Table 1.
Specification of the camera bodies used in the experimental trials.
Fig 1.
Set up of the photogrammetric station.
A mirrorless camera Canon EOS R5 with a macro lens (Canon EF 100mm f/2.8L Macro IS USM) is mounted on a tripod. The carpological material is affixed on a pin which is placed in the middle of a turntable. Three light-plates are placed on the top, left and right side of the carpological material. A portable photo studio is used as background.
Table 2.
Configuration of the camera Canon EOS R5 for image capturing.
Fig 2.
Camera position estimated by 3D Zephyr Lite.
97 images of the fruit of Ficus pumila Linnaeus (PH013) were captured from 0°, 20°, and -20° vertically towards the carpological material, with at a horizontal rotation interval of 11.25°. The red triangles represent the camera positions estimated by 3D Zephyr Lite.
Fig 3.
Composition of a 3D model reconstructed by photogrammetric platform.
(a) A carpological material of the fruit of Aleurites moluccana (Linnaeus) Willdenow (PH005). (b) Texture of the 3D model extracted by 3DF Zephyr Lite. (c) A 3D model with texture to show the genuine color of PH005. (d) A 3D model without texture to show the plain mesh of PH005.
Fig 4.
A workflow for 3D digitization of carpological material.
(a) The set-up of the image capturing system includes background color, lighting, camera position and the position of the carpological material. The adjustment of these parameters depends on the morphology of the carpological material. (b) Image capturing process of each material usually lasts for 5 to 10 minutes. (c) The images are imported to the software for 3D model reconstruction. The set-up should be adjusted if the reconstruction was failed. (d) Post-reconstruction editing would be carried out if the model is successfully reconstructed.
Fig 5.
Light reflection on shiny surface.
(a) Two light plates were placed on the left and right sides of the fruit of Heritiera littoralis Aiton (PH018) respectively. The light reflection created multiple white spots on the surface which resulted in failure of image recognition by the software. (b) A paper was added between PH018 and the left light plate, which diffused most of the direct light towards it. The light reflection on the left side was significantly reduced.
Fig 6.
Comparison of 3D models of long dimension carpological material reconstructed by different positioning.
(a) A carpological material of the fruit of Samanea saman (Jacquin) Merrill J. Wash (PH044). (b)&(c) The 3D model when PH044 was placed horizontally. The structure on the front side was reconstructed but the back side was not, as shown by the regions with darker color. (d) Decrease in sharpness of the images of PH044 was found due to greater depth of field from two ends, which reduced the coherence between images so the software could not recognize its structure for further reconstruction. (e)&(f) The 3D model when PH044 was placed vertically. The structure was reconstructed completely on both sides.
Fig 7.
A comparison of the indumentum of 3D models and the carpological materials.
(a) Side view of a carpological material of Melastoma malabathricum Linnaeus. (b) Side view of the 3D model PH193. (c) Side view of the mesh with texture turned off.
Fig 8.
Reconstruction trials of carpological materials with filament-like and thin structures.
(a) A Seed of Gymnema sylvestre (Retzius) Schultes with coma attached. (b)to(f) Results of the reconstruction trials of G. sylvestre showing the positions of the camera estimated by 3D Zephyr and the 3D models.
Fig 9.
Reconstruction trials of carpological materials with transparency.
(a) A seed of Oroxylum indicum (Linnaeus) Bentham ex Kurz. The background color passed through the transparent parts of the seed. (b) A 3D model of the same seed reconstructed with a black background. The transparent parts were stained with background color.
Fig 10.
Comparison between 2D image and 3D models of fruits of Sterculia lanceolata Cavanilles.
(a) A carpological material of the fruit of S. lanceolata (PH053). (b) 2D image of a specimen of S. lanceolata (Shiu Ying Hu 5569). (c) Front side of the 3D model of PH053. (d) Back side of the 3D model of PH053.
Fig 11.
3D models of Lophostemon confertus (R. Brown) Peter G. Wilson & J. T. Waterhouse at different time points after collection.
(a) A carpological material of the fruit of L. confertus was fixed for image capturing. (b) Top view of the 3D model (PH054) reconstructed on day 1. The valves on top of the fruit were closed. (c) Top view of the 3D model (PH054) reconstructed on day 2. The valves of the capsule started to retreat, and the seeds inside were exposed. Two sepals contracted and bended inwards. (d) Top view of the 3D model (PH054) reconstructed on day 5. Most of the seeds had fallen and the internal structure of the capsule was clearly seen. (e) Top view of the 3D model (PH054) reconstructed on day 10. All the seeds were released.