Research On 3D Reconstruction Technology Of River Water Surface Based On Point Cloud Data Of UAV Oblique Photogrammetry

The construction of virtual geographic environment at watershed scale is the primary research task of watershed information and intelligent management platforms such as "smart watershed" and "twin watershed",but the traditional three-dimensional modeling methods can not meet the requirements in terms of efficiency and accuracy.In recent years,UAV tilt photogrammetry technology has developed rapidly,which provides an effective means for rapid three-dimensional reconstruction of geospatial elements at watershed scale.Compared with other basin elements,the water surface area is easily affected by wind and waves,the reflection of shore elements and other factors,resulting in holes,sharp fluctuations,texture distortion and other phenomena in the threedimensional reconstruction results of oblique photography of basin elements,which affect the reconstruction effect.In order to solve this problem,the existing research usually uses "preset water surface" instead of "real water surface".Although the ideal reconstruction effect can be obtained from the visual aspect,it can not reflect the real water surface information.In fact,the real water surface information is very important for the comprehensive management of the basin.It can intuitively express the water color,floating objects,transparency and other information,and provide important decision-making basis for the supervision of water environment and pollutant emission reduction in the basin.Therefore,it has important research significance and practical value to realize the rapid reconstruction of real water surface information in the three-dimensional model of tilt photography.This thesis focuses on the fast three-dimensional reconstruction of water surface area in tilt photography three-dimensional model data,and studies the key technologies such as water edge extraction,water surface triangulation reconstruction and intelligent optimization and processing of texture image.The specific contents are as follows:(1)The data content and characteristics of UAV tilt photogrammetry three-dimensional model are analyzed,and the data preprocessing method and the accuracy evaluation method of waterline extraction are studied;(2)For the 3D model data of tilt photography,the waterline extraction algorithm based on vertex traversal and the waterline extraction algorithm based on region growth are designed respectively,and the performances of the two algorithms are comprehensively compared from two aspects of accuracy and efficiency.(3)An intelligent optimization and processing algorithm of water surface region texture image based on UAV multi view image data is designed,which improves the automation and intelligence level of water surface region model texture mapping.(4)The research results of key technologies in this thesis are combined with practical.Engineering projects,and demonstrated and applied in representative watersheds,which provides important support for the construction of real three-dimensional scenes of watersheds and the fine management of water environment.