| In recent years,UAV low-altitude remote sensing technology has gained wider and wider application in China.In the small-scale research area,UAV low-altitude remote sensing has many advantages compared with traditional satellite remote sensing,manned aerial photogrammetry,and manual measurement.The real-shot image is an important product of lowaltitude remote sensing of drones,which eliminates the problem of tilting shadowing.,has a wide range of application prospects.The research found that the real shot image is insufficient in the following aspects:(1)In terms of precision,most of the research at home and abroad is concentrated on small and medium scale,and the research on large scale real shot image is less,and the accuracy of the result is mostly outside 0.05 m.(2)In the production process,a complete set of real shooting video production workflows has not been proposed.(3)In terms of application,the application of real shooting images needs to be expanded.In view of the above deficiencies,this study proposes a workflow for making real-time images of UAV remote sensing,including image acquisition,motion recovery structure algorithm,multi-view stereo vision,digital surface model editing and multi-view image texture compensation.The details are as follows:(1)Taking the Miyun District of Beijing as the research area,the CORS-RTK method is used to lay control points on the ground,and the professional camera is used to collect images.(2)After obtaining the image,enter the internal processing,perform the puncture point on the image to identify the ground control point,and insert the coordinate information into the feature point;using the motion recovery structure algorithm,the position and posture of the camera exposure can be restored to generate a sparse point cloud;Based on the sparse point cloud,a multi-view stereo vision algorithm is used to generate a dense point cloud,and a dense point cloud can be used to create a digital surface model.(3)In the obtained digital surface model,there is a jagged shadow on the boundary of the house.This is caused by the blockage or multiple projection of tall trees.By editing the digital surface model,the shadow can be eliminated and the clear boundary of the house can be maintained.(4)Using the edited digital surface model to generate an orthophoto image,the orthophoto image is tilted,and the multi-view image texture compensation method is used to orthographically sort the image containing the tilted region,and the best orthophoto image is selected for texture.Compensation,obtaining a true shot image that eliminates tilt.After the real shot image is obtained,the randomly distributed check points are used for accuracy verification.Using the root mean square error formula to calculate,the horizontal accuracy is 0.0365 m,the vertical precision is 0.0437 m,and the resolution of the real shot image is 0.0166 m.A cadastral map was produced based on the generated real shot image as an application case of this study.This study proposes a complete UAV remote sensing real-image production process,and obtains a real shot image with an accuracy of less than 0.05 m.This is a high precision that is rarely achieved in other studies,and is done in rural homesteads.An application case.The workflow greatly saves manpower and material resources,fast and high precision,and can be widely used in rural real estate registration,national defense military,marine monitoring,land planning,geological disaster investigation,emergency rescue,land use investigation and other fields. |
PDF Full Text Request