【作者】 徐军；

【导师】 梁昌洪；

【作者基本信息】 西安电子科技大学 ， 电磁场与微波技术， 2003， 博士

【摘要】 本论文课题来源于国家自然科学基金项目“基于背景预测的红外弱小目标检测技术”（60007007）。红外图像中的弱小目标检测技术在军事领域已成为成像制导、告警系统的核心技术。本文从图像处理方法的角度对红外图像中弱小目标的检测进行了研究，分析了噪声、背景和目标的图像特征。在单帧图像和序列图像目标检测方面提出了基于背景预测的检测方法，最大化背景预测方法、最相似背景预测方法、位移式管道滤波方法、基于亚象元分析的目标运动特征算法和目标对比度相似性的帧间目标检测方法。以上算法无论在理论研究还是实际应用方面都具有重要价值。在检测效果评估方面首次推导了检测概率、虚警概率与信噪比、对比度和检测门限之间的关系。 所有的分析都结合了实际的红外图像进行仿真实验，并从实验结果验证了各种算法的优异特性和各种结论。更多还原

【Abstract】 The project of this thesis is supported by the National Science Foundation of China, named "Detection of small and dim target in infrared based on background prediction" (60007007). The small and dim target detection technique in infrared images has already become the key technology of the image guidance and the warning system.This paper study the detection of the small and dim targets in the infrared images from the view of image processing , analysis the imagery characteristics of noise, background and target. Detecting algorithm based on background prediction, maximum background prediction, most similar background prediction, displacement pipeline filter, target movement characteristics algorithm based on sub-pixel analysis and the method of the contrast similarity among frames are introduced for the target detecting in single image and sequence images. The above algorithms have very important academical value in either theory or in practicability. In first time, the relationship between probability of detection and probability of fault to SNR, contrast and detecting threshold is derived in the assessment of detecting result. All of the analysis are carried on the emulation experiment with the real infrared images, and are verified the excellent characteristics of various algorithms and conclusions from the experimental result.更多还原