红外点目标定量检测关键技术研究

发布时间：2018-01-07 23:14

本文关键词：红外点目标定量检测关键技术研究　出处：《中国科学院大学(中国科学院上海技术物理研究所)》2017年博士论文　论文类型：学位论文

【摘要】：点目标探测一直是各种军用、民用卫星遥感领域中较为活跃的课题,其中提升星载红外系统的点目标检测能力在我国航天事业中扮演着重要角色。由于探测点目标的特殊性,通过仪器特性参数和数据处理算法的优化提升其检测能力还存在很大空间,尤其是目前谈论的检测在目标的定量程度上还偏低。针对该薄弱环节,本文通过成像仿真模型和方法详细分析了工作波段以及采样体制对点目标检测性能的影响,提出了改进的单帧检测算法以及一种在先验信息匮乏情况下的亚像元特性反演算法,打破了常规检测问题的枷锁,从不同角度探讨了提升点目标检测定量程度的方法,探索性地研究了亚像元点目标的定量检测问题,相关研究成果可以为星载目标探测系统的研制以及其数据分析解决方案的设计提供参考。论文的主要研究成果和创新点是:(1)在星载目标探测系统的波段选择分析方面,对目标、有云和无云情况下地-气背景的红外辐射特性进行了详细分析和仿真计算,设计了基于最优信杂比的波段选择方法,针对目标的早期发现,得出了遥感器入瞳处的、不同杂波条件下的最优探测波段。(2)在星载目标探测系统的采样体制分析方面,建立了亚像元目标的采样模型,弥补了现有研究方法在数值仿真方面的不足;多方面分析、验证了两倍瞬时视场过采样的性能优势和劣势,提出了该过采样对目标信号和目标信杂比实际上是有抑制作用,不适用于低信杂比条件下的目标探测;提出了过采样在量化目标辐射强度方面的“过采集”误差。(3)在点目标单帧检测算法方面明确提出有效像元检测;提出了一种基于复杂背景可分性度量的信杂比准则用于目标的检出,该方法在具有阶跃特征的结构化背景中的性能更优;构造了马尔可夫随机场(MRF)对欧式空间距离的标号场概率响应模型,并通过高阶能量函数提高了目标概率对邻域标记变化的响应能力。改进方法在目标空间-辐射维度的检测能力更强。(4)在目标特性反演方面,将目标检测问题从“定性”提升到“定量”的层面,在一定假设条件下,提出了一种在先验知识匮乏情况下反演点目标亚像元特性的算法,并通过成像链路正演模型验证了反演算法可以比较准确地估计亚像元目标的温度和面积比例。
[Abstract]:Point target detection has been an active research topic in the field of military and civilian satellite remote sensing, which point target detection capability of spaceborne infrared system in China's aerospace industry plays an important role. Because of the particularity of point target detection, by optimizing the parameters and data processing algorithms to improve the detection ability of instrument characteristics still exist a lot of space, especially in the detection of the current talk about quantitative target is also low. In view of the weak link, the effect of imaging simulation model and method is analyzed in detail and the band sampling system on the detection performance of point target, put forward the single frame detection algorithm and a priori information under the condition of the lack of sub pixel characteristic inversion algorithm, to break the routine detection of chains, discusses the method of lifting point target detection quantitative degree from different angles, the exploration of the Study on the quantitative detection of the subpixel target, analysis of related research results can be developed for spaceborne target detection system and its data to provide a reference solution design. The main research results and innovations are: (1) in spaceborne target detection system for band selection analysis, the target infrared radiation the characteristics of the cloud and cloudless conditions - gas background was calculated with analysis and simulation, the design method to select the optimal SNR band based on the early detection of the target, the remote entrance pupil, the optimal conditions of different clutter detection band. (2) in spaceborne sampling system target detection system analysis, establish the sampling model of subpixel target, making up for the deficiency of the existing research methods in numerical simulation; multi analysis, verify the performance advantage over two times the instantaneous field sampling And put forward the disadvantages, over sampling of the target signal and the target SNR is actually inhibited, is not suitable for low SNR conditions of target detection; proposed over sampling error in quantitative acquisition "target radiation intensity". (3) proposed effective pixels detection in single point target frame detection algorithm; presents a complex background separability measure based on signal-to-noise ratio criterion for target detection, this method has the properties of the structured background step characteristic of better; construct a Markov random field (MRF) on the European Space label field response probability model, and improve the response the ability to target the probability of neighborhood change marker through higher order energy function. The improved method of target space dimension radiation detection capability. (4) the target characteristics inversion, will enhance the target detection problem from the "qualitative" To "quantitative" level, under certain assumptions, the sub-pixel characteristics of an inversion under the condition of the lack of prior knowledge of point target imaging algorithm, and through the forward link model proved that the algorithm can accurately estimate the temperature and area proportion of sub pixel target.

【学位授予单位】：中国科学院大学(中国科学院上海技术物理研究所)
【学位级别】：博士
【学位授予年份】：2017
【分类号】：V443.5;TN219

【相似文献】