Font Size: a A A

Simplification Algorithm Of Oblique Photogrammetry Point Cloud For Rendering Effect

Posted on:2022-10-08Degree:MasterType:Thesis
Country:ChinaCandidate:J S SuFull Text:PDF
GTID:2518306551470744Subject:Master of Engineering
Abstract/Summary:PDF Full Text Request
In recent years,oblique photogrammetry technology has been widely used in large-scale scene automation modeling and has become an increasingly important digital asset.In practical applications,because the model data may need to be continuously updated and intellectual property rights are protected,it is often necessary to centrally store the model data on the server.When the client application needs to render,the latest data is transmitted over the network in real time to complete the rendering.Due to the tight coupling of the triangle mesh in geometry,vertex index,and texture information,the scene data must be completely transmitted before rendering.When the network bandwidth is unstable,the data transmission time cannot be guaranteed,which may cause the client to experience stalls due to lack of rendering data.In this thesis,aiming at the problems existing in the establishment of LOD with triangular meshes,point cloud is used as the basic primitive of scene data to establish LOD models.To solve the problems of how to weight geometry and color attributes in existing methods based on minimizing geometric and color errors and the optimal rendering effect of the simplified model cannot be guaranteed,this thesis proposes a rendering-oriented simplification method,which conforms to the above-mentioned application scenarios.Point cloud is regarded as a surfel containing the attributes of position,normal,color,and radius,and the difference in the rendering effect of the model before and after the simplification in the local neighborhood is used as the error measurement factor.Simplify each neighborhood iteratively according to the size of the error,thereby reducing the amount of model data while ensuring the rendering effect of the model.The main contents of this thesis include:· First,this thesis proposes a simplified error measurement formula based on the model rendering effect.The geometric and color errors are unified from a visual perspective,and the difference in the rendering effect of the model in the local neighborhood before and after the simplification is used as the error measurement factor.In the error calculation,the viewpoint and the direction of the line of sight are generated for each local neighborhood.Project the surfels in the neighborhood and the representative surfel to a uniformly sampled plane,and calculate the color difference of the neighborhood surfels and the representative surfel at each sampling point to estimate the neighborhood rendering error.· Secondly,this thesis improves the iterative-based point cloud simplification process to improve the rendering effect of the simplified model.Aiming at the problem of uneven distribution of vertices on the surface of the simplified model in the existing iterative simplification process,this thesis improves the process.A surfel of the same size as the neighborhood is used to replace surfels in the neighborhood and the error value of the surrounding affected neighborhood is updated.In order to ensure the correctness of the recalculation of the rendering error in the surrounding neighborhood,only a small part of surfels in the neighborhood are deleted when replacing the surfels,and other surfels in the neighborhood are temporarily retained.· Finally,this thesis proposes a surfel radius estimation strategy based on line-of-sight,and realizes the generation of LOD models based on rendering effects.In this thesis,the radius of the surfel is used to divide the LOD level.The number of pixels occupied by a single surfel rendered on the screen is taken as a constant standard,and the radius of the current LOD-level surfel is estimated according to the distance between the camera and the model when the scene is roaming.The experimental results show that the error measurement method proposed in this thesis can uniformly measure the geometric and color errors of the model from the visual effect,and solves the problem of setting the error weights without standard values;The improved iterative simplified process can be better to solve the problem of uneven distribution of vertices on the surface of the simplified model,and improve the rendering effect of the simplified model;The multi-level point cloud obtained by using the surfel radius estimation strategy based on the viewing distance proposed in this thesis has good visual effects in large-scale scene visualization.
Keywords/Search Tags:Oblique photography, Point cloud simplification, LOD technology, Scene visualization
PDF Full Text Request
Related items