Implementation of MOSSE Object Tracking Algorithm on FPGA with High Level Synthesis Approach

Abstract

Video object tracking is significant topic of image processing aplications. A video tracking algorithm is should to track object in condition, variable light intensity, object occlusion , variable object size and variable object orientation. Considering these requirements, hardware with high processing capacity is needed. Hardware with high processing capacity creates disadvantages in applications such as mobile robot, flying vehicle, due to power consumption and hardware size. FPGA provides high performance in image processing applications by parallelism, processing big data capability and flexibility depending on design. In this paper, hardware architecture of Correlation filter based MOSSE video object tracking algorithm is desinged and designed algorithm is implemented on FPGA with high level synthesis aproach. Implementation of proposed hardware architecture on a single FPGA chip will create advantageous in applications where the payload is limited, such as mobile robots and unmanned aerial vehicles.