Precise acquisition of potted plant traits has great theoretical significance and practical value for variety selection and guiding scientific cultivation practices. Although phenotypic analysis using two dimensional(2D) digital images is simple and efficient, leaf occlusion reduces the available phenotype information. To address the current challenge of acquiring sufficient non-destructive information from living potted plants, we proposed a three dimensional (3D) phenotyping pipeline that combines neural radiation field reconstruction with path analysis. An indoor collection system was constructed to obtain multi-view image sequences of potted plants. The structure from motion and neural radiance fields (SFM-NeRF) algorithm was then utilized to reconstruct 3D point clouds, which were subsequently denoised and calibrated. Geometric-feature-based path analysis was employed to separate stems from leaves, and density clustering methods were applied to segment the canopy leaves. Phenotypic parameters of potted plant organs were extracted, including height, stem thickness, leaf length, leaf width, and leaf area, and they were manually measured to obtain the true values. The results showed that the coefficient of determination (R²) values, indicating the correlation between the model traits and the true traits, ranged from 0.89 to 0.98, indicating a strong correlation. The reconstruction quality was good. Additionally, 22 potted plants were selected for exploratory experiments. The results indicated that the method was capable of reconstructing plants of various varieties, and the experiments identified key conditions essential for successful reconstruction. In summary, this study developed a low-cost and robust 3D phenotyping pipeline for the phenotype analysis of potted plants. This proposed pipeline not only meets daily production requirements but also advances the field of phenotype calculation for potted plants.

Seed viability fundamentally determines crop establishment, stress resilience, and yield stability in pepper (Capsicum annuum L.), yet conventional assessment remains destructive, labor-intensive, and poorly scalable, while existing spectral learning approaches largely lack physiological interpretability, limiting their reliability for industrial seed quality management. Here, we present ViaNet, a lightweight, interpretable deep hyperspectral classification framework for 1038 naturally aged pepper seeds labeled via standardized 14-day germination tests. ROI-averaged hyperspectral reflectance vectors are modeled as a binary classification task, and ViaNet integrates Successive Projections Algorithm (SPA)-based wavelength sparsification with Efficient Channel Attention (ECA)-driven spectral weighting within a compact 1D-CNN architecture, enabling physiologically grounded feature learning under strict computational constraints. The model achieves recall for germinable seeds (79.75%) and outperforms classical machine learning methods. In addition, ViaNet consistently highlights reproducible spectral bands associated with natural-aging-related biochemical changes as reported in the literature (e.g., carotenoid-related absorption features in the near-UV region). By coupling spectral feature selection with attention-guided wavelength focusing, ViaNet establishes a closed analytical chain from spectral compression to physiologically interpretable inference. This framework balances predictive accuracy, interpretability, and deployability and provides a scalable, non-destructive, and biologically informed paradigm for hyperspectral seed viability assessment.