海面红外序列图像的预处理与目标检测方法研究

发布时间：2018-04-02 10:55

本文选题：红外图像　切入点：周期条纹噪声　出处：《深圳大学》2017年硕士论文

【摘要】：随着海洋资源的大力开发,海洋经济迅速发展,海洋权益争端日渐频繁,海上交通活动日益繁忙,迫切需要对本国领海进行实时感知监控。红外热成像技术以其作用距离远、隐蔽性强、穿透能力强、抗干扰性好、目标识别能力强、全天候工作等特点在海上目标检测、识别方面到了应用,成为其中一种重要的海上目标探测手段。本文主要研究了海上红外图像的周期条纹噪声去除方法,海上红外图像序列的海水区域提取方法,以及海上红外图像序列的海面目标检测方法。论文的具体工作和贡献可以概括为如下四个方面:1.针对红外图像中存在周期条纹噪声的去除问题,本文分析了周期条纹噪声的空间特性和频域特性,并推导了周期条纹噪声在幅度谱上的梳状冲激谱特性。在此基础上,提出了一种子图尺寸的最优估计方法,在滤除噪声之前,先将原始图像以最优尺寸裁剪为两个子图,使子图中的梳状谱具有理想冲激的性质,从而更有利于后期的噪声检测和噪声频率分量的滤除。此外,本文还给出了两种滤波器的设计方法,自适应梳状陷波器的设计方法和局部线性插值滤波器的设计方法。实验结果表明,本文提出的方法能够更有效地去除在红外图像中的周期条纹噪声,从而提高了红外图像的质量,更有利于后续的海上目标检测。2.针对海上红外图像序列的海水区域提取问题,本文分别沿水平方向和垂直方向对纯海水背景在空间域和频率域的几种特征进行统计分析,发现海水背景在各方向上的灰度标准差以及各方向上的频率分量幅度均值具有较稳定的特性,具有一定的聚集性。在此基础上,提出了一种基于空-频特征高斯建模的红外图像海水区域提取方法。该算法首先沿不同方向提取特征空间的主成分分量,然后利用Parzen窗函数对特征向量进行概率密度估计,并通过单高斯函数拟合概率密度函数,最后得到各方向的海水背景模型,并通过建立的海水背景模型提取出纯海水背景区域。3.针对海上红外图像序列的海上目标检测问题,本文提出了一种基于海水背景的双向半傅里叶域海面红外目标检测算法,该算法首先根据水平方向和垂直方向的海水背景模型构造双向半傅里叶域海水背景抑制滤波器,并分别对原始图像进行滤波处理,然后对滤波后的结果进行融合增强,最后采用基于Otsu的阈值分割方法进行目标提取。实验结果表明,本文提出的目标检测算法针对本实验室所处理的海上红外视频进行目标检测时,具有误检区域少,目标区域更完整,对非均匀光照噪声有一定的鲁棒性等优点。4.在上述研究工作的基础上,本文设计并实现了两个海上红外序列图像处理软件:红外序列图像预处理软件,海面红外序列图像的目标检测算法验证软件。
[Abstract]:With the rapid development of marine resources and the rapid development of marine economy, marine rights disputes become more and more frequent, and maritime traffic activities become more and more busy. It is urgent to monitor the territorial waters in real time. Infrared thermal imaging technology is far away from its role. Features such as strong concealment, strong penetration, good anti-jamming, strong target recognition ability, all-weather work, etc., have been applied to the detection and recognition of targets at sea. In this paper, we mainly study the method of removing periodic fringe noise from infrared image and the method of extracting sea water region of infrared image sequence. The specific work and contribution of this paper can be summarized as follows: 1. Aiming at the problem of periodic fringe noise removal in infrared image, In this paper, the spatial and frequency-domain characteristics of periodic fringe noise are analyzed, and the comb impulse spectrum characteristics of periodic fringe noise in amplitude spectrum are derived. On this basis, an optimal method for estimating the size of subgraph is proposed. First, the original image is cut into two subgraphs with the optimal size, so that the comb spectrum in the subgraph has the property of ideal impulse, which is more advantageous to the later noise detection and the filtering of the noise frequency component. In addition, In this paper, two kinds of filter design methods, adaptive comb notch filter and local linear interpolation filter, are presented. The experimental results show that, The method proposed in this paper can remove the periodic fringe noise in infrared image more effectively, improve the quality of infrared image, and be more favorable to the subsequent marine target detection .2. aiming at the sea water region extraction problem of the infrared image sequence at sea, the method proposed in this paper can improve the quality of the infrared image and improve the quality of the infrared image. In this paper, several characteristics of pure seawater background in spatial domain and frequency domain are statistically analyzed along the horizontal and vertical directions, respectively. It is found that the standard deviation of sea water background in each direction and the mean value of frequency component amplitude in each direction have relatively stable characteristics and have a certain degree of aggregation. A sea water region extraction method for infrared image based on space-frequency feature Gao Si is proposed. Firstly, the principal components of the feature space are extracted in different directions, and then the probability density of the feature vector is estimated by using the Parzen window function. By fitting the probability density function with single Gao Si function, the sea background model in each direction is obtained, and the pure seawater background region is extracted by the established sea water background model. The sea target detection problem based on the infrared image sequence is presented. In this paper, a bidirectional half-Fourier sea surface infrared target detection algorithm based on sea water background is proposed. Firstly, a bidirectional half-Fourier sea water background suppression filter is constructed according to horizontal and vertical sea background models. The original image is processed by filtering, then the result is fused and enhanced. Finally, the threshold segmentation method based on Otsu is used to extract the target. The experimental results show that, The target detection algorithm proposed in this paper has less error detection area and more complete target area, when the target detection algorithm is used to detect the target in the marine infrared video processed in our laboratory. On the basis of the above research work, we design and implement two marine infrared sequence image processing software: infrared sequence image preprocessing software. Verification software for target detection algorithm of infrared images of sea surface.
【学位授予单位】：深圳大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41

【参考文献】