基于单目摄像头移动机器人导航研究

发布时间：2018-07-05 05:21

本文选题：视觉机器人 + 图像处理　；参考：《辽宁工程技术大学》2017年硕士论文

【摘要】：如今,越来越多的机器人被应用在日常生活中。随着人工智能和图像处理技术的高速发展,越来越多的人也开始投入到视觉机器人的研究行列当中。本文主要研究视觉机器人在静态室内环境下的导航。首先应用单目摄像头获取机器人所在环境的图像,然后进行图像处理,提取特征点和障碍物。通过提取的特征点,应用尺度不变特征转换算法进行目标识别,识别出图像中的路标;通过摄像头3D距离计算公式得到机器人与障碍物和路标的距离。然后通过获取的路标对机器人进行定位,获取机器人的具体位置。本文针对移动机器人在定位过程中,传感器测量误差和机器人位姿误差导致系统定位精度急剧下降,提出一种多新息卡尔曼滤波算法。在标准卡尔曼滤波的基础上,引入抗差权因子和自适应因子抵制测量误差和位姿误差。同时,引入多新息,即多个时刻的新息向量,进一步提高系统的精度。最后通过模糊神经网络实现机器人的路径规划,使机器人可以从出发点避障并找到一个最优路径到达目标点。本文路径规划采用全局路径规划和局部路径规划相结合的路径规划方法。机器人在确定自己位置后利用全局路径规划寻找一个最优子目标点,然后利用模糊神经网络控制器对机器人进行局部路径规划,依次循环,直到机器人到达最终目标点。通过栅格法建立仿真环境,在无障碍物环境、少量障碍物简单环境和大量障碍物复杂环境三个环境验证算法的可行性。仿真结果表明本文算法可以使机器人从出发点快速找到一条最优路径到达目标点。
[Abstract]:Nowadays, more and more robots are used in daily life. With the rapid development of artificial intelligence and image processing technology, more and more people begin to study the visual robot. This paper mainly studies the navigation of visual robot in static indoor environment. Firstly, the single camera is used to obtain the image of the robot environment, and then the image processing is carried out to extract the feature points and obstacles. By using the feature points extracted, the target recognition is carried out by using the scale-invariant feature transformation algorithm, and the road signs in the image are recognized, and the distance between the robot and the obstacles and landmarks is obtained by the 3D distance calculation formula of the camera. Then the robot is located by the acquired road sign, and the specific position of the robot is obtained. In this paper, a multi-innovation Kalman filter algorithm is proposed to solve the problem that the sensor measurement error and the robot pose error lead to a sharp decrease in the positioning accuracy of the mobile robot. Based on the standard Kalman filter, the robust weight factor and adaptive factor are introduced to resist the measurement error and pose error. At the same time, the precision of the system is further improved by introducing the multi-innovation vector, that is, the multi-time innovation vector. Finally, the path planning of the robot is realized by fuzzy neural network, so that the robot can avoid obstacles from the starting point and find an optimal path to reach the target point. In this paper, global path planning and local path planning are used in path planning. After determining its position, the robot uses global path planning to find an optimal sub-target point, and then uses a fuzzy neural network controller to plan the local path of the robot and loop it in turn until the robot reaches the final target point. The simulation environment is established by grid method, and the feasibility of the algorithm is verified in three environments: no obstacle environment, a small number of obstacles simple environment and a large number of obstacles complex environment. Simulation results show that the proposed algorithm can quickly find an optimal path from the starting point to the target point.
【学位授予单位】：辽宁工程技术大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP242

【参考文献】