Real-time object tracking in a video stream using Field Programmable Gate Array

摘要

This paper aims at merging the best of both the technologies viz. Computer Vision and Embedded Systems by implementing a system which tracks a moving object in a live video stream using a Field Programmable Gate Array (FPGA) rapid prototyping platform. The system tracks an object in a real-time 240×320 USB-Video-Class (UVC) video stream offering portability. The tracking target selection is enabled using a touch-screen interface. The design is implemented on a 32-bit RISC soft-core processor available in FPGA and parts of the system are accelerated in reconfigurable hardware thus leveraging the hardware and soft-processor co-design capabilities offered by the FPGA. The HSV (Hue, Saturation, Value) color model is used to make the algorithm robust to changing lighting scenarios. In addition, the compute-expensive Color-Space transformation module is implemented through the use of pre-initialized look-up tables for accelerated performance. Finally, a simple, but robust algorithm is used to track a moving object using averaging, dynamic thresholding and center-of-mass model for updating the current location of the target object.