【作者】 张宝辉；

【导师】 常本康； 钱芸生；

【作者基本信息】 南京理工大学 ， 光学工程， 2013， 博士

【摘要】 红外与可见光图像融合技术能够提取不同波段图像的特征,增强人们在探测或者检测领域对图像的认知和理解能力。因此,在军用和民用领域该技术都受到了广泛关注。本文以研制红外与可见光的图像融合系统为核心,对红外与可见光成像原理,图像融合算法,融合系统设计以及融合系统应用展开了相关研究。首先,研究了融合系统前端探测器成像理论,针对原有的320×240非制冷红外探测器分辨率较低的问题,重新以UL03041型非制冷探测器为成像原件,设计了具有384×288非制冷探测器驱动电路。在FPGA中设计了具有自动增益的红外增强算法,解决了红外图像非均匀校正后对比度低的问题。在此基础上完成系统软硬件调试,为之后的融合系统提供良好的红外成像平台。其次,针对常见的图像融合算法,融合后信息量过大,影响人眼对目标或者背景观察的缺点,提出了一种新的红外与可见光图像融合算法。算法的核心思想是利用人眼视野的特征,以高分辨率的可见光图像信息为背景,以红外图像的热敏感信息为目标,用改进的形态学滤波对红外图像进行预处理,除去红外图像中的冗余信息,将剩余的红外目标轮廓信息与可见光图像进行融合,并且利用人眼对红色与绿色目标较为敏感的特性进行彩色融合。实验表明,本文算法有效可行,相对于常见的小波以及金字塔算法具有较少的运行时间,在增强目标的同时兼顾图像背景等细节信息。然后,研制了具有红外、微光与可见光CCD的多源图像融合系统。采用平行光轴的双目成像原理设计了系统成像前端,计算了不同视场图像的成像视差和仿射变换需要的参数,采用双线性内插算法缩减了不同图像的差异,通过硬件查找表实现快速图像实时配准。系统通过光纤连接成像前端与后端操控显示部分,实现远距离图像融合控制。系统设计两种工作模式：白天红外与可见光CCD图像融合；夜晚红外与微光图像融合。并利用实验验证了系统的实用性。最后,针对融合系统在实际中的应用,以典型目标飞行弹丸为例,研究了可见光CCD与红外对弹丸的探测距离,针对目标在飞行过程中的表面温度变化,建立融合系统对目标的探测距离计算模型。结合弹丸的弹道学理论,分别给出融合系统在不同应用需求时,探测飞行弹丸的最佳理论位置。更多还原

【Abstract】 Infrared and visible image fusion technology can extract image feature from different wavelengths, and enhance our ability of perceiving image information in the field of detection. Therefore, it is drawing great attention among civil and military fields. This paper is based on the development of infrared and visible image fusion system, and a comprehensive study of infrared and visible light imaging theory, image fusion algorithm, the design and application of fusion system is conducted.Firstly, the imaging theory of front-end detector of fusion system is studied. In view of low resolution (320×240) of existing uncooled infrared detector, new driver circuit for uncooled infrared detector with the resolution of384×288is redesigned based on UL03041. For the problem of image with low contrast after correcting infrared detector’s nonuniformity, the automatic gain algorithm is designed based on FPGA. After debugging software and hardware system, it provides a better infrared imaging platform for the following fusion system.Secondly, according to traditional image fusion algorithm, which has a great amount of information after fusion and influences our observation on target or background, a kind of imaging fusion algorithm on infrared and visible images is brought about. The algorithm’s basic idea is based on the character of human eye’s field. It takes visible image information with high resolution as background and thermal sensitivity information of infrared image as target, and then preprocesses infrared image with the improved morphological filtering to remove redundant information. Next, it fuses the contour information of infrared target and visible image. Finally, it finishes color fusion based on the human eye’s sensitivity to red and green target. The result of experiments shows that the algorithm put forward in this thesis is effective and feasible. It pursues both the detailed background and enhanced target information in fusion image, and takes less running time than wavelet and pyramidal algorithm.Then, multisource image fusion system which concludes infrared, low light level(LLL) and visible light CCD is developed. It uses binocular imaging theory of parallel optical axis to design the front-end of imaging system, to calculate imaging parallax of different view fields and parameters needed in affine transformations. It narrows down differences of different images by using bilinear interpolation algorithm and realizes rapid and real-time image registration by looking up table in hardware. The front-end imaging part and back-end control display part of fusion system are connected by fiber so as to implement remote image fusion control. Two working modes are designed:in the daytime, infrared and visible light CCD images are fused, while at night, infrared and LLL images fused. Furthermore, it checks the utility of this system by experiments.Finally, the fusion system is applied in practice. On the perspective of typical target like a flying projectile, detection ranges by visible light CCD and infrared are studied. Then, based on the temperature change on target’s surface during the flying process, computing model of the detection range by fusion system is established. With the theory of ballistics, the optimum region to detect the flying projectile in different requirements of applications is put forward.更多还原