| Using multi-rotor drone or other small mobile carriers in conjunction with monocular cameras,the spatial three-dimensional measurement of outdoor targets can be flexibly carried out,as three-dimensional information can be recovered through image processing and other means.Nevertheless,there are limitations about this approach,such as large computational cost,low accuracy,and susceptible to environmental influences.However,the popularization and lightweight of photoelectric sensors,such as Lidar,make it possible to carry multiple photoelectric sensors on a single small mobile carrier,which could combine the advantages of different sensors,achieving more efficient and accurate spatial three-dimensional measurements.This thesis implements a software for spatial three-dimensional measurement system,connecting hardware carriers which load Lidar,monocular camera,Inertial Measurement Unit(IMU)and GPS.Without the requirement of pre-calibrating different sensors,it realizes the registration and fusion of three-dimensional Lidar point clouds and two-dimensional color images.At final,the software generates a three-dimensional point cloud model with real color texture and scale information,while visualizing the model and other sorts of data.Principal works of this thesis is as follows:(1)Researching different methods about registering and fusing three-dimensional Lidarpoint cloud and two-dimensional color image,before determining the data processing method and procedure of different sensors;(2)Research on related algorithms of IMU/GPS fusion positioning,multi-view three-dimensional reconstruction,heterogeneous point cloud registration have been done.Firstly,based on Error State Kalman Filter(ESKF),an IMU/GPS odometer is implemented.Then,AKAZE and Structure from Motion(Sf M)technology are used to reconstruct the two-dimensional image sequence into three dimensions.Finally,using TEASER and Iterative Closest Point(ICP)algorithms to register and fuse the image point cloud and Lidar point cloud.(3)Complete the specific design and development of the software for spatial three-dimensional measurement system.On the basis of Robotic Operating System(ROS),the software realizes the data interfaces for different sensors,the encapsulation of algorithms mentioned above,as well as the automatic process to deal with incoming data.In addition,the software combines Foxglove to realize a web visualization interface,for modular data display,interaction and control functions.(4)Verify the feasibility of the software.Using a DJI M300 multi-rotor drone which loads a DJI L1 measurement unit to measure an outdoor building target by flying around.After then the data is transferred into the software deployed on a local computer to generate the measurement result.Through comparison and analysis,the software implemented in this paper has considerable running efficiency and output accuracy. |
