红外/可见光跟踪算法的研究及实现

【作者】 张雪；

【导师】 吴援明；

【作者基本信息】 电子科技大学 ， 信号与信息处理， 2007， 硕士

【摘要】 图像跟踪器是图像制导系统的核心部分。传统图像跟踪器用电视摄像机作信号传感器,而通常的可见光电视不能在夜间工作,不能“看到”烟雾遮蔽的目标。现代先进图像跟踪器系统中采用红外热像仪成像装置,克服了以上缺点,大大扩展了图像跟踪器的作用时域。本论文的图像跟踪器采用CCD和红外成像传感器进行目标跟踪,它可有效地提高系统的全天时/全天候工作能力,有着广泛的应用前景。另外,早期的图像跟踪器一般采用简单的跟踪算法,只有单一工作模式,只能对一定条件下的目标进行跟踪,不同的应用环境,跟踪的效果会有很大差别。随着光电转换器件和高速实时图像处理技术的迅速发展,图像跟踪器向着多模式跟踪发展。本文改进了传统的相关和对比度跟踪,并采用二者相融合的双模式跟踪算法,有效提高了跟踪置信度和稳定度,拓宽了图像跟踪器的应用环境。本课题的来源是西南技术物理研究所开展的XX红外吊舱图像跟踪器的研究项目。该图像跟踪器以高性能的DSP(ADSP-TS201) + FPGA(Stratix系列的EP1S40F1020芯片)为核心,以高效、优化的软件设计,实现实时、高精度、高稳定的图像跟踪。本论文在总结国内外研究现状的基础上,基于DSP做了大量研究和调试工作,使图像跟踪器的功能得以实现,并在硬件系统上联调成功,在外场挂飞试验中取得了很好的效果。论文的工作主要体现在以下几点:1.研究CCD和红外数字图像信号的特性,及可能遇到的实际情况,设计图像预处理算法,对跟踪效果起到加强的作用。2.改进了传统的相关跟踪算法,设计了按列模板刷新策略以及基于连通域的相关目标提取算法,解决了相关跟踪存在的瞄准点漂移等问题。3.改进了基于对比度的形心跟踪算法,设计了目标极性判断以及多个对比度目标的提取和判断方法。4.设计并实现了双模式跟踪算法:将相关跟踪和基于对比度的形心跟踪算法融合,得出了最优的跟踪结果。双模式跟踪算法使图像跟踪器在跟踪过程中能够根据目标信息智能地选择跟踪算法,大大提高了图像跟踪器的综合性能。更多还原

【Abstract】 Image tracker is the core component of image guiding system. Traditional image tracker uses CCD as signal sensor, but normally CCD can’t work at night or can’t "see" the targets obscured by smoke. Modern advanced image tracker which uses infrared image device can overcome these weaknesses, so the applications of image tracker are greatly expanded. For CCD and infrared imaging sensor used in image tracker, the all-weather working capability of the system has been effectively improved. Additionally early image tracker usually works in a single work model with simple tracking algorithm, so it can only work under certain conditions; even more the tracking effects are greatly different in various environments of application. With the rapid development of photoelectric convector and high-speed real-time image processing technology, image tracker inclines to multi-mode way. In this dissertation traditional correlation and contrast tracking algorithms are modified, and a dual-mode tracking algorithm combined by these two models is designed. The accuracy and stability of image tracker, which can work in complex environments, are effectively improved.The dissertation comes from the research on project XX infrared image tracker of South-West Technology and Physics Institute. Based on high-performance hardware platform of DSP (ADSP-TS201) + FPGA (The EP1S40F1020 of Stratix) and structural design of software, the image tracker realizes real-time, high-precision and high-stability image tracking. After summarizing the modern situation of research at home and abroad, then debugging DSP, the functions of this image tracker comes true finally. The main points are listed below:1.The image processing algorithm is designed to strengthen tracking effects according to the characteristics of CCD and infrared digital image signals.2.The traditional correlation tracking algorithm is modified, then a template column-changing algorithm by and correlation targets extraction algorithm based on connecting areas are designed. These strategies overcome the tracking points drifting of correlation tracking.3.Centroid tracking algorithm based on contrast is modified, and judgment methods of objectives’polarities, as well as extraction and judgment methods of several contrast targets, are designed.4.Dual-mode tracking algorithm is designed and realized: correlation tracking and centroid tracking algorithms are combined to obtain optimal tracking effects. The image tracker which uses dual-mode tracking algorithm can choose tracking algorithms smartly according to the information of targets, so the performances of the image tracker are greatly improved.更多还原