过采样下红外弱小目标检测算法研究

发布时间：2018-06-18 19:33

本文选题：过采样 + 目标检测　；参考：《国防科学技术大学》2015年硕士论文

【摘要】：红外弱小目标检测是目标检测领域的一项重难点问题。由于弱小目标信噪比比较低,淹没在背景和噪声中,检测难度大。本文研究了过采样体制下的目标检测算法。首先建立过采样扫描红外图像模型,并分析了红外弱小目标特性。其次介绍了几种典型的背景抑制算法,并对空域高通滤波,均值滤波,最大中值滤波进行详细的介绍。再次,用矩形结构元素及线性多结构元素形态学抑制背景,通过分析:线性多结构元素可以抑制云层背景边缘,但无法消除小于目标的亮点噪声,而矩形结构元素对云层背景边缘抑制效果较差,但是可以消除小于目标的亮点噪声。在此基础上,利用过采样图像中目标的特点,结合线性多结构元素和矩形结构元素的优点,提出了基于过采样体制的形态学多级滤波背景抑制算法,并对结构元素进行最优化构造。同时对该算法的结构及原理进行了具体地阐述,并通过仿真对算法进行分析。仿真实验得出该算法在提高过采样扫描图像的信噪比方面效果是比较好的,而且候选目标点也比较少,算法的检测能力明显较高,在多帧检测中,对传统的邻域判决法进行改进,提出基于联合状态估计的过采样目标序列检测算法,采用改进的算法对单帧检测阶段检测出的目标进行轨迹检测,改进的算法不仅检测出了目标轨迹,而且虚警率也较低,同时验证了单帧检测中提出的基于过采样体制的形态学多级滤波背景抑制算法是可行的。
[Abstract]:Infrared small and weak target detection is an important and difficult problem in the field of target detection. Because the signal-to-noise ratio of small target is low, it is difficult to detect because it is submerged in background and noise. In this paper, the target detection algorithm under oversampling system is studied. Firstly, the oversampling scanning infrared image model is established, and the characteristics of infrared dim targets are analyzed. Secondly, several typical background suppression algorithms are introduced, and the spatial high pass filter, mean value filter and maximum median filter are introduced in detail. Thirdly, the morphology of rectangular structure element and linear multi-structure element is used to suppress the background. Through analysis, the linear multi-structure element can suppress the edge of cloud background, but it can not eliminate the bright spot noise which is smaller than the target. The rectangular structure element has a poor suppression effect on the cloud background edge, but it can eliminate the bright spot noise which is smaller than the target. On this basis, taking advantage of the characteristics of targets in over-sampled images and combining the advantages of linear multi-structure elements and rectangular structural elements, a multi-level morphological filtering background suppression algorithm based on over-sampling scheme is proposed. The structure elements are constructed optimally. At the same time, the structure and principle of the algorithm are described in detail, and the algorithm is analyzed by simulation. The simulation results show that the algorithm is effective in improving the signal-to-noise ratio of the over-sampled scanned images, and the candidate target points are less, the detection ability of the algorithm is obviously higher, in the multi-frame detection, The traditional neighborhood decision method is improved, and an oversampling target sequence detection algorithm based on joint state estimation is proposed. The improved algorithm is used to detect the track of the target detected in single frame detection stage. The improved algorithm not only detects the trajectory of the target, but also has a low false alarm rate. At the same time, it verifies the feasibility of the multi-level morphological filtering background suppression algorithm based on over-sampling in single-frame detection.
【学位授予单位】：国防科学技术大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TP391.41;TN219

【参考文献】