【作者】 谭晓宇；

【导师】 陈谋；

【作者基本信息】 南京航空航天大学 ， 模式识别与智能系统， 2008， 硕士

【摘要】 弱小目标的检测与识别,是近年来图像处理与目标检测识别领域的一个研究热点,也是一项难度大且有着重要战略应用价值的课题。本文主要研究图像中的弱小目标检测识别技术。通过图像处理,检测出弱小目标;通过信息融合,最终识别出弱小目标。首先,研究了含弱小目标的图像的基本特性,包括图像噪声模型、背景空间分布模型和弱小目标模型。同时研究了目标检测前的图像预处理方法。其次,研究了弱小目标检测的“先检测后跟踪”算法,主要用于复杂背景弱噪声图像中的弱小目标检测。针对背景复杂时,弱小目标检测不能通过单一的图像处理方法实现的特点,提出基于小波分析和信息融合的弱小目标检测算法和基于小波形态学的弱小目标检测算法。仿真结果表明这两种算法都能取得较好的检测效果。随后,研究了弱小目标检测的“先跟踪后检测”算法,主要用于背景简单强噪声图像中的弱小目标检测识别。针对使用常规的动态规划算法累积能量时效果较差的缺点,提出一种新的基于状态稳定性的动态规划能量累积算法。仿真结果结果表明,新算法较好的解决了目标能量累积时的能量扩散问题。接着,研究了弱小目标识别算法。针对有效信息较少,弱小目标识别困难且致信度低的特点,提出一种新的基于弱小目标速度、灰度、轨迹信息的识别方法。仿真结果表明,新方法明显提高识别置信度。最后,基于VC++和OpenCV语言开发了弱小目标检测识别软件,实现几种基本的目标检测识别算法和新设计的算法,同时也验证了新算法的正确性。更多还原

【Abstract】 The detection and recognition of dim and small target is a hot research subject in the field of image process and target detection and recognition. This research also has an important strategic application value. In this dissertation, detection and recognition of dim and small target in images are studied. Dim and small targets are detected by processing images and be recognized with information technique.Fist of all, the basic characteristics of the images with small targets are studied. These characteristics include the models of image noise, space background and small targets. The image pre-processing before targets detection is studied too.Then,“Detect-Before-Track”algorithm is studied. It is used for detecting dim and small targets in complex background images with week noise. As the dim and small targets can not be detected with a single image processing technique, two new methods of dim and small targets detection are proposed. One is base on the wavelet transform algorithm and information fusion technology. The other is base on the wavelet transform algorithm and mathematical morphology. Simulation results show that the two new algorithms can obtain the good detection results.Following,“Track-Before-Detect”algorithm is studied. It is used for detecting dim and small targets in simple background images with strong noise. In allusion to the flaws of general dynamic programming algorithm with maximizing energy, a new improved method is proposed, which is based on the most steady states. Simulation results show that the proposed algorithm can well solve the energy scattering problem.And then, recognition algorithm of dim and small targets is studied. Because of the lack of the small targets’information, it is difficult to recognize small targets and the recognition reliabilities always are low. A new improved method of small targets recognition is proposed,which is base on the the information of targets’velocity , gray and tracks. Simulation results show that recognition reliabilities are raised obviously.Finally, a kind of software of detection and recognition of dim and small target is developed with VC++ and OpenCV. Several basic dim and small targets detection and recognition algorithms and new methods are implemented by it. The effect of these new methods in this dissertation is validated by using the software.更多还原