助老助残机器人视觉智能定位系统研究

发布时间：2018-03-24 10:29

本文选题：助老助残机器人　切入点：嘴部检测　出处：《山东科技大学》2017年硕士论文

【摘要】：助老助残机器人是面向老年人、残疾人的一种服务型机器人,具有助老、助残、娱乐等功能,随着社会上老年人数量的增加,助老助残机器人的需求不断加大,对助老助残机器人的研究也受到了广泛关注。本文主要对一种应用到不可移动的助老助残机器人上的视觉智能定位系统进行研究。选用Bumblebee2双目相机构建立体定位系统。在深入理解双目立体视觉的相关原理之后,对Bumblebee2相机进行了二次开发,实现了对指定的任意目标物体的空间三维定位以及对受助者嘴部的空间三维定位。为实现对指定目标的定位,本文设计实现了通过位于受助者头部的激光笔指引的方式来指定目标,该种设计方法简单、方便、实用,避免了对大量不同目标的分类学习。对光斑的检测过程经历了图像的平滑处理、阈值分割、腐蚀处理、膨胀处理等过程,最后采用重心法求取光斑中心在图片中的位置。通过该位置可利用双目立体视觉得到指定目标在相机坐标系中的空间三维坐标。对于嘴部的检测定位主要根据Haar-like特征和Adaboost训练方法,采用OpenCV提供的Haartraining算法训练得到人脸分类器和人嘴分类器,利用得到的分类器进行人脸检测,再基于检测到的人脸区域进一步检测人嘴的位置,通过该位置即可利用双目立体视觉获得人嘴的空间三维坐标。本定位系统基于台式机,利用OpenCV和Bumblebee2的API在Microsoft Visual Studio 2008编程环境中编程实现。实验结果表明,在没有强光干扰的情况下,能很好地检测出激光光斑和受助者嘴部,Bumblebee2双目相机对嘴部和其他目标的定位精度在5mm范围内,满足项目需要。
[Abstract]:The disabled robot is a kind of service robot for the elderly and the disabled. It has the functions of assisting, assisting and entertaining. With the increase of the number of the elderly in society, the demand of the robot is increasing. The research of assistive robot has been paid more and more attention. In this paper, a visual intelligent positioning system applied to immovable robot is studied. Bumblebee2 binocular camera is used to construct stereo positioning. After deeply understanding the related principles of binocular stereo vision, The second development of Bumblebee2 camera is carried out to realize the spatial 3D positioning of any target object and the three-dimensional location of the recipient's mouth. In this paper, we design and implement the method of directing the target by the laser pen in the recipient's head. This design method is simple, convenient and practical. The process of spot detection has gone through the process of image smoothing, threshold segmentation, corrosion treatment, expansion processing and so on. Finally, the position of the spot center in the image is obtained by using the center of gravity method. Through this position, the three-dimensional coordinates of the specified target in camera coordinate system can be obtained by using the binocular stereo vision. The detection and location of the mouth are mainly based on Haar-like. Features and Adaboost training methods, The face classifier and mouth classifier are trained by Haartraining algorithm provided by OpenCV. The face detection is carried out by using the obtained classifier, and the position of human mouth is further detected based on the detected face region. Through this position, the three-dimensional coordinates of human beak can be obtained by binocular stereo vision. This positioning system is based on desktop computer, and realized in Microsoft Visual Studio 2008 programming environment by API of OpenCV and Bumblebee2. The experimental results show that, In the absence of strong light interference, the location accuracy of laser spot and Bumblebee2 binocular camera to mouth and other targets can be well detected in the 5mm range, which can meet the needs of the project.
【学位授予单位】：山东科技大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41;TP242

【参考文献】