Implementación de un prototipo para la clasificación automática de tomates riñón basado en la norma INEN 1745, aplicando técnicas de visión artificial

The process of manually classifying tomatoes by size and color is not only time-consuming but also inefficient and costly. The present study proposes a prototype that uses computer vision techniques to automatically classify tomatoes based on the INEN standard 1745 to address this issue. The implementation of the prototype was done in three main phases. Firstly, a detection and classification system was developed using the YOLOv5 algorithm. The system was trained using images acquired from tomatoes grown in a greenhouse located in the Chambo canton, province of Chimborazo. In the upcoming stages, physical components such as the Central Processing Unit, image acquisition system, and various electronic-mechanical elements were integrated with software developed in Python and C. The software was used for detection and recognition, while the hardware was managed by the software. In order to evaluate the system, 200 tomatoes were randomly selected and subjected to tests for size and color classification. The system's performance was compared to two traditional methods: observation based on experience alone and the use of a precision instrument. For the size test, the tomatoes' transverse diameter was meticulously measured using a digital caliper and manually classified by the operators. For the color classification test, the expert operator manually categorized the tomatoes as green or red based on his criteria. Finally, the results obtained from the system were compared to the results obtained from the traditional methods for size and color classification of the prototype. The proposed system achieved an effectiveness of 92.5% in size classification, which is notably higher than the 77% effectiveness achieved by operators. In color classification, the prototype achieved an effectiveness of 93.5%, while an expert is 100% effective. In conclusion, the implemented system significantly improves the classification process of tomatoes when compared to traditional methods. The system can also work for much longer continuous intervals, which is a significant advantage over an operator

• Autor Principal: Manosalvas Ramos, Cristian Christopher (author)
• Formato: bachelorThesis
• Idioma: spa
• Publicado: 2024
• Subjects: Visión Artificial; Procesamiento de Imágenes; Norma INEN 1745
• Acceso en liña: http://dspace.unach.edu.ec/handle/51000/12190