Non-destructive internal quality inspection of apple fruit by X-ray imaging and deep learning

Apple fruit is part of the daily diet of many people in the world. Unfortunately, internal disorders regularly develop during growth and storage. Nowadays, commercially available quality inspection systems inspect only external quality, while the inside of the apple cannot be inspected reliably with non-destructive techniques. To deal with this limitation, apples are sampled and cut-open to visually inspect the internal disorders. If too many fruits are affected by defects, the whole batch is often discarded. Evidently, this leads to high financial losses as there are yet unaffected fruits in the batch. In addition, affected apples can also occur in batches that passed the previous selection. Presence of internal disorders so also negatively affects the consumers’ perception of quality. Therefore, the aim of this PhD project is to develop a non- destructive internal quality inspection system to identify and classify apples with various internal disorders, such as internal browning and watercore, through a unique combination of X-ray imaging and deep learning. Apples of different cultivars and with different disorders will be produced and characterized in 3D, a simulation framework will be developed for data augmentation and different deep learning models will be built and validated to segment and identify specific disorders. The method will be implemented and tested on a prototype for fruit from different origins and seasons.