| In recent years,a large number of relevant science and technology have been invested in the field of meteorology.Doppler weather radar is one of the important observation tools for detecting meso and micro-scale convective weather systems,which can help meteorologists to make accurate and timely analysis and judgment of severe convection weather and reduce economic losses caused by meteorological disasters.With the upgrading and upgrading of China's geostationary orbit meteorological satellites,it is a severe challenge for meteorologists to acquire valuable information from the massive meteorological data detected within a limited time and then transform the boring and obscure data into intuitive graphs or images.At present,in the existing mature Doppler weather radar business system,radar echo data is often displayed in the form of simple and intuitive two-dimensional plane images,which limits the radar echo details and spatial three-dimensional contour structure distribution.Throughout a few existing studies on radar 3d visualization,most scholars take the limited regular grid as the frame model of 3d reconstruction.Its limitations include two aspects: first,interpolation consumes memory resources and reduces processing efficiency;Secondly,the more data volume increases after interpolation,the more data points are involved in subsequent processing.In addition,a full volume Doppler weather radar scan can detect not only weather targets,but also non-weather targets,such as ground objects,birds and insects.Therefore,in order to improve the accuracy of echo data,it is necessary to filter non-meteorological targets.Based on the above,this paper is aimed at the research of Doppler weather radar echo data visualization.The main work and results are as follows:1.This paper introduced the characteristics of radar volume scanning data,and then analyzed various interpolation methods from polar coordinates to Cartesian coordinates.Various interpolation methods were evaluated based on the spatial continuity and structural characteristics of the reflectivity field.The experimental results showed that the Fourier interpolation method based on spectral analysis was the closest to the observed value.2.Aiming at the inefficiency of the original raster 3D visualization,a new radarmodeling scheme based on "point,line,sector,radar cluster" structure was proposed.Based on the new modeling scheme,the traditional iso-surface algorithm was improved in this paper.Firstly,the cube in the radar volume scan data was replaced by hexahedron,and then the intersection point of the contour line and the quadrangle vertex were compared to solve the ambiguity of the surface.Compared with the traditional MC algorithm,this algorithm had the advantages of resolving the ambiguity of the surface,reducing the number of voxel and improving the reconstruction speed.3.Based on the radar modeling scheme,the light cone projection algorithm was accelerated and improved by introducing a "cone cap" bounding box,modifying the sampling end condition,and adopting a variable step sampling method.Compared with the GPU-based ray-casting algorithm,this algorithm could not only ensure the quality of image rendering,but also reduce the time of first drawing and redrawing.4.Firstly,the characteristics of radar echoes were introduced.Then,the horizontal texture,vertical gradient and radial velocity were calculated by preconditioning radar echo,and the non-precipitation echoes existing in the echo data were identified and eliminated.Finally,the upper-layer elevation echo data was used to fill and smooth the “echo hole” to improve the recognition effect of the algorithm.Based on the radar data characteristics,this paper proposed a new modeling scheme combined with echo quality control algorithm to efficiently and accurately visualize radar echoes in two or three dimensions,which played an important role in weather prediction in the future. |
PDF Full Text Request