红外视频图像判读的预处理技术研究

【作者】 刘春霞；

【导师】 邢忠宝；

【作者基本信息】 中国科学院研究生院（长春光学精密机械与物理研究所） ， 机械电子工程， 2003， 硕士

【摘要】 数字式视频判读系统利用多媒体技术和数字图像处理技术，分析和处理光电经纬仪的测量数据，对测量结果进行修正，是靶场光测设备测量系统的一个重要组成部分，而判读系统中最重要的部分则是图像的预处理部分，本文主要对红外视频图像预处理技术进行研究。 首先介绍了判读原理和方法，根据图像特点和用户需求，提出了一种新的判读方式：中轴线判读——通过选取目标中轴线上的点作为特征点进行判读，这样判读后的目标位置一定位于目标中轴线上。由于目标中轴线的相对稳定性，使判读数据的准确性得到提高。接着讨论了系统所用的数字图像处理中平滑、阈值分割、边缘检测和形态学处理等算法，并对处理结果进行比较，选择每一种处理方法中最适合系统图像的算法。针对图像中目标周围分布不均匀弱光区的现象，本文在最大类间方差法的基础上，对该算法进行了改进，得到了一种高阈值的二值化算法，在阈值选取中能够得到较大的阈值，可有效去除目标周围的弱光区和目标后面亮度不高的拖尾部分，使目标形状和大小在图像处理后基本保持不变。最后，简要介绍了系统构成和系统流程、主要功能及相关重要问题的解决方法。 文中实验所用图像红外摄像机采集的导弹发射图像和实验室模拟图像，软件在Windows 2000、Visual C++6.0环境下编程实现。更多还原

【Abstract】 Digital video interpretation system is an important part of range optical measuring system. It utilizes multimedia and digital image processing technology, processes measuring data and amends measuring results. In this paper discussed the most crucial part - infrared images pretreatment technology.First, introduced reading principle and reading method. In allusion to the image’s characters and user’s requirement, advanced a new reading method: axis reading. Through selecting the point on the target’s axis as the trait point to read, the target’s position we got must be on the axis of target. So the data was more exact because of the relative stability of target’s axis. Then talk over the correspondent algorithm such as smoothness, threshold division, edge checking, morphology processing and so on. And chose the best appropriate algorithm of each kind of method. In allusion to the, weak light zone unevenly distributed target, improved a new binarization algorithm of high threshold based on the method of maximum classes square error. This algorithm got high threshold in threshold selecting and wipe off the uneven light zone and the tail of low brightness, so the shape and scale of the target were unaltered after image processing. At last introduced the structure of digital frame frequency reading system, and system flow, primary function and the questions met with.The images used in this paper are images token by infrared vidicon and simulate images created in laboratory. The software is programmed in Visual C++6.0 and Windows 2000.更多还原