Multi-object tracking for mobile navigation in outdoor with embedded tracker

摘要

Robotic tracking is a key problem in the robotic system which is more challenging in outdoor environment. This paper presents a new type of embedded smart tracker with a method of Omni-vision tracking. The central components the smart tracker equipped with are: a Digital Signal Processor (DSP), a Field-Programmable Gate Array (FPGA), a CMOS image sensor and a Fisheye lens. The natural target we choose is the overhead street lamps. An improved tracking method based on Mean Shift Embedded Particle Filter is transplanted into the embedded platform to realize the stable and real-time multi-objective tracking. Experimental results show that compared with traditional tracking algorithm, the tracking instruments have good tracking performance, basically meet real-time performance, accuracy and robustness requirement.