Robust Object Tracking Using Alternative Correlation Filters

Baskurt K. B., SAMET R.

25th Signal Processing and Communications Applications Conference (SIU), Antalya, Türkiye, 15 - 18 Mayıs 2017 identifier identifier

  • Yayın Türü: Bildiri / Tam Metin Bildiri
  • Doi Numarası: 10.1109/siu.2017.7960179
  • Basıldığı Şehir: Antalya
  • Basıldığı Ülke: Türkiye
  • Anahtar Kelimeler: visual object tracking, correlation filter, real-time video processing
  • Ankara Üniversitesi Adresli: Evet

Özet

This study proposes an unsupervised real-time object tracking method that is minimally affected from environmental conditions and target's appearance changes. Proposed object tracking is realized by two independent correlation filters which are estimating location and scale. Alternative correlation filters representing different appearances of the target are also used in order to increase robustness of the method across to scene and target changes. Sustainability of tracking is provided by putting into use alternative correlation filters when quality of tracking output decreases to critical level. Proposed method is tested on Object Tracking Benchmark (OTB) dataset and results are shared.