An Indoor Positioning System Based On Ground Feature For Robot Vision

With the development of science and technology,more and more robots have begun to serve our lives,which has brought great convenience to our lives,saved the company a lot of labor costs,and made outstanding contributions for the national defense.Robots can be arbitrarily positioned using GNSS(Global Navigation Satellite System)outdoors,but in indoor environments,satellite signals are susceptible to interference,making it impossible for robots to use GNSS for precise positioning indoors.Because of the urgent need for indoor positioning services,indoor positioning technology has become a hot topic at home and abroad.With the rapid development of computer technology and image processing technology and the popularization of image sensors,computer vision-based positioning technology has become a research hotspot of indoor positioning technology and gradually replaces the traditional communication-based indoor positioning system because of its low deployment cost,strong autonomy and high positioning accuracy.Based on this,this paper aims to solve the problem of indoor positioning of robot through machine vision technology,designs and develops a indoor positioning system based on machine vision with good real-time performance and high positioning accuracy.The main research contents and innovations are as follows:(1)By reviewing a large number of literatures,this paper systematically analyses the principle of the current mainstream indoor positioning technology and the domestic and international research status,summarizes the shortcomings and limitations of the mainstream indoor positioning technology,and then proposes a new indoor positioning system based on machine vision.(2)A method of indoor positioning based on ground feature for robot vision is designed.This method captures ground pictures by ultra wide-angle camera installed at the bottom of the robot,and then detects the edge of the captured pictures by Canny operator to obtain the contour of the pictures.The camera's intrinsic and extrinsic parameters are obtained by camera calibration,and then the point correction algorithm is innovatively used to restore the contour points of the image to the undistorted plane,which significantly reduces the running time of the positioning algorithm;The floor straight line is successfully fitted from the set of contour points containing noise points by innovatively combining the RANSAC algorithm with the least squares method,and then the fitted floor straight line is restored to the real physical space by IPM algorithm.Then the real position and pose data of the floor straight line in the camera coordinate system are calculated.Finally,the extended Kalman filter is used to fuse the real position and pose data of the floor straight line and the wheel encoder data of the robot car to obtain the optimal positon and pose estimation of the robot.Using extended Kalman filter to fuse data is also one of the innovations of this paper.(3)In order to verify the validity of the positioning algorithm proposed in this paper,a visual positioning system is designed and implemented independently.It mainly includes the construction of hardware platform,the selection of related hardware and the design and implementation of corresponding software algorithms.Then experiments are carried out on this experimental platform to verify the performance of the proposed visual indoor positioning algorithm.The experimental results show that the proposed visual indoor positioning algorithm has high positioning accuracy,good real-time performance and low deployment cost,and is very operable.