移动机器人运动目标检测与跟踪系统研究

河北工业大学

硕士学位论文

移动机器人运动目标检测与跟踪系统研究

姓名：冯乐涵

申请学位级别：硕士

专业：机械工程

指导教师：张明路

2010-12

河北工业大学硕士学位论文

移动机器人运动目标检测与跟踪系统研究

摘 要

在人类所感知的环境信息中，视觉信息占了非常大的比重，而动态视觉信息更是其主要组成部分。计算机视觉是用计算机模拟人眼功能，从图像或者图像序列中提取信息，对外界环境中的事物进行形态和运动识别。在各种光学图像中，彩色图像是对外部客观世界最为逼真的描述。相比灰度图像来说，对彩色图像的处理可以获得更大的信息量。

本文针对彩色运动目标检测与跟踪技术开展研究工作，解决了色彩分割、形状检测、目标跟踪等技术问题，搭建了基于移动机器人的彩色目标识别与跟踪系统，并完成了相应的实验。论文的主要工作及研究成果如下：

首先，分析了移动机器人运动目标检测与跟踪系统的功能及性能指标，设计了系统的总体组成结构并构思了各子系统的实现形式。

其次，研究了运动目标检测和跟踪算法。一方面，设计了基于颜色特征和形状特征的目标物分割算法，实现了对图像中彩色目标物的有效分割。另一方面，对经典Camshift跟踪算法进行了改进：实现了该算法的全自动化；将Kalman滤波器融入其中，提高了算法在目标物遮挡情况下的鲁棒性；通过参数设置，使该算法在出现跟踪失败的情况时能够及时重新初始化，进入新一轮的目标检测及跟踪。本文通过OpenCV实现了运动目标检测及跟踪算法，并通过实验验证了该算法的可行性。

最后，建立了移动机器人运动目标检测与跟踪系统。该系统将上位机图像处理的结果以无线方式传送给下位机，再由下位机控制移动机器人实现对运动目标的跟踪。该系统还装有超声测距装置，可以实时感知机器人与运动目标之间的距离，从而有效避免了机器人与目标物之间的碰撞。

关键词：运动目标检测，图像分割，运动目标跟踪，Camshift算法，移动机器人，OpenCV

i

移动机器人运动目标检测与跟踪系统研究

RESEARCH OF MOVING TARGET DETECTION AND

TRACKING SYSTEM BASED ON MOBILE ROBOT

ABSTRACT

The visual information, especially the dynamic visual information, occupies most of the composition among the environment information perceived by human beings. Computer vision is the use of computer to simulate human eyes system. It could get information from image sequence, and recognize the configuration and feature of the goal object. Because color image can describe objects better than the gray-level image, more information can be obtained from processing color image.

The paper focuses on the research of moving target detection and tracking system, solving the technical problems such as the color segmentation, form detection and target tracking. It finally achieves the color target recognition and tracking system based on the mobile robots and completes the corresponding experiments. The major work and results of the paper are as the following:

Firstly, the performance and function indexes of the mobile robot moving target detection and tracking system are analyzed. And the general structure composition of the system and the realization forms of all subsystems are designed.

Secondly, moving target detection and tracking algorithm are investigated. The paper designs the target segmentation algorithm based on the characteristics of color and shape, which successfully divides the colored targets in the images. The classic Camshift Algorithm is improved: achieved the complete automation of Camshift. The Kalman filter is added into the Camshift, which increases the robustness of the algorithm when the target is sheltered. By setting the parameters, the algorithm can reinitialize and begin a new round of detection and tracking immediately when its track failed. The feasibility of the algorithm of the target detection and tracking designed in this paper is proofed by the experiments.

Finally, the moving target detection and tracking system based on mobile robots is established. The system transmits the image results dealt by the upper computer to the lower computer on wireless way. The mobile robot controlled by the lower computer tracks the object. This system has ultrasonic distance measurement system, which could perceive the distance between the robot and the target, avoiding the impact of the robot and target.

KEY WORDS: moving target detection, image segmentation, moving object tracking, Camshift algorithm, mobile robot, OpenCV

ii

河北工业大学硕士学位论文

第一章 绪论

§1-1 研究背景和意义

视觉是人类信息交流的一种重要方式，它不仅指对光信号的感觉，还包括了获取、传输、处理、存储和理解视觉信息的全过程。网络及多媒体技术的飞速发展进一步扩大了视频图像的应用范围，使之逐渐成为信息传播的主流载体之一。由于视觉在人类感知中的地位非常重要，视频图像技术一直都是科学研究的热点所在[1]。

机器视觉可以代替人类的视觉从事目标检验、目标跟踪、机器人导航等方面的工作，在一些需要重复、迅速的从图象中获取精确信息的场合下，机器视觉与人类视觉相比，具有更加敏锐的洞察力和持续工作的能力。尽管在目前软硬件技术条件下，机器视觉功能还处于初级水平，但其潜在的巨大应用价值已经引起了世界各国的高度重视，发 …… 此处隐藏：15623字，全部文档内容请下载后查看。喜欢就下载吧 ……