Recognition of Lychee Fruit Maturity Based on Object Detection Algorithm

Picking robots have solved the problem of insufficient labor force and reduced fruit picking costs in orchards. The recognition of fruits in lychee orchards is a key technology for harvesting robots. This article takes tropical fruit lychee as an example and uses the Edge Impulse platform to collect data using OpenMV. Object Detection, BrainChip Akida, YOLOv5 for Renesas DPR-AI, and YOLOv5 are used as training models. Through experimental comparison and analysis of data collection quantity, DSP preprocessing, and training rounds, the accuracy of Object Detection is higher and the recognition of lychee fruit maturity is more effective. Therefore, the Object Detection model is quantified and deployed to the OpenMV end to complete the recognition of mature and immature lychees.