红外热图像拼接技术的研究与应用

发布时间：2018-03-14 15:21

本文选题：红外热像仪　切入点：特征提取　出处：《南京理工大学》2017年硕士论文　论文类型：学位论文

【摘要】：红外图像反映了物体之间的温度差异分布,所以红外监测技术具有远距离不接触和不受恶劣环境影响的特点,在军事目标探测与预警、电力设备在线监测和工业检测等范围都得到了较为广泛的应用。但红外热像仪大多数成像范围窄且分辨率不高,一张图像往往很难显示较大目标物体的全部样貌,给后续的图像分析和被测目标的温度趋势判断带来了不便,而图像拼接技术为其提供了一个较为合适的解决方案。论文首先对红外热像的拼接技术进行了研究,着重研究了红外热像的特征点提取与图像配准,在此基础上提出了相应的改进算法。并设计了与之配套的红外热像采集与拼接系统,通过实验验证了其实用性与可靠性。在图像预处理的研究中,论文首先分析了红外成像原理,对红外图像的去噪和增强技术进行了研究,并在此基础上对红外图像进行仿真实验与结果分析。在图像配准的研究中,论文首先对三大配准方式进行了介绍与仿真说明,研究确定了在论文中使用基于特征的图像配准方案;在特征点的提取中,对经典的角点提取算法和局部不变特征提取算法进行了深入的研究,进行了编程实验仿真与结果分析,并提出了一种改进的多尺度空间的特征点提取方式,通过编程实验验证了其可行性;在图像的特征匹配中,对待匹配图像之间的粗匹配和结合改进的RANSAC算法的特征点的提纯方法进行研究,并用编程实验验证了其可行性。在图像融合的研究中,论文对基本的融合方式进行了介绍,来完成图像拼接的最后一步,并进行编程实验生成两幅图的拼接图。在软件系统的设计中,论文根据之前对算法的研究成果,设计了一套基于红外热像仪的热像采集与拼接系统。该系统不仅能完成热像仪的控制、实时视频的点温度检测和热像的采集,还能够选择相应的算法对热像仪采集的图像进行处理与拼接,给后续图像的分析工作带来了方便,具有一定的价值和意义。
[Abstract]:Infrared images reflect the distribution of temperature differences between objects, so infrared monitoring technology has the characteristics of remote non-contact and unaffected by adverse environment, in the detection and early warning of military targets, Power equipment on-line monitoring and industrial detection have been widely used. However, most infrared thermal imagers have narrow imaging range and low resolution, so it is difficult for an image to display the full appearance of a large target object. It brings inconvenience to the subsequent image analysis and the temperature trend judgment of the target under test, and the image stitching technology provides a more suitable solution for it. Firstly, the paper studies the technology of infrared thermal image stitching. In this paper, the feature point extraction and image registration of infrared thermal image are emphatically studied, and the corresponding improved algorithm is put forward, and a matching infrared thermal image acquisition and splicing system is designed. The practicability and reliability are verified by experiments. In the research of image preprocessing, firstly, the principle of infrared imaging is analyzed, and the denoising and enhancement techniques of infrared image are studied. On this basis, the infrared image simulation experiments and results analysis. In the image registration research, the paper first introduces the three major registration methods and simulation. In the feature point extraction, the classical corner extraction algorithm and the local invariant feature extraction algorithm are deeply studied, and the programming experiments and results analysis are carried out. An improved feature point extraction method in multi-scale space is proposed, and its feasibility is verified by programming experiments. The methods of coarse matching between matching images and feature points of improved RANSAC algorithm are studied, and the feasibility of this method is verified by programming experiments. In the research of image fusion, the basic fusion methods are introduced in this paper. In the software system design, according to the previous research results of the algorithm, A thermal image acquisition and splicing system based on infrared thermal imager is designed. The system can not only control the thermal imager, detect the point temperature of the real-time video, but also collect the thermal image. It can also select the corresponding algorithm to process and join the images collected by the thermal imager, which brings convenience to the analysis of the subsequent images, and has certain value and significance.
【学位授予单位】：南京理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41

【相似文献】