【作者】 缪军同；

【导师】 李吉成；

【作者基本信息】 国防科学技术大学 ， 电子与通信工程， 2006， 硕士

【摘要】 本论文根据红外成像系统信息处理的实际需求为背景,围绕红外图像的非均匀性校正技术展开研究,即在研究和分析常用的非均匀性校正算法的基础上,基于红外焦平面探测器的非线性特性,将“S”曲线拟合技术应用于红外图像的非均匀性校正,并进行了Matlab仿真;在归一化BP神经网络校正方法的基础上,结合非均匀性的特征分析,提出了BP神经网络的改进算法,最后应用Matlab对BP神经网络改进校正算法和归一化BP神经网络算法进行了计算机仿真实验,并进行了性能分析,为红外成像系统非均匀性校正算法的选择提供了理论支撑和准确的实验数据。全文共分为五章,具体内容安排如下:第一章:绪论。介绍课题背景和非均匀性校正技术的基本知识。第二章:对红外系统非均匀性产生的原因以及其表现特征进行了分析,并在此基础上给出了非均匀性的定义和度量方法,介绍了几种常用的非均匀性校正算法。第三章:基于红外焦平面探测器的非线性特性,将“S”曲线拟合技术应用于红外图像的非均匀性校正,并进行了Matlab仿真,实验结果验证了算法的有效性。第四章:研究和分析了归一化BP神经网络校正算法的收敛性、迭代步长及最大误差值的选择技术,在此基础上,论文提出了一种改进的BP神经网络校正算法,并进行了仿真实验。理论和实践结果表明:改进的BP神经网络算法克服了归一化神经网络不能有效消除低频)噪声的缺陷。第五章:运用本文讨论的校正算法对实际图像进行了校正,并对各种校正算法的性能进行了比较。最后为结束语。更多还原

【Abstract】 This paper make a deep research in nonuniformity correction of IRFPA detector, according to treating with the information in infrared imaging systems . The causations and features of nonuniformity in infrared imaging systems are described. Researches on several methods of NUC are made respectively. a new way for NUC on the basis detector’s response nonlinearity is used. Based on the simulation results, the algorithm behave more satisfyingly. The unitary BP algorithm is introduced, which has more adaptability than classical BP algorithm. The unitary BP algorithm’s effectiveness and superiority were verified by simulation .To avoid the disadvantage of the traditional spatial adaptive nonunifonnity correction denoising the low frequency dominated nonunifonnity noise,a new way is used .Simulation show that the proposed scheme can achieve good performance when low frequency noise is dominant .There are five chapters in this paper.the main work can be summarized as follows:In chapterⅠ, the technology of IRFPA’s for NUC is described in the first part of this paper.In chapterⅡ, the causations and features of nonuniformity in infrared imaging systems are described .Researches on several methods of NUC are made respectively.In chapterⅢ, a way for NUC on the basis detector’s response nonlinearity is used. Through analysis on the data obtained from simulation, the“S curve”and its arithmetic model are found, based on the simulation results, the algorithm takes the detector’s natural essence into account so can behave more satisfyingly.In chapterⅣ,the unitary BP algorithm is introduced, which has more adaptability than classical BP algorithm.Simulation were performed.The unitary BP algorithm’s effectiveness and superiority were verified by simulation .when low frequency noise is dominant ,a new way for NUC is used .Simulation show that the scheme can achieve good performance .In chapterⅤ, simulations on several methods of NUC are made respectively with really IR image several methods of nonuniformity correction (NUC) for FPA are analyzed and compared in detail.更多还原