Food Quality Detection with Convolutional Neural Network

Authors

  • Abed Atassi McGill University
  • Aurelia Haas McGill University
  • Lukas Durand McGill University
  • michel Cantacuzene

Keywords:

Food Quality Detection, CNN, Computer Vision, Multispectral Camera, Image Recognition

Abstract

Food health and safety have always been a major concern for producers and distributors alike. From farm to table, nearly all store-bought fresh products are evaluated in multiple stages before reaching the end consumer. The industry has continuously adapted new technologies in order to increase efficiency and throughput while also emphasizing quality control. With the development of computer vision techniques and convolutional neural networks, many papers have investigated computerized identification and classification of produce freshness, but these mainly incorporated traditional RGB cameras with mixed results. Knowing that the agricultural industry has been embracing multispectral cameras with drones to survey crop health, we will identify and propose a method for the identification and classification of agricultural produce using images from multispectral sources. This project is separated into two main sections. Firstly, the choice of the convolutional neural network model and an investigation into its scalability and ability to increase the range of produce handled, as the industrial application of computer vision for agricultural produce requires retraining and extension of models. Secondly, we investigated the grading of quality based on images from a multispectral camera source. The first semester goals were to secure funding for a multispectral research camera, and give the members of this team a thorough understanding of the different convolutional neural networks for image recognition, while also determining problems we would encounter further on; The second semester goals were to create a multispectral camera able to take pictures of produces with different wavelength lenses and, by using the CNN models previously studied, analyze these pictures to obtain a certain accuracy of image recognition. A significant amount of testing for both CNN and Multispectral phases was conducted in order to find the most efficient and precise algorithm and capture process to meet our project’s needs.

Additional Files

Published

2021-04-15

How to Cite

Atassi, A., Haas, A., Durand, L., & Cantacuzene, michel. (2021). Food Quality Detection with Convolutional Neural Network. International Journal of Computational and Biological Intelligent Systems, 3(1). Retrieved from https://ijcbis.org/index.php/ijcbis/article/view/1639