Automated System for the Classification of Cherimoyas by Neural Network

According to Midagri, cherimoya production in Peru was 20 thousand tons, being one of the most consumed fruits and difficult to meet quality standards in the selection stage due to its rapid ripening. This work develops an automated system for the classification of cherimoyas according to the degree of ripeness and size. For the simulation of cherimoya grading, the bottleneck was determined by direct observation and then the algorithm was programmed using the neural network and trained in YOLO V5 to recognize the external characteristics of cherimoya in green, ripe stage, so a mechanical system was considered for the classification by size, to then obtain the simulation in Factory IO and TIA PORTAL with connection to PLC S7-1200 1214 DC/DC/DC and a HMI TP700. Finally, the classification proposal was implemented in which 100% of the cherimoyas were recognized through the interactive HMI screen, being able to classify them in state, green, ripe, small and then automatically count them in 25 units per box, which has a graphical environment so that the operator can manipulate it.