Efficient Hardware Implementation of Real-Time Object Tracking

Njuguna J. C., Alabay E., ÇELEBİ A., Celebi A., Gullu M. K.

30th Signal Processing and Communications Applications Conference, SIU 2022, Safranbolu, Türkiye, 15 - 18 Mayıs 2022 identifier

  • Yayın Türü: Bildiri / Tam Metin Bildiri
  • Doi Numarası: 10.1109/siu55565.2022.9864768
  • Basıldığı Şehir: Safranbolu
  • Basıldığı Ülke: Türkiye
  • Anahtar Kelimeler: FPGA, hardware architecture, HLS, MOSSE tracker
  • Kocaeli Üniversitesi Adresli: Evet

Özet

This paper presents a novel hardware architecture for the minimum output sum of squared error (MOSSE) tracker algorithm. The proposed hardware architecture is built using a high-level synthesis (HLS) approach to generate hardware blocks from C++ code. Furthermore, real-time testing and verification are done using an FPGA device fabricated at a 16nm technology node. The microbolometer thermal imaging sensor which has 17 μm pixel pitch, 640×480 video resolution and 30 frames/second (fps) frame rate is used. Using the HLS tool, the maximum clock frequency and frame rate of 300 MHz and 60 fps are achieved respectively. Generated register transfer level (RTL) and Matlab simulations pose negligible differences between hardware and software implementations. Experimental results reveal that the proposed architecture balances hardware resource utilization and target tracking accuracy for consumer electronics applications.