Vegetable Visual Quality Evaluation System Based on Artificial Intelligence

Abstract

Thanks to the rapid development of neural networks in recent decades, applications for this technology have been found in various fields, from medicine to waste management. The same applies to agriculture, where artificial intelligence enables agribusinesses to make decisions based on objective statistical data and through that helps to increase the productivity of the businesses. An example of precision agriculture is the visual quality evaluation of vegetables with the use of machine learning based classifiers. This thesis aims to make such tools more accessible and affordable for small agribusinesses as existing solutions are generally too expensive or cannot be easily integrated into existing processing lines. A new system, Vegeval, is designed and developed to overcome these issues and to provide real-time statistics to agribusiness owners about the quality of their produce. With the use of edge computing, it is shown that a relatively inexpensive system can be built for a hassle-free adoption of precision agriculture processes in existing vegetable processing lines. Consequently, based on the results of the thesis, it can be observed that hardware with low computing resources can successfully be deployed for fulfilling computer vision and object detection tasks in the discussed use cases. The latter additionally indicates that applying artificial intelligence to make everyday tasks more efficient does not necessarily have to come at a large expense.