面向移动机器人的全方位视觉系统关键技术研究

发布时间：2018-06-04 10:06

本文选题：移动机器人 + 全方位视觉　；参考：《北方民族大学》2017年硕士论文

【摘要】：随着科学技术的不断发展,机器人技术在人类生产生活、军事、救援和反恐等领域发挥着越来越重要的作用。机器人要完成预定的任务就需要获取相关场景信息,由于机器人系统的可扩展性,有很多方式可以为机器人提供决策信息,其中最主要的方式就是机器视觉。传统的机器人视觉系统具有视角单一的缺点,不能一次性获取全局信息。但是,机器人在众多实际应用领域,需要超广角甚至是360°全视角的图像,以便获取更多的视场信息。全方位视觉技术应运而生,由于它具有大视野的显著特征,通过将全方位视觉技术运用于机器人的视觉系统设计中,能够为机器人提供更加全面和高效的工作场景信息,具有良好的应用前景。鱼眼镜头因其物理结构特点,使得它具有比普通镜头更大的视角。在相同面积下,鱼眼镜头捕获的图像信息量更大。因此在相同视场信息要求下,所需鱼眼镜头比普通镜头数量少,相应拼接图像数量减少,拼接次数也随之减少,由此可以显著降低系统对视频图像处理的计算量,提高系统的实时性。本文在分析传统的全方位视觉系统原理的基础上,采用鱼眼镜头获取机器人工作场景信息,再通过改进的畸变校正和图像拼接算法,进而为机器人提供全方位、高效的场景信息。本文重点对以下两个方面的内容开展研究:(1)图像算法研究:对原始畸变鱼眼图像进行有效区域提取并校正,提出了改进的基于经度坐标的鱼眼图像快速校正方法;对于校正后图像,采用SURF算法进行拼接,对拼接的过程设计与实现进行优化,最终得到全方位图像;(2)仿真实验研究:利用专业的移动机器人仿真平台Webots,对移动机器人和实验环境进行仿真建模,并对所提图像算法的效果进行对比研究。将改进的鱼眼图像校正和拼接算法移植于此仿真环境进行实验,验证系统方案的可行性。仿真实验结果表明,本文改进的鱼眼图像校正算法与校正后图像的拼接算法,实时性较高,能够得到比较理想的全方位环视图像,满足了移动机器人对全方位视觉系统的需求。
[Abstract]:With the development of science and technology, robot technology is playing a more and more important role in the fields of human production, military, rescue and anti-terrorism. The robot needs to obtain the relevant scene information in order to complete the predetermined task. Because of the expansibility of the robot system, there are many ways to provide the decision information for the robot, the most important way is machine vision. Traditional robot vision system has the disadvantage of single visual angle, so it can not obtain global information at one time. However, in many practical applications, robots need wide-angle or even 360 掳full-angle images in order to obtain more field of view information. Omnidirectional vision technology emerges as the times require. Because it has the remarkable characteristics of large field of vision, it can provide more comprehensive and efficient working scene information for robot by applying omni-directional vision technology to the design of robot vision system. It has good application prospect. Because of its physical structure, fish-eye lens has a larger angle of view than ordinary lens. In the same area, the fish-eye lens captured more image information. Therefore, under the same field of view information requirement, the number of fish-eye lenses required is less than that of ordinary shots, the corresponding number of stitched images is reduced, and the number of stitching times is also reduced, which can significantly reduce the calculation amount of video image processing in the system. Improve the real-time performance of the system. On the basis of analyzing the principle of traditional omnidirectional vision system, this paper uses the fish-eye lens to obtain the working scene information of the robot, and then through the improved distortion correction and image stitching algorithm, it provides the robot with omni-directional. Efficient scene information. This paper focuses on the following two aspects of the research on the image algorithm: extract and correct the effective region of the original distorted fish-eye image, and propose an improved fast correction method based on longitude coordinates for the fish-eye image; For the corrected image, the SURF algorithm is used for stitching, and the design and implementation of the stitching process are optimized. Finally, we get the omni-directional image simulation experimental research: we use the professional mobile robot simulation platform Webotsto simulate and model the mobile robot and the experimental environment, and compare the effect of the proposed image algorithm. The improved fish-eye image correction and stitching algorithm was transplanted to the simulation environment to verify the feasibility of the system. The simulation results show that the improved fish-eye image correction algorithm and the corrected image mosaic algorithm are more real-time and can obtain ideal omni-directional circle view image, which meets the needs of mobile robot for omni-directional vision system.
【学位授予单位】：北方民族大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41;TP242

【参考文献】