Kalman滤波在CME图像处理中的应用

发布时间：2018-04-25 07:15

本文选题：日冕物质抛射 + 背景建模　；参考：《昆明理工大学》2017年硕士论文

【摘要】：日冕物质抛射(CME)是太阳爆发活动中规模最大、对地球环境影响最为严重的活动现象。这种现象可能引起通讯中断、宇航失灵、卫星故障、以及电网输送网络和电力设备的崩溃。因此研究CME的提取与探测对于空间天气预报有着十分重要的意义。CME的提取和探测过程实际上就是对观测的序列图像,运用数学建模的方法和图像处理技术探测CME的过程。CME的探测可以看作是动态背景下的运动目标检测,其中背景建模技术对CME的探测有着至关重要的作用。针对CME的序列图像中背景环境复杂的特点(行星和彗星、冕流的变化),本论文在对比和分析光流法、传统帧间差分法和背景差分法的特点和不足后,改进了传统的Kalman滤波背景更新算法。以极坐标下的序列图像为背景建模对象,通过对传统Kalman滤波在模型初始化、模型学习率做出改进,建立动态变化的图像背景,并运用背景差分法和形态学技术对差分图像进行后处理来检测和识别CME,并对CME进行跟踪。相对于光流法、传统帧间差分法和背景差分法,本论文提出的方法检测到的C ME轮廓比较完整、准确率史高、误差小,史能适应太阳序列图像中复杂的背景环境,针对变化的动态环境能更好地去除噪声、冕流等干扰,建立更加趋近真实的背景。本论文的主要研究内容和结果如下:(1)首先,讲述CME研究的背景和意义,概述运动目标检测的研究方法、CME的国内外研究方法的现状,以及概述运动目标检测的研究难点。(2)对运动目标检测中当前主流的帧间差分法、光流法、背景差分法和传统的Kalman算法进行简单的介绍,并对算法的优势和不足进行深入地对比和分析。(3)针对当前运动目标检测算法存在的主要问题,改进了传统的Kalman滤波背景更新算法,并详细地介绍了该算法的改进过程。改进的内容主要包括背景模型的初始化参数估计、背景模型的更新以及后处理中CME运动区域的提取与处理等。并将改进的算法应用到CME的提取处理中,主要步骤包括图像的预处理、改进Kalman背景的提取、运动物质的确定,运动目标的连通性分析、孔洞填充,阈值分割等。(4)将本文提出的改进Kalman背景差分法与传统的算法进行对比和分析,以手工检测CME的CDAW为基准,分别与SEEDS、CACTus、帧间差分法、传统Kalman滤波背景更新法进行对比和分析。实验表明,它不仅能有效的探测LASCO C2序列图像中的CME,还能够探测到CDAW手工目录上探测不到的C3序列图像上的CME。与手工识别相比,本论文中的探测方法更为快速,鲁棒性更好。在精度方面,它不仅可以探测到CDAW手动目录上列出的CME,还可以检测到亮度比较弱的CME,在一定程度上提高了自动检测算法的准确率。同时,对检测到的CME的跟踪效果也良好。本论文方法的不足之处是仍旧存在较小的误检率。
[Abstract]:Coronal mass ejection (CMEs) is the largest active phenomenon in the solar burst and has the most serious impact on the earth's environment. This could cause communications disruptions, aerospace failures, satellite failures, and the breakdown of power networks and power equipment. Therefore, it is very important to study the extraction and detection of CME for space weather forecast. In fact, the process of extracting and detecting CME is the sequence image of observation. The detection of CME by mathematical modeling and image processing can be regarded as moving target detection in dynamic background. Background modeling plays an important role in the detection of CME. In view of the complex background environment of CME sequence images (planets and comets, coronal current changes), this paper compares and analyzes the characteristics and shortcomings of optical flow method, traditional inter-frame difference method and background difference method. The traditional background updating algorithm of Kalman filter is improved. Taking the sequence images in polar coordinates as the background modeling object, by initializing the traditional Kalman filter in the model, the learning rate of the model is improved, and the dynamic image background is established. The background differential method and morphological technique are used to detect and recognize the differential image and track the CME. Compared with the optical flow method, the traditional frame difference method and the background difference method, the CME profile detected by this method is relatively complete, the accuracy is high, the error is small, and the history can adapt to the complex background environment in the sun sequence image. The noise and coronal flow can be removed better in the dynamic environment, and a more realistic background can be established. The main contents and results of this thesis are as follows: (1) first of all, the background and significance of CME research are described, and the research methods of moving target detection are summarized. And the research difficulty of moving target detection is briefly introduced, such as inter-frame differential method, optical flow method, background differential method and traditional Kalman algorithm. The advantages and disadvantages of the algorithm are compared and analyzed deeply. Aiming at the main problems of the current moving target detection algorithm, the traditional background updating algorithm of Kalman filter is improved, and the improvement process of the algorithm is introduced in detail. The improvements include the initialization parameter estimation of the background model, the updating of the background model, and the extraction and processing of the CME motion region in the post-processing. The improved algorithm is applied to the extraction of CME. The main steps include image preprocessing, the extraction of improved Kalman background, the determination of moving material, the connectivity analysis of moving object, the filling of holes. Threshold segmentation, etc.) compare and analyze the improved Kalman background difference method and the traditional algorithm, and compare and analyze the traditional Kalman filtering background updating method with the SEEDSCACTus, inter-frame difference method and traditional Kalman filter background updating method, respectively, based on the manual detection of CME CDAW as the benchmark. Experiments show that it can detect not only the LASCO C2 sequence image, but also the C3 sequence image which can not be detected on the CDAW manual directory. Compared with manual recognition, the detection method in this paper is faster and more robust. In terms of accuracy, it can detect not only the CDAW listed on the manual directory, but also the CMEs with weak brightness, which improves the accuracy of the automatic detection algorithm to a certain extent. At the same time, the tracking effect of the detected CME is also good. The shortcoming of this method is that there is still a small false detection rate.
【学位授予单位】：昆明理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：P35;TP391.41

【相似文献】