嵌入式全方位视觉跟踪器

Embedded Tracker Based on Omni-directional Vision

为了实现自引导车AGV(automatic guided vehicle)的自主导航,设计一种基于全方位视觉的嵌入式跟踪器。该跟踪器将鱼眼镜头与现场可编程门阵列(FPGA)、数字信号处理器(DSP)以及CMOS感光芯片相集成,实现一种小型化、结构化的设计。在此跟踪器上移植改进的均值漂移和粒子滤波相结合的跟踪算法,实现基于鱼眼图像的动态多目标跟踪,并通过与传统算法的分析、比较来体现改进算法的优越性。实验结果表明本文所研究的跟踪器具有良好的跟踪效果,满足实时性、准确性和鲁棒性的要求。

An embedded navigator is presented to achieve AGV automatic navigation in this paper.The Tracker integrated a fisheye lens,a field-programmable gate array（FPGA）,a digital signal processor（DSP） and a single complementary metal oxide semiconductor（CMOS） is compacted,smart and modular.An improved algorithm based on mean shift embedded particle filter is replanted into this embedded tracker for multi-object tracking.To find the advantages of our integrated method,we compared with the traditional algorithm.The experimental results indicate that the said tracker could tracking multiple targets well and be adequate to real time,accuracy,robustness.