Multiview Geometric 3D Reconstruction Based On UAV Sequence Images

In recent years,unmanned aerial vehicle(UAV)remote sensing technology has developed rapidly because of low cost and high efficiency,and the scope of its application has been gradually expanded.At present,UAV images can match vertical images of objects with their tilted images for constructing texture.This is the mainstream method of 3D models of urban scenes.There are many research methods to match the vertical and tilt images of unmanned aerial vehicles,but the research on the three-dimensional reconstruction after the matching of this method is less.How to make efficient and accurate feature matching and build 3D models based on UAV imagery is the key problem of UAV in 3D modeling application.It is also one of the research hotspots of UAV Remote Sensing in recent years.In this paper,the following research is carried out on the preprocessing of the UAV image and the matching and three-dimensional reconstruction after matching of the oblique and vertical images :1.Mask uniform light algorithm and Retinex uniform light algorithm is widely used in image uniform light,uniform color,image enhancement,in the comparative analysis of these two algorithms,found that the mask uniform light image more in line with human visual habits,the light,Color is more uniform.Retinex images are focused on highlighting the texture and other features of the image,but their color distortions make the image visually poor.In this paper,a three-dimensional reconstruction process based on the hybrid method of UAV uniform light,uniform flow process.The improvement process utilizes the characteristics of Mask and Retinex images,and discards the defect that the uniform light image is not effective in three-dimensional reconstruction using a single method.2.Based on the detailed introduction and analysis of the SIFT algorithm and the ASIFT algorithm,the simulation process of ASIFT algorithm is simplified based on multi-view geometry,and an improved method of UAV multi-view geometric feature point matching algorithm is proposed.View angle images can be quickly matched.Through experiments and experience,the judgment threshold of UAV multi-view geometric feature point matching method is set as 70%,on the basis of which,the algorithm is applied to the non-augmented 3D reconstruction experiment.The results show that the number of point clouds in the image is obviously increased after using the method proposed in this paper.The comprehensive experiment shows that the method proposed in this paper is feasible and effective and achieves the expected results.