Vision object tracker

Abstract

This thesis details designing, implementing, and evaluating a flexible object-tracking solution for conveyor belt systems in partnership with Marel, a prominent food processing systems provider. This work, part of a B.Sc. final-year project, addressed the challenge of efficiently tracking objects on third-party conveyor belts. The project involved extensive research into various tracking methodologies, leading to the selection and implementation of the Simple Online and Real-time Tracking (SORT) algorithm. Initially validated using data from linear belts, the solution demonstrated its robust and real-time tracking capabilities. Comprehensive testing was conducted using a Unity-based conveyor simulation environment to emulate complex conveyor configurations. Significant contributions include the effective integration of the simulation environment for rigorous testing and the improvement of an image recognition model to maximize tracking accuracy. The final product is a high-performance, adaptable tracking solution with broad industrial applicability.