【作者】 杨德贵；

【导师】 肖顺平； 黎湘；

【作者基本信息】 国防科学技术大学 ， 信息与通信工程， 2010， 博士

【摘要】 随着红外传感器技术的发展,凝视红外探测系统在军事领域中的应用越来越广。相对于空中和海面而言,凝视红外探测系统在地面应用难度更大,其主要难点有:(1)目标及地物背景种类繁多;(2)目标易被复杂背景淹没或遮挡;(3)探测平台相对运动;(4)成像条件影响检测性能;(5)被动系统难以测距等。针对上述难点,本文开展了地物类型反演、地面复杂背景下红外目标检测以及被动测距究等工作,主要研究成果如下:1.构建了单层热能交换模型,仿真分析了典型地物的辐射统计特性,利用外场实测数据验证了模型的准确性;基于单层热能交换模型建立了相应地物表面的热平衡函数,利用相对稳定的地物光电特性参数以假设检验方式进行类型反演。2.对块匹配方法进行改进,实现了地面运动背景的快速补偿;针对单波段红外探测系统,基于图像理解的思想提出基于感兴趣区域的地面弱小运动目标检测算法;针对双波段红外探测系统,提出基于差分图像融合的地面弱小目标检测算法,结合了差分算法的时域优势和多传感器处理的空域优势,两种算法均有效提高了弱小运动目标的检测性能。3.提出一种基于改进的圆投影向量的图像匹配算法,不仅实现了运动背景的快速补偿,还可以构建面目标描述模型。在此基础上,针对运动面目标易出现的遮挡、目标机动等问题,提出一种基于圆投影向量和粒子滤波的地面运动面目标检测与跟踪算法。另外,针对地面运动目标非合作的情况提出一种基于综合信息度量的的地面运动目标检测和跟踪算法,不仅保证了算法有效性,还对地面运动小目标和面目标均具有通用性。4.分析了探测系统成像条件对图像像素灰度的影响,利用红外图像的灰度映射变换得到能够真实反映背景辐射量的表观温度图像;基于表观温度图像在像素邻域滑窗内提取相应特征集,采用假设检验方法检测静止目标,并针对不同成像条件提出了自适应滑窗的选择方法。5.构建了凝视红外成像系统的目标作用距离模型,证明了面/小目标与背景的辐射通量差具有相同表达形式,利用红外图像像素灰度与目标辐射通量之间的关系,推导了单站单波段与双波段条件下的被动测距方法。最后对凝视红外探测系统的组成、功能、研制及其外场试验等情况进行了总结。更多还原

【Abstract】 With the development of infrared sensor technology, staring infrared detection systems are being applicated more and more widely in the military field. Compared with the air and sea applications, it is more difficult in terrestrial applications. And the main problems are shown as follows: (1) the types of the target and the typical ground are complicated, (2) the target is easy to be blocked or submerged, (3) the system platform is relative locomotory commonly, (4) the imaging condition affects the detection performance, (5) the passive ranging is difficult, etc.According to the abovementioned difficulties, the inversion of the typical ground, the target detection under complicated ground condition and the passive ranging are investigated in this thesis. And the main contributions are shown as follows:1. The uniform single-level heat-exchange model is put forward, and the statistics characteristic for the typical ground infrared radiance is emulated, whose veracity is verified by the outfield infrared images. Then the thermal balance function is constructed, and the ground type is inversed by the hypothesis testing method based on the typical ground optic-electric parameters.2. The block matching method is improved to compensate the backgrouond motion quickly. Then, for the single-band IR detection system, the detection algorithm based on region of interest was put forward. Finally, for dual-band infrared detection system, a detection algorithm for ground weak small targets, which integrates the advantages in temporal-spatial domain, is propsed based on fusion of inter-frame difference images. Both of the algorithms can improve the detection capability.3. Aimed at the moving surface target, firstly, the round projection vector to compensate the background motion is improved. Then associated with the particle filter, the method combineing detection and tracking the ground moving surface target is proposed. In addition, according to uncertain size of the non-cooperative target, an algorithm is proposed based on the measurement of integrated information, which can not only guarante the detection effectiveness, but also be universal to both small and surface targets.4. The impact of imaging conditions on the gray scale of pixels is analyzed. In order to get the real apparent temperature of the background, the IR images are transformed based on the imaging condition. Then the feasure stes can be extracted from the pixel neighboring window, and the target was detected by hypothesis testing with the adaptive window chosen according to different imaging condition.5. The range model is founded based on the working principle of the staring IR imaging system. Considered the relationship between the pixel gray and the target’s radiating power, the two ranging algorithms are derived based on one band and two band of the single station. Then the algorithms are validated by the outfield IR image, and the ranging error is analyzed.Finally, the composition, function, development and field experiment of the staring infrared detection system is summarized.更多还原