基于视觉里程计和自然信标融合的移动机器人定位算法

发布时间：2018-05-20 20:47

本文选题：机器视觉 + 机器人　；参考：《农业工程学报》2017年10期

【摘要】：针对未知室内环境下移动机器人的定位问题,提出了一种基于单目视觉和自然信标融合的移动机器人定位算法。该方法采用普通摄像头和全景摄像头结合的方式进行定位。采用普通摄像头提取天花板特征,进行视觉里程计定位。全景普通摄像头提取周边环境及天花板信息,采用Harris-SIFT构建自然信标库,并且进行自然信标检测与识别。在两者的基础上,移动机器人进行实时定位。利用自然信标定位时,提出方向滤波器结合视觉里程计和惯导系统估计旋转角度,利用自然信标求取定位特征点,两者结合进行定位。该方法有效地结合了相对定位的快速性和绝对定位的准确性。在室内环境进行试验,其中自然信标定位在x,y方向的平均误差分别为38.66和31.54 mm。相比视觉里程计而言,结合自然信标定位的平均误差分别减小了32.53%和68.68%。融合视觉里程计的自然信标定位运算速度相比仅靠自然信标定位而言提高了约6倍左右,同时与视觉里程计定位处于同一数量级。试验结果表明相对于视觉里程计而言,所提算法大大提高了定位精度,同时不影响算法的实时性。
[Abstract]:Aiming at the localization problem of mobile robot in unknown indoor environment, a mobile robot localization algorithm based on monocular vision and natural beacon fusion is proposed. This method uses common camera and panoramic camera to locate. The ordinary camera is used to extract the ceiling features and locate the visual mileage. The panoramic common camera extracts the information of the surrounding environment and ceiling, constructs the natural beacon database with Harris-SIFT, and detects and recognizes the natural beacon. On the basis of both, the mobile robot carries on the real-time localization. When using natural beacons to locate, a directional filter combined with vision odometer and inertial navigation system is proposed to estimate rotation angle, and natural beacons are used to obtain localization feature points. This method effectively combines the relative location speed with the accuracy of absolute location. The experimental results show that the average errors of natural beacon positioning in XY direction are 38.66 and 31.54 mm. respectively. Compared with visual mileometer, the average error of combining with natural beacon is reduced by 32.53% and 68.68% respectively. Compared with the natural beacon location, the speed of the fusion vision odometer is about 6 times faster than that of the natural beacon location, and it is in the same order of magnitude as the visual mileage location. The experimental results show that the proposed algorithm greatly improves the positioning accuracy and does not affect the real-time performance of the algorithm.
【作者单位】：浙江大学流体动力与机电系统国家重点实验室;
【基金】：国家自然科学基金创新群体基金(51521064) 杭州市创新链产业链重大科技创新项目(20132111A04) 杭州市重大科技创新项目(20142013A56)
【分类号】：TP242;TP391.41

【相似文献】