Bibliographic citations
Perez, S., Rosell, L. (2017). Elección de características de interés en la clasificación de granos de café mediante un sistema de visión por computadora [Tesis, Universidad Privada del Norte]. https://hdl.handle.net/11537/12650
Perez, S., Rosell, L. Elección de características de interés en la clasificación de granos de café mediante un sistema de visión por computadora [Tesis]. PE: Universidad Privada del Norte; 2017. https://hdl.handle.net/11537/12650
@misc{renati/514206,
title = "Elección de características de interés en la clasificación de granos de café mediante un sistema de visión por computadora",
author = "Rosell Llanos, Luis Adrián",
publisher = "Universidad Privada del Norte",
year = "2017"
}
ABSTRACT The classification of quality on coffee before toasting, one of the most important operations to define its quality and price in the market, is done manually by personnel trained in the recognition of coffee defects. However, the subjective nature, the cost and the time that it involves generates an important research field for the application of technologies such as artificial vision. The objective of this study was to evaluate the ability to identify defects and classify coffee beans using a computer vision system in the red-green-blue (RGB) space. For this purpose, a system for acquisition and analysis of images was implemented, developing a computer application in Matlab 2015ª. Samples of green coffee were purchased on local market, each grain being classified according to NTP 209.027 2001. Images of each class were acquired and analyzed by determining in each grain six shape parameters, six color parameters, in the RGB and HSV spaces, and two normalized indices or differences. Statistical relevance of parameters was deterined using the software for data analysis named WEKA and using these three models of classification, vector machines (SVM), decision trees and nearest K-neighbor (K-neighbor), were implemented. The three types of classifier used in the present investigation show accuracy between 89% and 92.3% which probe the possibility to implement systems based on RGB image to classify coffee been.
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.