【作者】 赵鹏；

【导师】 甘育裕；

【作者基本信息】 重庆大学 ， 通信工程， 2007， 硕士

【摘要】 目标分割和跟踪技术是现代科学技术领域的一个重要研究分支,它的相关研究成果可以满足某些军事或民用的应用需求。论文阐述了目前目标分割、跟踪系统的研究现状及相关技术发展动态,提出了基于C6202高速DSP和大规模可编程器件的复杂场景中的目标分割和跟踪的软硬件实现方案。系统设计包括:硬件系统设计、硬件系统的程序设计以及目标分割与跟踪的DSP软件设计。论文讨论的重点是硬件平台的设计,整个硬件系统的设计充分体现了现代电子系统设计的思想。在设计中强调了硬件系统的功能可扩展性和稳定性。硬件系统采用模块化设计方式,这既降低了设计的复杂度,又有利于项目的分工。对硬件系统中的几个关键模块设计,如视频接口的设计、数字图像模块的设计以及系统的总裁与控制逻辑设计进行了比较详细的讨论。硬件系统支持的功能有:视频信号的数字成像、数字图像处理、与主机进行数据交换。论文在图像处理部分主要讨论了预处理模块和边缘检测两个模块。在预处理模块中,利用形态学滤波中的形态开消除图像中的边缘毛刺和孤立斑点,形态闭填补图像中的漏洞以及裂缝;在边缘检测模块中,利用小波变换对图像进行边缘检测。将其应用于DSP系统,从算法程序、硬件特点等方面阐述了这些算法的具体实现过程。论文最后对硬件系统各功能模块接口的调试方法、调试过程进行了讨论。包括:视频接口的调试、输入缓冲FIFO读写操作的调试、图像帧存储器的读写操作的调试以及DSP与PCI9054的HPI接口等的调试。论文所设计的系统应用了小波多尺度边缘检测技术,并引入了形态学滤波算法,解决了图像中的奇异点及边缘细节检测问题,系统具有一定的先进性和较强的实用性。但由于课题涉及的内容较多,诸多问题有待进一步深入研究和探讨,因此系统还存在着许多需要改进和完善的地方。更多还原

【Abstract】 The technology of video object segmentation and tracking is an important branch being studied in modem science and technology field. The relative research project can be served for the need of national defence and civil applications. The thesis describes the state of automatic target recognition ,track system and the developing trends of the relative technology, deeply systemic research the image sementation,recognition and tracking algorithm. The solution of relative software system and hardware system based on the C6202 high-speed DSP and large-scale PLD is developed. The system design can be divided: hardware platform design、hardware system’s program design and DSP software design.The design of hardware platform is mainly discussed in this paper. The method of modern electronic system design is taken into the design of whole hardware system, In the development of hardware system, the function expansibility and stability are emphasized greatly. The modularized structure is used in the design of hardware system .It not only reduces the complexity of design, but also benefits the group work. The function of hardware system includes conberting video signal to digital image deta and digital image processing and data exchange with the host computer.We discuss pretreatment module and edge detection module in the domain of image processing . In pretreatment module, we use morphologic-opening to eliminate edge burrs and island spots.,and morphologic-closing to fill up heaks and splits; In edge detection module, we use wavelet transform to complet edge detection. And they are applied in the DSP system, and the detailed realization steps were expounded in the aspects of programming and hardware characteristic etc.At last, we discuss debugging of hardware system, including debugging of video frequency interface, debugging of reading and writing FIFO, debugging of image frame memorizer,and debugging of interface between DSP and PCI9054.The system has some advanced and practicability because of using of wavelet multi-scale edge detection and morphologic filter arithmetic. While the content too much the task involved, there are many problems have to lucubrate and discuss. So there are many problems need to be improved and consummate.更多还原