Fig. 1. Prototype of "CMD-1 non-contact detection device for crop growth status"

Fig. 2. The process of image information acquisition， processing and fusion

Fig. 3. Improved semantic segmentation network model structure based on U-Net

Fig. 4. Schematic of the imaging principle and training set

Fig. 5. Prediction effect of 4 cm， 6 cm， 8 cm hemispherical surface

Fig. 6. Prediction output of 8 cm diameter hemispherical surface and 3D shape recovery map

Fig. 7. Evaluation of 3D reconstruction effect

Fig. 8. Acquisition and prediction effect under the projection of tomato stripes with a radius of 3.51 cm

Fig. 9. Brix analysis of tomatoes based on hyperspectral data

Fig. 10. Spectral curves of different pixels in the experimental spectral data

Fig. 11. Spectral image of the same tomato at 480 nm， 520 nm， 640 nm

Fig. 12. Edge extraction preprocessing steps

Fig. 13. The distribution curve of the pixel value of the 640 nm spectral image section

Fig. 14. Edge detection effect based on height map

Fig. 15. Edge detection effect based on spectrogram

Fig. 16. The edge pixel distribution of the sample and its curve fitting effect

Fig. 17. The process and effect of image registration

Fig. 18. Characterization of tomato fruit height-reflectance （at 640 nm） registration fusion