摘要随着我国人口老年化和少子化的发展,社会对服务机器人的需求不断增长,以助老助残为目标的服务机器人研究得到了广泛重视。为了满足全向移动式服务机器人在家庭环境下对导向技术的需求,本文开展基于机器视觉的目标识别与目标定位技术研究。 本文在分析当前基于 PC 机平台的视觉导向系统,在家庭环境下存在的不足的基础上, 提出以 DSP为核心的嵌入式双目视觉目标识别与定位方案, 建立了以 TMS320DM642芯片为图像处理器的嵌入式硬件平台,分析了该平台在视觉导向系统中的功能结构,详细介绍了 TMS320DM642 处理器的内核资源和外部接口,以及双目摄像头、视频输入模块、视频输出模块等核心部件的功能。在此基础上,设计了双目视觉目标识别与导向系统的软件功能及其处理流程,并介绍了 CCS 软件开发和调试平台的特点及应用方法。论述了与目标识别相关的图像预处理、特征提取理论,提出了目标识别算法及其流程。重点研究了双目视觉定位理论、立体匹配原理及双目测距方法,并设计了双目视觉定位算法及其流程。最后,通过双目摄像头的标定实验,获得了相机的内外参数,实验验证了所设计的目标识别、定位及测距算法在本文搭建的嵌入式软硬件平台上是可行的。 27342 毕业论文关键词 图像处理 机器视觉 目标识别 目标定位 双目视觉 Title Home Omnidirectional mobile service robot—— Object recognition and visual guidance system design Abstract With the development of aging population in our country, the social demand for service robots is growing, which leads the service robot research to extensity attention. In order to meet the requirements of the mobile robot in the home environment, this paper studies the target recognition and target localization technology based on computer vision. This article follows the current visual guide system based on PC platform, considering the shortcut of family environment, using DSP as the core of embedded binocular visual object recognition and localization scheme is established based on TMS320DM642 chip for image processor embedded hardware platform. Analyses the vision guidance system in the functional structure of the platform, processor TMS320DM642 inside resources and external interface, binocular camera, video modules are the core component. Based on this, the software function and the process of the binocular vision target recognition and guidance system are designed, and the features and application of the CCS software development and debug platform are introduced. The theory of image preprocessing and feature extraction is discussed, and the target recognition algorithm is proposed. The binocular vision positioning theory, stereo matching principle and binocular ranging method are studied, the binocular vision positioning algorithm is designed. Finally, the experiments show that the target recognition, positioning and location algorithm in this paper build of embedded software and hardware platform is feasible. Keywords Image processing Computer vision Object recognition Target location Binocular vision
目次
1 绪论 1
1.1 课题的研究背景及意义 . 1
1.2 家用服务机器人视觉系统研究现状 . 1
1.3 论文总体思路与内容安排 . 2
2 目标识别与视觉导向系统设计 4
2.1 系统分析及总体设计 . 4
2.2 硬件系统 . 6
2.3 软件系统 . 10
3 目标识别与定位理论及算法研究 12
3.1 目标识别理论 . 12
3.2 目标识别算法设计 . 15
3.3 目标定位理论 . 16
3.4 立体匹配理论 . 22