Long-term target tracking combined with re-detection

Abstract Long-term visual tracking undergoes more challenges and is closer to realistic applications than short-term tracking. However, the performances of most existing methods have been limited in long-term tasks. In this work, we present a reliable yet simple method, which extends state-of-the-art learning adaptive discriminative correlation filters (LADCF) algorithm with re-detection component based on support vector machine (SVM) model. The LADCF localizes target each frame, re-detector able efficiently re-detect whole image when fails. We further introduce robust confidence degree evaluation criterion that combines maximum response average peak-to-correlation energy (APCE) judge level predicted target. When generally high, SVM updated accordingly. If drops sharply, re-detects perform extensive experiments OTB-2015 UAV123 datasets. experimental results demonstrate effectiveness our