Journal of Mineral and Material Science
[ ISSN : 2833-3616 ]
Flotation Froth Monitoring Using Unsupervised Multiple Object Tracking Methods
1FRC CSC of the RAS, Lead Data Scientist, Norilsk Nickel, Moscow, Russian Federation
2Head of AI Solutions, Norilsk Nickel, Moscow, Russian Federation
3Moscow Institute of Physics and Technology, Data Scientist, Norilsk Nickel, Moscow, Russian Federation
Corresponding Authors
Keywords
Abstract
The popularity of computer vision algorithms applied to the metals and mining industry has grown drastically in recent years. This article will cover the application of computer vision models, video processing techniques, and methods of tracking many objects without data labeling (so-called unsupervised multiple object tracking) using the flotation froth, by example. In more detail, you will find in this article description of this kind of data domain as well as some words about the flotation process, the segmentation approach for many similar objects and an approach to its simultaneous tracking, an overview of existing tracking methods without data labeling and quality metrics comparison of these methods.