicit geometric priors can improve the representation of projected scale and local structural information, providing a feasible approach for low-cost, non-disturbing automated piglet growth monitoring.
Zitieren
APA
Zhong, X., Jiang, X., Liu, Q., Chen, H., Chai, J., Luo, Z., & Xu, S. (2026). Non-contact weight estimation of piglets using RGB images and explicit geometric priors via cross-attention fusion. Smart Agricultural Technology, 15, 102691. https://doi.org/10.1016/j.atech.2026.102691
Vancouver
Zhong X, Jiang X, Liu Q, Chen H, Chai J, Luo Z, et al. Non-contact weight estimation of piglets using RGB images and explicit geometric priors via cross-attention fusion. Smart Agricultural Technology. 2026;15:102691. doi:10.1016/j.atech.2026.102691
BibTeX
@article{Zhong2026,
author = {Zhong, Xiaodie and Jiang, Xin and Liu, Qingzhi and Chen, Hongji and Chai, Jie and Luo, Zonggang and Xu, Shunlai},
title = {Non-contact weight estimation of piglets using RGB images and explicit geometric priors via cross-attention fusion},
journal = {Smart Agricultural Technology},
year = {2026},
volume = {15},
pages = {102691},
doi = {10.1016/j.atech.2026.102691}
}
EUVIMED