Data for geometry-based greenness index comparisons in lettuce biophysical estimations

The lettuce images were captured daily using consumer-grade Vivo17 Pro smartphone camera for full six-week crop life cycle. Four lettuces are both cultivated in indoor and outdoor aquaponic setup. The distance of camera to lettuce canopy is 10 inches. The presence of artificial and natural photosynthetic light is present on the images. The raw images have corresponding vegetation channel for triangular greenness index (TGI) and artificial bee colony-optimized visible band oblique dipyramid greenness index (vODGIabc). The vODGIabc channel must exhibit more enhanced vegetation pixels than TGI transformed channel. The data shows that using swarm intelligence based optimization and the introduction of saturation and value component rectification increases the sensitivity of vODGIabc to green color pixels resulting to subtracted non-vegetation pixels in the image even with the presence of shadow and bright light sources. TGI and vODGIabc images are discriminated against the ground truth images annotated using superpixel and watershed transformation. To further the analysis, each ground truth image is accompanied with actual measurement of biophysical signatures of lettuce. The use of computational intelligence for lettuce fresh weight, height, number of spanning leaves, leaf area index and growth stage estimation is of significant consideration. Matlab codes for optimization, and machine learning and deep transfer network estimations are also included.