Vertical farming offers new opportunities for efficient and sustainable indoor produc-tion of plant-based foods. It requires continuous and non-destructive preharvest quality control to ensure high production quality. For this purpose, a computer vi-sion-based system for crop quality control was developed based on the OrbiPlant® vertical farming system.

The system consists of two regular cameras mounted on top of the Orbiplant® loop structure that capture color and depth images of each individual row of plants from the front and back as they pass through the conveyor belt-driven system. The cam-eras are enclosed in a box with light sources to ensure reproducible lighting condi-tions. To simulate rejectable plant material, leaf area damage of the plants was in-duced by herbicide application. Recorded images are analyzed by deep learning models to extract and analyze relevant quality information.

For each individual plant, an estimate of the leaf surface area is performed to evalu-ate the growth process against the plant population. In addition, discolored areas (e.g. due to disease or nutrient deficiencies) are detected and measured. The quality indicators are processed and made available to users in the form of a web-based dashboard for quality control with information on the position of the plant in the conveyor belt.