Artificial Vision System Based on Neural Networks to Identify Defects in Yungay Potato and Huevo de indio Caused by Premnotrypes Vorax (#1057)

Authors

Bobadilla Valderrama, Milton Omar

Diaz Alfaro, Andrea Ximena

León León, Ryan Abraham

Abstract

In the present investigation a proposal of an artificial vision system is presented, based on neural networks to identify defects in the yungay potato and Indian egg, caused by the Premnotrypes Vorax. The objective of this research is to develop an artificial vision algorithm that allows the detection of potatoes with defects caused by the white worm. For the realization of the software, the Python programming language and the Tensorflow platform were used. For this, an i5-1.60 GHz, 1800 Mhz computer was used for a correct image processor, it also has a capacity of 4 Gb of RAM memory, a 1080-pixel camera was also used to be connected to the equipment. Through an experimental and quantitative trial-and-error methodology, the designed software allowed the classification to be carried out with an efficiency level of 96.33% for the Yungay potato, while for the Huevo de Indio potato, the efficiency was 95.12%, observing which is more effective in the yungay potato. Concluding that the implementation of software with artificial vision is a good opportunity for improvement for potato farmers because they would be more efficient when classifying potatoes