目标搜索跟踪系统中图像配准算法的研究与硬件化实现

发布时间：2018-05-03 16:50

本文选题：特征检测 + 特征匹配　；参考：《南京理工大学》2017年硕士论文

【摘要】：红外搜索跟踪系统作为传统雷达系统的补充,在现代化战争中有着无可替代的作用。而图像配准作为系统信号处理中的重要一环,也有着非常重要的地位。它是完成可靠探测、跟踪目标等一系列操作的前提和基础。本课题围绕图像配准的算法展开,并通过高速硬件平台实现算法的硬件化。本课题从图像配准的理论知识出发,首先简单介绍了图像配准的不同种类以及它的发展历史,然后结合了红外搜索跟踪系统的成像特点,详细研究了基于特征的图像配准算法和基于互信息的图像配准算法。本文从特征提取和特征匹配两个方面对这两类算法进行了研究,并对相应的算法进行了仿真。实验结果表明基于特征的配准算法在运行速度上要优于基于互信息的配准算法,而基于互信息的配准算法在配准精度上则更为优秀。为了将算法实现硬件化,本课题首先研究了 DSP的相关基础和高速PCB设计的相关知识。接着根据配准算法的要求进行硬件系统的原理设计和器件选型,然后进行布局设计以及布线和优化,最后制作出了一套以TMS320C6678为核心的硬件系统,并且对设计出来的硬件系统进行了全方位的调试,确保了硬件系统能正常运行。最后,将图像配准的算法成功的移植到了本课题设计的硬件系统中。实验结果表明本课题设计的系统能够快速且有效地实现图像配准的功能。
[Abstract]:As a supplement to the traditional radar system, the infrared search and tracking system has an irreplaceable role in the modern war. Image registration, as an important part of the system signal processing, also has a very important position. It is the prerequisite and foundation for a series of exercises to complete reliable detection and target tracking. Based on the theoretical knowledge of image registration, this paper first introduces the different types of image registration and its history, and then combines the imaging characteristics of the infrared search and tracking system, and studies the image registration algorithm based on feature and based on the feature based image registration. The image registration algorithm of mutual information is studied in this paper from two aspects of feature extraction and feature matching, and the corresponding algorithm is simulated. The experimental results show that the feature based registration algorithm is better than the mutual information based registration algorithm in the running speed, and the registration algorithm based on mutual information is in the registration essence. In order to realize the hardware of the algorithm, this topic first studies the related foundation of DSP and the related knowledge of high speed PCB design. Then, according to the requirements of the registration algorithm, the principle design and device selection of the hardware system are made, then the layout design, wiring and optimization are carried out. Finally, a set of TMS320C6678 is made. The core hardware system and the designed hardware system are debugged in all directions to ensure the normal operation of the hardware system. Finally, the image registration algorithm is successfully transplanted into the hardware system designed in this subject. The experimental results show that the system designed by this project can achieve the work of image registration quickly and effectively. Yes.

【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

【参考文献】