复杂背景下的红外弱小目标检测算法

发布时间：2018-01-30 06:05

本文关键词： 红外成像弱小目标检测各向异性扩散滤波奇异值分解多尺度几何分析　出处：《西安电子科技大学》2015年硕士论文　论文类型：学位论文

【摘要】：红外探测技术利用目标和背景之间的红外辐射差异来进行目标探测,具有全天时工作、隐蔽性强和抗干扰性好等优点,因而得到了广泛的应用。然而,远距离下目标成像面积非常小,同时成像目标的辐射强度也相对较弱,特别是当目标处于复杂的背景环境中,目标甚至可能被复杂的背景所淹没,信杂比更低,使得红外弱小目标检测的难度极大地提高。因此,对复杂背景下的红外弱小目标检测技术进行深入研究,有着极为重要的理论意义和实际的工程应用价值。本文首先对红外图像中的目标和背景的辐射特性进行了分析,并运用多尺度几何分析的方法研究了目标与背景在不同尺度和不同方向的表现形式,为后面提出新的目标检测算法提供理论支持。其次,介绍了偏微分方程理论在图像与信号处理中的应用,由于其中的各向异性扩散方程具有显著的各向异性,可以将其进行离散化处理,运用到滤波中,本文对扩散系数进行修正,并通过与原始图像作差得到一种改进的各向异性扩散差分滤波。然后,将非下采样Contourlet变换与奇异值分解引入到红外弱小目标检测中,提出了一种基于非下采样Contourlet变换与奇异值分解的红外弱小目标检测算法。该算法首先将原始图像进行奇异值分解,选取适当数目的奇异值重构背景信息;再通过与原始图像的差分运算来抑制背景;然后进行非下采样Contourlet变换,在变换域中再次利用奇异值分解来保存目标信息、抑制背景和滤除噪声;最终经过非下采样Contourlet反变换即可实现复杂背景下的红外弱小目标检测。采用真实的红外图像进行了仿真实验,结果表明了该算法的有效性。最后,研究了Surfacelet变换,提出了一种基于Surfacelet变换与各向异性扩散方程的红外弱小目标检测算法。该算法首先利用Surfacelet变换对原始图像进行分解,得到一系列的高频方向子带和低频子带;然后分别采用各向异性扩散差分滤波和局部去均值滤波对高频方向子带和低频子带进行处理,以此来突显目标和抑制背景,最终实现复杂背景下的红外弱小目标检测。采用真实的红外图像进行了仿真实验,实验结果表明该算法可以有效地实现红外弱小目标的检测,同时具有非常不错的鲁棒性。
[Abstract]:The infrared detection technology uses the infrared radiation difference between the target and the background to carry on the target detection, has the advantage of all-day operation, strong concealment and good anti-jamming, so it has been widely used. The imaging area of the target is very small at a long distance, and the radiation intensity of the imaging target is relatively weak, especially when the target is in the complex background environment, the target may even be submerged by the complex background, and the signal-to-clutter ratio is lower. The difficulty of infrared dim target detection is greatly improved. Therefore, the infrared dim target detection technology under complex background is studied in depth. It has very important theoretical significance and practical engineering application value. Firstly, the radiation characteristics of target and background in infrared image are analyzed in this paper. And the multi-scale geometric analysis method is used to study the representation of target and background in different scales and different directions, which provides theoretical support for the proposed new target detection algorithm. Secondly. The application of partial differential equation theory in image and signal processing is introduced. Because the anisotropic diffusion equation has obvious anisotropy, it can be discretized and applied to filtering. In this paper, the diffusion coefficient is modified, and an improved anisotropic diffusion differential filter is obtained by differentiating with the original image. Non-downsampling Contourlet transform and singular value decomposition (SVD) are introduced into infrared dim target detection. This paper presents an infrared small and weak target detection algorithm based on non-downsampling Contourlet transform and singular value decomposition, which firstly decomposes the original image by singular value decomposition. Select a proper number of singular values to reconstruct the background information; Then the background is suppressed by the difference operation with the original image. Then the non-downsampling Contourlet transform is carried out, and the singular value decomposition is used again in the transform domain to save the target information, suppress the background and filter the noise. Finally, the infrared dim target detection under complex background can be realized by non-downsampling Contourlet inverse transform. The real infrared image is used for simulation experiment. The results show that the algorithm is effective. Finally, the Surfacelet transform is studied. An infrared small and weak target detection algorithm based on Surfacelet transform and anisotropic diffusion equation is proposed. Firstly, the Surfacelet transform is used to decompose the original image. A series of high frequency subbands and low frequency subbands are obtained. Then the anisotropic diffusion difference filter and the local de-mean filter are used to process the high-frequency directional subband and the low-frequency subband respectively to highlight the target and suppress the background. Finally, the infrared small and weak target detection under the complex background is realized. The real infrared image is used to carry out the simulation experiment, and the experimental results show that the algorithm can effectively realize the infrared small and weak target detection. At the same time, it has very good robustness.
【学位授予单位】：西安电子科技大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TP391.41;TN215

【参考文献】