【作者】 李飞；

【导师】 王华锋；

【作者基本信息】 北方工业大学 ， 电子科学与技术， 2018， 硕士

【摘要】 近年来,随着科学技术的发展,给机器人应用相关行业带来巨大的社会需求。由于机器人能够快速、准确地完成一些具有重复性、危险性的任务。不同于工业机器人,服务类机器人具有比较灵活的应用场景。在技术与性能要求方面,两者的侧重点有所不同:工业机器人侧重于动作的重复执行精度,而服务机器人更侧重于对环境的感知能力和人机交互等方面的能力。由于服务机器人的工作环境具有复杂、不确定和不受控的特点,这就要求它必须具备对环境和事物高效、准确的识别、感知、理解、判断及自主行动能力。本文首先对服务型机器人导航技术近几年的发展现状以及未来的发展趋势进行了研究。经过对比后发现,基于视觉的导航方式相比于其他非视觉类传感器的导航方式,由于获取的是视觉信息,以及后续处理手段更接近于人类的思考方式,因而,成为目前较为前沿的研究领域。本研究结合计算机视觉和深度学习技术的发展趋势,针对目前移动机器人室内环境下的导航应用需求,提出了一种把双目立体视觉和基于机器学习的物体识别技术相结合的机器人导航方法,扩展了双目测距和物体识别技术的应用范围,提高了机器人对环境的感知能力。相较于以往的基于单目视觉的机器人导航方式,本研究利用基于视觉的物体识别技术,通过构建融合识别及决策模型,赋予机器人一定的环境认知能力;并结合三维立体重建技术,实现场景中的目标测距;把视觉平面识别信息与三维重建信息相融合,最终实现增强机器人对环境感知能力的目的。为了验证实验模型的效果,本研究以自主搭建的移动平台为载体,设计并实现了一个将双目立体视觉和物体识别技术相结合的实验系统。提出的双目立体视觉和物体识别相结合的导航方法,当应用到搭建的实验平台上时,能够实现对实验场景下物体的识别和测距,能够实现移动平台的自主导航控制。经验证,效果达到预期目的。更多还原

【Abstract】 In recent years,with the development of science and technology,it brings huge social demand to robot related industries.Because the robot can complete some repetitive and dangerous tasks quickly and accurately.Different from industrial robots,service robots have more flexible application scenarios.In terms of technical and performance requirements,the focus of the two is different:industrial robots focus on the repetition precision of action,while service robots focus more on the ability to perceive and interact with the environment.Because the working environment of the service robot is complex,uncertain and uncontrolled,it requires that it must have the ability to recognize,perceive,understand,judge and act independently of the environment and things.In this paper,the development status and future development trend of service-oriented robot navigation technology in recent years are studied.After comparison,it is found that the vision based navigation mode is a more frontier research field than other non-visual sensor navigation methods,because it is the visual information and the follow-up methods are closer to the human thinking mode.In view of the development trend of computer vision and deep learning technology,this paper proposes a robot navigation method combining binocular stereo vision with object recognition technology based on machine learning in view of the current navigation application requirements under the indoor environment of mobile robots,which extends the needs of binocular range finding and object recognition.With the scope,the robot’s ability to perceive the environment is improved.Compared with the former method of robot navigation based on monocular vision,this research makes use of visual based object recognition technology to give the robot certain environmental cognitive ability by constructing the fusion recognition and decision model,and combining the three-dimensional reconstruction technology to realize the target location in the scene,and the visual plane recognition information.Combined with 3D reconstruction information,the purpose of enhancing the environment perception ability of robots is finally achieved.In order to verify the effect of the experimental model,this study designs and implements an experimental system that combines the binocular stereo vision with the object recognition technology with the independent mobile platform as the carrier.The combined navigation method of binocular stereo vision and object recognition,when applied to the experimental platform,can realize the identification and distance measurement of the object under the experimental scene,and can realize the autonomous navigation control of the mobile platform.The results have been verified to achieve the desired purpose.更多还原