Research On Three-dimensional Reconstruction Of Airfield Based On Oblique Photogrammetry

With the rapid development of computer graphics and related hardware industries,especially the rapid development of 3D reconstruction technology based on oblique photogrammetry,there are new requirements for 3D modelling of airports.Airports are special sites where aircraft operate 24 hours a day,so it is a challenging research topic to model airports in 3D with high efficiency,high quality and low cost.Traditional airport modelling using satellite images or manual on-site photography has the disadvantages of high labour cost,long time consuming and poor modelling results.Large-angle oblique stereo images are ideal for 3D modelling of airports due to their stable geometric model,wide coverage area and rich texture information.However,factors such as weather changes and camera lens distortion make it difficult to generate 3D models automatically by computer.Therefore this thesis addresses the key issues of how to improve the accuracy and reduce the modelling costs of 3D models for this particular site of the airport.The research in this thesis is as follows:1.Firstly,a fixed-wing UAV with five lenses is used to take large-scale aerial photographs of the airport area and build a 3D model of the whole airport,and then a small rotary-wing UAV with a single lens is used to take aerial photographs to model a single facility of the airport-the sliding platform,so as to make up for the shortcomings of the lack of accuracy of a single facility in the wide-scale modelling by means of supplementary photography.In this thesis,the corresponding UAV flight routes are planned according to the above two shooting characteristics to make the acquired images better and more complete;secondly,the acquired images are partially removed according to the image preprocessing principles and then reconstructed in 3D to form a preliminary 3D model of the airport.2.On the basis of the preliminary airport 3D model,two different UAV encircling flight routes are planned according to the size and height of the airport sliding platform and the UAV encircling flight height calculation formula,and then the point cloud data obtained from different UAV flight parameters are used to improve the 3D model effect of the sliding platform through the two nearest neighbour matching algorithm to obtain a more complete sliding platform shape;secondly,for the airport containing Secondly,based on the inverse process modelling,a lightweight texture mapping method is proposed to extract the repetitive cell texture information and reconstruct the image texture by means of rotation and translation to reduce the consumption of computer hardware.The experiment proves that the UAV flight route planned in this thesis can better obtain the complete image of the airport,and the tilt photography 3D modelling method adopted in this thesis has higher modelling efficiency and stronger visual realism of the model compared with the traditional airport modelling method,it can be obtained that the two-two nearest neighbour matching algorithm adopted in this thesis has 133657 point clouds before matching and 135171 point clouds after matching.The point cloud is denser and the shape of the object is more clearly visible;the lightweight texture mapping algorithm adopted in this thesis can achieve an image compression rate of over 75%,which can effectively reduce the consumption of computer hardware when modelling.