红外图像增强和目标检测跟踪技术研究

【作者】 畅博；

【导师】 李晓峰；

【作者基本信息】 电子科技大学 ， 通信与信息系统， 2006， 硕士

【摘要】 红外成像制导是当今精确制导技术中使用最多的一种,是当今红外技术在军事上的主要应用之一。制导系统涉及面广,包括红外物理学,随机信号处理、图像分析以及高速数字信号处理等多门学科。其中,利用红外成像技术来实现对目标的检测和跟踪,是军事武器现代化、自动化和智能化的重要标志之一。它通过对探测器输入的红外图像序列的实时性处理,来完成目标的低信噪比的检测和目标的精确跟踪等任务。本文主要的研究内容就包括针对红外图像进行图像增强预处理的具体方法和技术,对于预处理后的图像序列进行目标检测和跟踪的具体方法和技术。本文首先讨论了红外辐射基本特性以及目标和背景的红外辐射特性的基本区别,进一步探讨了热成像仪的红外图像特征,归纳总结出来了红外图像的主要特点,为后续工作打下良好的理论基础。论文在红外图像预处理方面,首先分析了传统的空间滤波技术,并着重介绍了平滑空间滤波器。其次,从直方图特别是红外直方图的特点入手,分析了传统的直方图均衡方法。并在此基础上,对平台直方图增强算法中自适应求取平台值方法和基于遗传算法的红外图像增强算法进行了研究。论文在红外目标检测方面,以对经过预处理的图像进行分割为切入点,采用二维类间方差法和局域生长法来进行图像的分割,消除了原始图像中一些不可能为目标的孤立噪声点和大面积的背景区域。对于经过分割以后的目标图像,由于存在噪声和其他干扰,因此很难利用单帧信号来实现对目标的可靠检测,故此利用序列图像中上下多帧的相关信息,实现对目标进行有效的检测。论文在目标跟踪方面,首先是针对跟踪系统的功能介绍,分析了图像目标跟踪的机理和实现条件。其次介绍了几种常用的跟踪算法,如波门法、匹配跟踪法以及目标位置预测法等。最后在此基础上,进行模板匹配跟踪算法的研究。着重针对初始模板的获取,匹配算法的修正、自适应更新模板和模板更新的阈值要求等方面进行研究,从而实现对图像序列进行目标跟踪。更多还原

【Abstract】 IR imaging guidance is the most wildly used among precise guidance technique and is one of application in military affairs of IR technique. The guidance system involved many areas such as IR physics, random signal processing, image analysis, high speed digital signal processing and so on. In them, the target detection, and tracking using the IR imaging technique is one of the symbol of modernization, intelligent, autoimmunization of modern weapon. It fulfills the mission of the lower s/n target detection and the precise tracking by use of real time process of IR image serial which input by detector. The main content in this paper is about the idiographic methods and research of IR image enhancement, target detection and tracking of IR imaging series and some achievements have been gotten.First, this paper discusses the basal characteristic of IR, the difference of radiation of target and background and the characteristic of infrared thermal imager. Then this paper concludes the characteristic of IR image and builds the academic foundation.In the facet of pre-transact of IR image, the traditional smooth spatial filter and histogram equalization are analyzed and intruded first. Then the research of image enhancement based on of Plateau equalization algorithm and Genetic Algorithm is done then in this chapter.Image segmentation with algorithm of 2D-maximal variance between-class and region growing is accepted to eliminate the background and the isolated noise points which can not be the target. The image after segmentation still has the noise and the other disturber, it is hard to detect by only one frame image. So the method by use of the relation by frame series is accepted to detect the true target.The function of tracking system is described in the facet of target tracking. The mechanism and implement condition are analyzed then. After the introduction of traditional algorithms, the algorithms of template matching is adopted to accomplish the target tracking and some emphases have been put on the field such as the first obtain template, the renovated template, the modified matching algorithms and the requirement of modification in the process of target tracking.更多还原