【作者】 李庆；

【导师】 刘上乾；

【作者基本信息】 西安电子科技大学 ， 物理电子学， 2006， 博士

【摘要】 随着红外焦平面阵列(IRFPA)技术的发展,IRFPA成像技术也获得了迅速的发展,并在成像制导等领域获得了广泛的应用,然而,阵列的非均匀性仍然是限制IRFPA成像系统性能的一个重要因素,因此,在实际应用中必须对其实施校正。由于在大动态范围下IRFPA响应的非线性较为突出,以及探测器的响应会随时间的推移和工作环境的变化而发生缓慢飘移,从而使得非均匀性校正技术成为具有挑战意义的课题。本文针对IRFPA器件在成像制导状态下所面临的特殊情况,即输入信号动态范围大、环境条件复杂多变及工作帧频高等问题,研究并提出五种基于场景的非均匀性校正算法及工程实现技术,解决或降低了上述问题对非均匀性校正的不良影响。具体为:针对焦平面阵列在大动态范围内的应用,提出了具有双重校正性能的校正算法——基于二次曲线模型的卡尔曼滤波校正算法,对焦平面阵列实现非均匀性和非线性双重校正;针对焦平面阵列应用在场景变化不充分的特殊环境,提出了基于图像配准技术的校正算法,所提出的算法不需要满足各探测器单元场景统计值相一致的条件,算法具有较好的收敛性;针对传统基于场景统计的校正算法存在的不足以及为满足实际应用需要,提出了基于场景统计的非均匀性校正新算法,它较典型的统计校正算法具有更好的收敛性,且计算量小,所需内存少,易于工程实现。另外,该算法也适合场景变化不充分下焦平面器件的非均匀性校正;针对多变的环境条件引起的焦平面器件响应特性和稳定性的变化,以及传统神经网络校正算法难以消除低频空间噪声的不足,在利用小波对由非均匀性导致的空间噪声进行频率特性分析的基础上,提出了两种改进的非均匀性校正算法——基于均值与神经网络相结合的校正算法及基于递归最小二乘法的校正算法。两种算法都具有较强的非均匀性校正能力,其中,前一种算法较简单,而后一种算法收敛性较好,跟踪能力强,且能有效地抑制“人工虚影”和目标退化现象的产生;此外,针对国产128×128IRFPA器件实际应用的具体要求,设计并研制了一套高帧频的IRFPA成像信号实时处理系统,以对IRFPA成像信号实现实时非均匀性校正等成像信号处理。所设计的成像系统具有信号动态范围大、运算速度快、处理精度高、可靠性能高以及系统灵活性和通用性强等优点。更多还原

【Abstract】 Along with the development of the infrared focal-plane array (IRFPA) technology, the IRFPA imaging technology has also obtained rapid development, and has obtained widespread application in imaging guidance domain. However, the array nonuniformity is still a important limiting factor for the performance of the IRFPA imaging system. Therefore, nonuniformity correction must be implemented in the practical application.Because the nonlinearity of the detectors’response under the large dynamic rank and the slowly and randomly drift with time of the detectors’response, all of the above make the nonuniformity correction become an even more challenging problem. Aiming at the peculiar problems which the IRFPA faces under imaging guidance condition, namely the large dynamic rank of input signal, the complex changeful condition & surroundings and the high work frame-rate and so on, five based-scene algorithms for nonuniformity correction and one technology for project realization are presented in this dissertation, and the bad affect of above problems on nonuniformity correction is solved or reduced. More precisely: Aiming at the application of focal-plane array in the large dynamic rank, a algorithm holding dual correction performance is presented---Kalman filtering algorithm for nonuniformity correction based on quadratic model, and can implement both nonuniformity and nonlinearity correction to IRFPA; Aiming at the application of IRFPA in the special environment of the inadequate scene change, a correction algorithm based on image registration technology is presented. The algorithm does not need to satisfy constant-statistics constrain for various detectors, and has a better convergence; Aiming at the shortage of the traditional correction algorithm based on scene statistics and in order to the actual need, a new correction algorithm based on scene statistics is presented. Comparing it with the typical algorithm based on scene statistics, this algorithm has a better convergence and smaller operation, and need less memories, and is easier to realize. Moreover, this algorithm similarly adapts the nonuniformity correction under the inadequate scene change conditions for IRFP; Aiming at the change of the response characteristic and stability of IRFPA caused by the changeful condition & surroundings, and the shortage which the traditional neural network algorithm eliminates the low-frequency spatial noise, on the basis of wavelet analysis for the frequency characteristic of the spatial noise caused by the nonuniformity, two improved correction algorithms are presented---the correction algorithm based on average and neural network combination, and the correction algorithm based on RLS. Both have the stronger nonuniformity correction ability. Thereinto, the former is simpler. The latter has a better convergence and stronger tracing ability, and can effectively suppress the"ghosting" and the target fade-out phenomenon; Moreover,更多还原