3D plant phenotyping has garnered significant interest for its ability to quantify key structural traits such as plant volume and canopy architecture. However, standard monocular 3D reconstruction techniques suffer from inherent scale ambiguity, requiring an additional step to recover the true metric scale of the plants. Existing scale recovery methods, whether based on precisely fabricated 3D objects or planar patterns such as checkerboards, have been successfully applied in controlled environments but face practical constraints in certain real-world scenarios: some require costly fabrication or pre-reconstruction calibration, which can limit throughput in dynamic field environments. Here, we present MagicRing, a novel, affordable, and physically reliable post-reconstruction scale recovery approach that addresses these specific constraints and provides a complementary solution for high-throughput, mobile, and field-based phenotyping. MagicRing features a simple red ring printed on A4 paper with a known diameter. By leveraging color-based segmentation and geometric curve fitting, our approach automatically detects the ring within 3D point clouds, recovers the metric scale, and establishes a standardized world coordinate system without the need for pre-calibration. Its planar, isotropic design ensures robustness even under significant occlusion. We demonstrate the utility of MagicRing through MobilePheno3D, an integrated smartphone-based pipeline that performs fully automated 3D reconstruction, scale recovery, and phenotypic extraction from video sequences. This system, which was validated across multiple plant species, including vegetables, wheat, rice, and maize in both indoor and field settings, reliably reconstructs aboveground and root structures and supports continuous growth monitoring. MagicRing decouples data collection from data analysis, enabling a workflow transition from conventional step-by-step, scene-specific calibration toward more scalable, high-throughput 3D plant phenotyping.
Keywords: 3D plant phenotyping; 3D reconstruction; Affordable phenotyping; MagicRing; Metric-scale recovery.
© 2026 The Authors.