High-Precision Visual-Tracking using the IMM Algorithm and Discrete GPI Observers (IMM-DGPIO)

Abstract

In this work, we propose the integration of a bank of Discrete Generalized Proportional Integral Observers (DGPIO) within an Interacting Multiple Model (IMM) structure in order to improve the precision of visual-tracking tasks. Applications such as visual servoing, robotic assisted surgery and optronic weapon systems require accurate and high-precision measurements provided by real-time visual-tracking systems. In this case, the DGPIO-Bank was designed using two kinematic models based in constant velocity (CV) and constant acceleration (CA) motion profiles. The main feature of the DGPIO-Bank is the active disturbance rejection (ADR) feature which reduces noise in the position signal of a moving object. The resultant algorithm uses a fusion of four important features: state interaction, Kalman filtering, active disturbance rejection and multiple models combination. For performance comparison, we evaluated our proposed IMM-DGPIO algorithm and other state of the art IMM algorithms. Experimental results show that our proposed strategy had the best performance.