Desarrollo de un prototipo para la detección de kikuyo en tiempo real mediante visión artificial en cultivo de hortalizas.

Abstract

One of the biggest problems that arise in crops and must be combated with technological tools is the presence of weeds, which compete directly with plants for nutrients, light and water, rather than being pest hosts. This project focuses on the design of a kikuyu detector using a raspberry pi 4 as the control system chosen for its small size and great computing power. This project deals with the detection of kikuyu in crops, an analysis of Ecuador reveals that it is a country where agriculture represents a great support to the economy, so it is important to focus on facilitating some of the processes necessary to obtain fruits, vegetables and other products obtained from the agricultural sector. Since the algorithm developed for the detection of kikuyu is designed to operate in a controlled environment, the values obtained may vary depending on lighting conditions and the presence of other weeds in the crop, taking into account these considerations, the recognition of kikuyu is good in the proposed scenarios, with an average of 79.25% correct detections and 20.75% incorrect detections, however in scenarios 3 and 4 more incorrect detections were reflected in the results, which indicates that there are steps such as irrigation or high environmental humidity that make detection difficult.