Development and Application of a Low-Cost 3-Dimensional (3D) Reconstruction System Based on the Structure from Motion (SfM) Approach for Plant Phenotyping

Purpose: This study aimed to develop a low-cost non-destructive plant growth observation system based on 3-dimensional (3D) modeling using the Structure from Motion (SfM) method and apply it to the volume estimation of Chinese cabbage. Methods & Results: The system included a camera arm, a stepper motor, a web camera, LED strip lights, and image processing software for generating 3D models. The system was evaluated using Chinese cabbage over a 35-day growth period. The developed system successfully generated 3D models closely resembling the actual plant, with volumetric validation results of R2 = 0.9956, RMSE = 13.97 cm3, and MAPE = 71.33. Furthermore, a one-way ANOVA was conducted to assess the significance of the volume measurements obtained using the developed system. The ANOVA results showed no significant difference in volume compared to the conventional system (ρ-value = 0.946). Conclusion: The system showed promising results for generating 3D models of plants and accurately estimating their volume. However, further improvements are necessary for precise measurements and validation with other crop types. The results suggested that the proposed system has the potential to be an efficient tool for precision agriculture in plant growth observation and monitoring. © The Author(s), under exclusive licence to The Korean Society for Agricultural Machinery 2024.