Automatic Failure Detection and Correction for Real-Time Object Tracking with Kernelized Correlation Filter

Jungsup Shin, Heegwang Kim, Dasol Jeong, Joonki Paik

Abstract: This paper proposes an algorithm that estimates tracking failure and track the target object again. The proposed algorithm consists of three steps: 1) Tracking target object with KCF tracker. 2) Estimating tracking failure by analysing correlation values. 3) Re-capturing the target object using multiple search windows and re-tracking. Experimental results shows that the proposed algorithm effectively corrects tracking failure situations.