Kalman Filter based Optimized Object Tracker with Auto-tuning of Process Noise Covariance Parameter and Performance Verification
Sk Babul Akhtar, Supriya Dhabal, P. Venkateswaran · 2023
Object tracking is a critical task in many computer vision and robotics applications. Kalman filters are commonly used for tracking, but their performance can be significantly improved by optimizing the tuning parameters of the system. In this article, an optimized Kalman filter object tracking approach is presented that automatically utilizes an algorithm to find the best tuning parameter for the filter. This article demonstrates how the Kalman filter can be improved by selecting an appropriate and proper process noise covariance matrix, and the verification of the chosen parameter is shown using sensitivity and robustness metrics. Firstly, an algorithm over a simple one dimensional tracking subject and also the proposed approach is evaluated on several standard object tracking systems. The results show that Kalman Filter with the correct process noise covariance is reliable even with low computation power compared to other methods in terms of tracking accuracy and robustness.