【作者】 庞晓宇；

【导师】 李淑娟； 郑勐；

【作者基本信息】 西安理工大学 ， 机械制造及其自动化， 2008， 硕士

【摘要】 移动机器人的视觉导航技术是目前科学研究的热点和重点,随着视觉导航技术研究的深入,工业AGV搬运机器人、智能车辆和国防技术研究取得了重大的突破。由于视觉系统所采集的图像具有信息丰富,对运动环境描述全面的特点,使得移动机器人控制智能化成为发展趋势。本文根据移动机器人的运动控制要求,对图像预处理、图像分割、特征区域匹配的方法进行了研究,并完成了对目标标识物的识别。运动模型的分析为移动机器人的运动控制提供了理论依据,在对CCD摄像机模型在图像采集中所产生的畸变进行分析过程中,采用Tsai两步法标定出摄像机的内外参数,并通过实验对标定结果进行验证。结果表明,该标定方法可以达到CCD摄像机的标定要求。由于视觉系统采集的图像受到周围环境的影响,会产生噪声和畸变。因此在进行图像分割和特征区域匹配前对所采集的图像进行了灰度变换、中值滤波等预处理,提高了图像处理的效率;在图像分割过程中汇总对比了多种边缘检测算子的处理效果和效率,最终选定采用Sobel算子进行图像的边缘检测。在标识物的识别过程中,进行基于颜色分割时,选取自动阈值分割法,采用基于最大面积法选取初始阈值的最大类间、类内方差比法对运动路径图像进行初始分割。在图像修正中又先后使用了图像膨胀和图像细化的方法,便于目标标识物边缘信息的更好提取。在标识物识别和运动控制系统中,提出了颜色特征区域匹配的方法,通过图像中像素RGB值与样板中的RGB值的绝对偏差的计算,实现了对标识物的识别。利用Visual C++的开发环境和图像处理技术,开发了移动机器人运动检测和视觉导航两个系统,实现了对移动机器人的运动仿真和运动控制。移动机器人视觉导航实验结果表明,论文中所提的方法和思路有效可行。更多还原

【Abstract】 Vision-based navigation technology is a new hot spot in the area of mobile robot research. Researching on the AGV technology, Intelligent Vehicle and National defense technology have acquired breakthrough with the Vision-based navigation technology be in-depth studying. Because the images collected by the vision system can describe roundly the environment, and carry a tremendous amount of information about the scene, it becomes a Development Trend for intellectual control of the mobile robot. According to the controlling requirements on the movement of the mobile robot, this article focus on the recognition of target objects by the methods of Image preprocessing、Image Segmentation、matching for feature area.The analysis of movement model provides theoretical proof for the movement control of the mobile robot. On the consideration of distortion of CCD model in images collection, the Tsai two-step method is accepted to calibrate the internal and external parameters, and test the validity of above prediction with experiment. The result indicates that the calibration could meet for the calibration of the CCD Video Camera.Because of the environmental influences, the collected images have noise and distortion, which must be preprocessed with the gray level transformation、histogram handling and airspace filtering before the image segmentation and matching for feature area to increase the efficiency of image processing. Then by the comparison of the outcome and efficiency of comparison the image segmentation process, the Sobel operator is chosen to do the edge detection on images.In the target recognition, when processing the image segmentation on colors, the auto-threshold method is adopted and the maximum area method is practiced in the selection of the initial threshold, and the max ratio of class among variance and class internal variance is applied to carry through the initial segmentation. In the image modification the methods of image expansion Image Expansion and Image Thinning is used in order to ease the obtain of the edge information of the target. The method of matching for feature area by color techniques is proposed and practiced in the movement testing and control system on Mobile Robot. The system could achieve the recognition of target, though computing the absolute deviation of RGB from Pixel-point in the image to target template.With the utilization of the Visual C++ and image processing technology two systems, movement testing on mobile robot and vision-based navigation are established and Motion Simulation and control are realized. The experiment on the vision-based navigation of mobile robot indicates the prediction and methods proposed in the article is feasible.更多还原