Research On Construction Of Virtual Battlefield Environment Based On Multi-source Data

With the development of computer graphics,virtual reality and other technologies,combat simulation technology plays an important role in the military field by virtue of its high efficiency,economy,security,and science.The virtual battlefield environment is a key part of the combat simulation system.A high-quality virtual battlefield environment can also bring real visual perception to trainers and improve the experience of the combat simulation system.As the requirements for the scale,accuracy,and real look of virtual battlefields increase,the cost of producing large-scale,high-quality virtual battlefield environments is also increasing.In order to improve the construction efficiency and effect of the virtual battlefield environment,this paper studies from the aspects of procedural road modeling and terrain landscape synthesis.In procedural road modeling,this paper proposes a procedural road modeling method based on OSM(OpenStreetMap)data.By analyzing the topological structure and attributes of road vector data in OSM data,a set of road network structure extraction algorithms,road height estimation strategies,and road centerline representations based on Hermite splines are proposed.In the construction of the road model,a procedural road modeling method based on the deformation of the Hermite spline grid is further proposed,which more flexibly and efficiently models the road network sections and intersections.In the synthesis of terrain landscape,this paper studies from the classification of multispectral remote sensing image and the synthesis of procedural terrain texture.The classification results of multispectral remote sensing image are used as the basis of procedural terrain texture synthesis to improve the efficiency and quality of the virtual battlefield environment.In the classification of multispectral remote sensing images,this paper proposes a QWLD-GLCM fusion feature,which combines WLD(Weber Local Descriptor)and quaternion to extract spatial characteristics with band correlation in multispectral remote sensing image classification tasks.In combination with GLCM(Gray-level Co-occurrence Matrix),more effective QWLD features are collected in smaller neighborhoods,which improves the classification effect under the condition of small samples.In procedural terrain texture synthesis,this paper proposes a method for synthesizing terrain texture based on Voronoi noise,which can use a single texture to effectively eliminate the sense of texture repetition existing in ordinary texture tiling.At the same time,a layering based on height and slope mixing is proposed.The procedural terrain texture enables a natural transition between different texture layers and provides more detail to the terrain texture.