The Research And Implementation Of 3D Reconstruction Based On Line Matching

How to make the computer feel the three-dimensional environment information has been one of the goals in the field of computer vision.3D reconstruction technology based on image is the process that the computer automatically calculate and match the two-dimensional geometric information and depth information of the scene according to the two-dimensional image of the scene and finally establish the three-dimensional model.The existing 3D reconstruction techniques generally use algorithms based on structure from motion(SFM,Structure from motion),and multi-view stereo vision algorithms(MVS,Multi-view stereo).By using SFM algorithm to process image sets,the sparse 3D point cloud model and camera pose information can be obtained,and the processing speed is fast.Through the MVS algorithm,the 3D dense point cloud model of the scene can be get,and the processing speed is slow.It often takes a lot of time and memory of the computer.After careful observation,we can see that the surrounding man-made PhotoScan cases have prominent geometric features of the line segments.Currently there are many mature line detection method,such as LSD(Line Segment Detector)line detection method and Hof line detection method,if the 3D information related to the use of these unique segment geometric feature extraction and matching,3D reconstruction efficiency can make more efficient.The lines of PhotoScan cases are very obvious,and the geometric structure of man-made structures can be more highlighted by using the 3D line model.In addition,the 3D point cloud model has poor reconstruction accuracy in the absence of texture,and the resulting 3D model often has holes.The 3D line segment model can provide more structural information and reflect the geometric topological relation of the scene.The unmanned aerial vehicle(UAV)has the advantage of small volume,high flexibility,convenient use and other advantages.At the same time,because aerial images captured by the unmanned aerial vehicle have advantages of large scale,wide viewing angle and high current,so the unmanned aerial vehicles have been widely used in various industries in recent years.It is convenient to acquire pictures of different views from large scenes through unmanned aerial vehicle.However,the UAV still needs to formulate relevant strategy when it performs the task,so that it can give full play to its optimal efficiency.Therefore,it is of great significance to make an automatic flight plan for the aerial images captured by the unmanned aerial vehicle.Through querying data,3D has been developed from point based motion recovery structure algorithm to line based multi view stereo vision algorithm,but each algorithm has its own advantages and disadvantages.At the same time,through analyzing pinhole camera model,epipolar geometry relationship and multiple line segment detection algorithms,it is found that 3D reconstruction based on line matching is feasible.This paper aims at the existing background and development,a 3D reconstruction technique of the unmanned aerial vehicle based on line matching is proposed.First of all,through the formulation of unmanned aerial vehicle automatic flight plan to achieve the goal of PhotoScan cases pictures collection.Then,the computer then performs correlation matching by identifying and extracting two-dimensional line segments from the image,then a 3D line segment is obtained.Finally,a 3D line segment model is obtained through a series of filtering,fusion and optimization.At last,four sets of the surrounding PhotoScan cases images are collected by the unmanned aerial vehicle,and the algorithm is tested and evaluated.The experiment compares the picture sets by adjusting several key parameters,after the observation and analysis,it is found that the algorithm can obtain a more accurate three-dimensional line model,and it has greatly improved the time consuming,basically achieved the desired results.This algorithm can still be improved in the following research.