Desarrollo de un prototipo para el control de calidad de la carne bovina determinada por sus características organolépticas, basado en un sistema automático de inspección por visión artificial.

Abstract

The prototype for the quality control of the beef determined by its organoleptic characteristics based on an automatic system of inspection by artificial vision, was developed with the objective of guaranteeing less manipulation and estimation of time by human intervention in the sensorial meat valuation process. For the execution of the project a linear transportation system was considered, using inductive sensors to determine the location of the tray along the system and a laser sensor that detecs the presence or absence of meat. The development of the artificial vision system (AVS) was implemented on a computer Raspberry Pi (RPi), with a programming language multi-paradigm (Python) using OpenCV libraries. The acquisition of images was done with the RPi camera and a side lighting system. The Gaussian method was used for noise cancellation prior to the development of the Otsu technique and morphological operations for the identification of color and texture parameters, which allowed a better description of the images. A data base witch was classified into five color and texture scales respectively was used to categorize the characteristics. The system evaluation was carried out with a sensorial analysis comparison of fifteen samples, between a panel of five selected, semi-expert judges and the prototype which showed a time-save of approximately 67%compared to the human evaluation with a margin of error of 23.4% in the evaluation. Therefore, the implemented prototype proposed a non-invasive objective method, with smaller time intervals of evaluation. This shows any future evaluator panels could be replaced by a single user. The use of fast processors is recommended to minimize execution times and to perform an exhaustive research on color spaces for better estimation of characteristics in foods.