Automated system for classifying Hass avocados based on image processing techniques

Farming of Hass-variety-avocado has become an important economic activity in Colombian agriculture for covering local market requirements but also for exporting high quality fruits to foreign markets. Many processes still need to be improved in different regions of Colombia where Hass variety is cultivated in order to increase the efficiency during harvesting, transportation and storage. Human-performed classification of avocados takes time that could be used by farm´s personnel to perform other activities aimed to achieve higher quality standards and production efficiency. Product price and target market depends to a large extent on classification stage. We propose a system based on image processing of avocados´ pictures for their classification after they have been picked off the tree. The classifying algorithm runs on a computer which receives fruit´s images wirelessly from a camera, the result is transmitted using IEEE 802.15.4 packets and avocados are placed in conveyor belt rails according to their categories. Matching percentage of the image processing algorithm for selecting avocados was of 88% compared to the test set classified by expert farmers. Confusion matrix showed 82.22% accuracy when classifying avocados.