A blueberry classification algorithm using convolutional neural networks developed in Python and a Raspberry Pi 4

The present article proposed an algorithm of Deep Learning for blueberry (Vaccinium Corymbosum) classification based on their quality characteristics. The problematic is the high export demand of the fruit in Peru and the lack of an automatic quality inspection in most of the small and medium business. The objective of this research is to achieve an accuracy greater than 90% in order to compete with the losses of the companies that uses manual inspection. Also, the processing time needs to be low so the program can be used in an industrial machine. For the developed of the network, an enclosure for image acquisition with indirect lighting and an inclination of 20 was designed. The images captured by a Raspberry Pi are analyzed by a convolutional neuron network (CNN) and according to the result a signal is sent to a PLC s7-1200 of a sorter machine in order to filter the non-exportable blueberries y pneumatic actuators. The results of the research indicate a validation accuracy of 95.64% and an average of 156.68ms of processing time.