Ground-based Monocular Vision Odometer Positioning Research

With the rapid development of science and technology,people need more and more positioning services.While providing location service requirements,obtaining location accuracy is also important and has become a hot spot in this technology area.In the indoor environment,in the robot field,GPS and Bei Dou signals are shielded due to factors such as buildings,which makes it difficult to obtain GPS and Bei Dou signals in indoor environments.Therefore,under such circumstances,many researchers have proposed solutions based on visual indoor positioning in recent years.The image is used in various complex environments to sense and recognize,and the relevant information of the feature points in the image is obtained,and the visual indoor positioning also uses indoor positioning on the mobile robot,so the visual positioning can also reduce the deployment cost.In order to solve the problem of navigation and positioning of robots in the room,this paper proposes a method based on the ground feature point robots monocular vision odometer positioning research aims to rely on the visual and odometer to better improve the positioning accuracy and reduce hardware costs.In order to expand the vision of the camera,the fisheye camera is used in this project.Then the fisheye camera is installed under the chassis of the car to correct the pictures taken by the fisheye camera in real time.The internal and external parameters are obtained by calibration of the camera,and the camera is further calculated by a new method.Install the attitude angle on the trolley,establish the plane imaging model to obtain the correspondence between the world coordinate system and the pixel coordinate system;and carry out the point correction image method proposed in this project experiment,this method can enable the entire system to achieve real-time operation and reduce system processing time.The least square method and RANSAC algorithm are then used to fuse and extract the feature points and feature lines of the floor.This part is also one of the experimental innovation points in this paper;then the relevant feature information is obtained and transformedinto the physical space through the inverse perspective transformation algorithm.Get real physical information;obtain real information and carry out extended Kalman fusion with the mobile robot odometer,and obtain the relative position information with the initial position of the mobile robot as a reference point in real time.The system runs in the relatively low-cost Raspberry Pi and STM32.In order to reduce the hardware cost,the system adopts an improved ground feature point algorithm and adopts image difference to remove the interference information of the mobile robot itself.The camera adopts a cheap fisheye camera to expand the pair.The ground field of view and real-time correction of fisheye cameras and real-time positioning,this system is easy to operate and high positioning accuracy,low cost.The experimental results show that the above-mentioned visual odometry location method can obtain the relative position information of the camera and the floor edge line more accurately,and use the Kalman expansion algorithm for odometer fusion,making the method more accurate than the traditional odometer positioning.Using the experimental method of the subject can make the low-cost Raspberry Pi processing the camera to collect ground images in real time.The PC remotely controls the Raspberry Pi to set the destination position information,which brings convenience to the operator and is cheaper and less expensive than the SLAM positioning robot.Cumulative error occurs.The conclusion is to use the experimental method of the subject to improve the positioning accuracy on the basis of the original odometer location method and to reduce the cost in terms of hardware,and also to meet the real-time nature of the system.