Prediction of the main quality parameters of fishmeal to automate the drying process using hyperspectral imaging and artificial neural networks

Moscol I.; Peltroche G.; Ruesta V.; Sanchez J. · Proceedings of the 2020 IEEE Engineering International Research Conference, EIRCON 2020 (2020) · DOI: 10.1109/eircon51178.2020.9254070

Resumen

Hyperspectral Imaging (HSI) is used to detect parameters showing further information than conventional instruments. This information is represented as a spectral signature, whose reflectance values can be applied in an artificial neural network (ANN) for detection, classification and prediction of the most important parameters in fishmeal. Currently, in many fishmeal companies there is a lack of control aimed at real-Time detection of quality parameters such as protein, moisture, ash, fat, among others, whose correction is feasible during production. The problem is established at the exit of the drying process, where the optimization of the final quality parameters is possible through the controlled re-entry of the product into the dryer to reduce moisture to the optimum percentage, in a homogeneous way and without burns on the product. The present investigation applies the methodology of Hyperspectral imaging (HSI) and Artificial Neural Networks (ANN) to solve the problem by developing an algorithm able to predict the main parameters to determine the quality level of fishmeal in a non-invasive way giving the possibility of improving the drying process. As a result, a good correlation between reflectance, provided by Hyperspectral Imaging, and the main quality parameters is obtained after implementing a Multilayer Perceptron Neural Network algorithm.

Instituciones: Universidad de Piura