【作者】 常洪花；

【导师】 张建奇；

【作者基本信息】 西安电子科技大学 ， 光学工程， 2006， 博士

【摘要】 随着光电成像技术的不断发展，现代成像系统能够在战术有效距离内显示出越来越具体的目标细节和场景内容，在大大改进了成像距离和成像清晰度的同时，也使得“场景内容”成为影响光电成像系统目标获取性能不可忽视的重要因素。这些干扰目标探测的场景内容就是所谓的背景杂波。而传统的成像系统性能表征模型对背景杂波效应没有考虑或考虑不足，因而，非常有必要对其加以改进，以适应不断发展的新型光电成像系统性能评估的需要。 本论文主要在以下几个方面进行了研究工作：（1）基于人眼视觉对图像结构信息的高度自适应性，综合考虑背景和目标在平均强度，对比度及强度分布方面的相似性，通过构建数学统计模型的方法来衡量背景杂波的强弱，提出了一种新的杂波量化描述尺度——目标结构相似性（TSSIM）尺度，该尺度既有传统数学统计方差尺度简单易行的优点，又充分考虑人眼在探测目标时的基本感知特性。由于该尺度对场景内容没有特定的限制，因而可广泛应用于各类光电图像。同时，借助于误差分析和相关性分析理论，从两个方面比较分析了该杂波尺度与现有杂波尺度，一方面是比较分析其与观察者试验目标探测概率的相关性；同时也比较其与目标平均探测时间的相关性。从而充分验证本文所提出的杂波尺度与人眼实际感知结果的一致性。（2）以目标结构相似性尺度为背景杂波度量因子，以荷兰TNO人类因素研究所提供的Search₂试验数据库为基础，通过构建数学模型的方法，定量分析了背景杂波对观察者目标虚警概率的影响，并在此基础上，讨论了人眼视觉在背景杂波干扰下的虚警概率表征模型。（3）基于NVESD经典性能模型在背景特性表征方面的基本特点，以人眼视觉搜索探测特性领域的最新研究成果为理论依据，借助于相关性分析理论，确定对观察者目标探测概率具有最佳预测性能的信杂比表征尺度，以该尺度作为背景杂波修正因子，对NVESD静态目标探测概率模型进行修正；以同样的方法选择目标平均探测时间的背景杂波修正因子，对NVESD搜索模型进行修正，最终得到定量描述背景杂波效应的人眼视觉目标获取性能修正模型。同时，利用现场实验数据对修正模型和原模型预测结果的准确性进行了比较，从而以实际试验证明了修正模型的合理性。（4）初步构建包含背景杂波效应的机器视觉成像系统目标探测性能模型。本文以系统信号干扰比（SIR）为理论基础，引入巴特沃斯模型功率谱密度（PSD）模型来描述背景杂波，综合考虑光学系统、探测器、电子线路的功率传递函数、目标辐射强度和噪声信号的数学统计特性，及探测算法对杂波和系统噪声的抑制作用，初步建立了考虑背景杂波效应的机器视觉目标探测性能模型。更多还原

【Abstract】 With the development of electro-optical technology, imaging systems are capable of showing some internal detail on targets and some more detail in the scene at tactically significant ranges. While these changes improve the imaging range and resolution, they make the content of the scene become one of the most important factors affecting target acquisition performance. The scence content, which interferes with target acquisition performance, is defined as background clutter. Traditional performance characterization methods or models become insufficient for this type of imaging systems since they only consider a little or even no background clutter. Hence, it is necessary to improve traditional performance evaluation methods to satisfy the performance evaluation of the novel imaging system. In this paper, the following works are studied.（1） Based on the research trying to understand how human eyes perceive clutter in background, we propose a new clutter metric which estimates the degree of clutter in an image by comparisons between the target and background areas in the following aspects: intensity, statistical variance and structure. We define this metric as target structure similarity measure （TSSIM）. The metric can be easily calculated and efficiently consider the human eye perception. Since requiring no specific properties about the viewing conditions, they can be easily used in various kinds of electro-optical images. Error and correlation analysis are used to experimentally validate this metric. Attempts to correlate this metric output with both experimental target detection probability and detection time yield rather good results.（2）Based on the Search₂ dataset provided by TNO Human Factors Research Institute of Netherlands, via a simple mathematical model, the target structure similarity measure is used to quantitatively describe the effect of background clutter on the human false target detection probability. Furthermore, a human false target detection decision behavior model in cluttered environments is established.（3） Based on the theory of correlation analysis and the up-to-date research findings in human visual perception, we present a new modifying method of the classic NVESD target acquisition model with more comprehensive treatment of the target signature and background clutter relationship. At the same time, the Search₂ dataset is used to compare the prediction capability for experimental target detection probabilities of the modified model to the classic one. The former shows much better predictiveperformance than the latter, and reasonability of our modification method is validated.（4） Background clutter seriously affecting the target detection performance of the machine-vision-based imaging systems, such as IR seeker. We present a robust performance evaluation technique for staring IR seekers based on signal-to-interference ratio （SIR）, with the quantitative description of the background clutter emphasized. The power transfer functions of the optical system, detector and electronic are established to describe energy transmission of the signal and interference （noise and clutter） through the IR seeker, and the targets’ radiant intensity statistics as well as the noise’s statistical characteristics are also taken into account. In order to quantify the background clutter, we use a clutter measure based on its energy content —power spectral density （PSD）. Based on this measure, a SIR is developed to analyze detection performance.更多还原