红外小目标图像预处理及检测算法研究

发布时间：2018-04-29 11:00

本文选题：图像预处理 + 红外小目标　；参考：《哈尔滨工程大学》2016年硕士论文

【摘要】：红外小目标检测与跟踪作为红外预警、红外制导等领域的核心技术,在现代高科技战争中发挥着越来越重要的作用。受探测器噪声、大气传输的影响,加上探测距离比较远,使得红外小目标图像整体信杂比低,边缘信息缺失,因此红外小目标的检测相对困难。而且在复杂背景条件下,检测难度将进一步加大。本文正是在复杂背景的条件下,对红外图像预处理和小目标检测算法进行了深入的研究。首先,本文通过对比的方法分析了红外图像的特征,并进一步针对红外图像的三个方面进行了详细的分析与研究。然后,在这些分析研究的基础上,构建了红外图像的数学模型。其次,在研究经典的中值滤波算法和其多种改进算法的基础上,本文提出了一种适用于简单背景红外小目标图像的滤波算法——均值中值滤波算法。该算法以图像灰度值均值作为依据来选择阈值,有效区分了背景和目标信号,并分开处理,取得了良好的效果。接着,将该算法结合数学形态学开Top-Hat变换使用,借助开Top-Hat变换的背景抑制能力,成功将该算法引入到复杂背景条件下,显著提高了复杂背景下预处理后红外图像的信杂比。最后,在小目标检测时,着重分析了小波变换对图像突变点的检测算法。并借助方向小波优良的方向解析能力和时频局部化分析能力,在保留目标信号细节信息的同时实现小目标的检测。本文在小目标的判定方面做出了改进,通过标记多个方向小波变换结果中灰度值跳变大的点,并结合这些点在图像中的分布区域来联合判定小目标。该方法不仅有效识别出了小目标,而且对噪声虚警也起到一定的限制作用。
[Abstract]:As the core technology of infrared early warning, infrared guidance and other fields, infrared small target detection and tracking plays an increasingly important role in modern high-tech warfare. Due to the influence of detector noise and atmospheric transmission and the relatively long detection distance, the whole signal-to-clutter ratio of infrared small target image is low, and the edge information is missing, so the detection of infrared small target is relatively difficult. And in the complex background conditions, the difficulty of detection will be further increased. In this paper, infrared image preprocessing and small target detection algorithms are studied in detail under the condition of complex background. Firstly, this paper analyzes the features of infrared images by comparing them, and makes a detailed analysis and research on the three aspects of infrared images. Then, on the basis of these analyses, the mathematical model of infrared image is constructed. Secondly, on the basis of studying the classical median filtering algorithm and its many improved algorithms, this paper presents a filtering algorithm for simple background infrared small target image, which is the mean median filter algorithm. The algorithm selects the threshold value based on the average value of image gray value, and effectively distinguishes the background signal from the target signal, and processes them separately, and achieves good results. Then, the algorithm is used in combination with the open Top-Hat transform of mathematical morphology. With the help of the background suppression ability of the open Top-Hat transform, the algorithm is successfully introduced into the complex background conditions, and the signal-to-clutter ratio of the pre-processed infrared image in the complex background is improved significantly. Finally, in the small target detection, the wavelet transform algorithm for image mutation detection is analyzed. With the help of directional wavelet's excellent ability of direction analysis and time-frequency localization, the detection of small targets can be realized while preserving the detailed information of the target signal. In this paper, the decision of small targets is improved by marking points with large gray values in the results of multi-directional wavelet transform, and combining the distribution of these points in the image to determine small targets jointly. This method not only effectively identifies small targets, but also limits noise false alarm.
【学位授予单位】：哈尔滨工程大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TP391.41

【参考文献】