Research On Three-dimensional Ocean Visualization Based On Multi-scene Modeling

As the resources on the earth's continent are gradually depleted,more and more countries are turning their attention to the ocean which is full of natural resources.Therefore,the frequency of maritime disputes between China and neighboring countries at sea has gradually increased over the past few years.The rational supervision and development of Chinese territorial sea resources is of great significance to Chinese step towards a maritime power.Traditional marine surveillance technology mainly relies on two-dimensional video images,which is increasingly unable to fully meet the needs of relevant departments.With the rapid development of science and technology such as virtual reality and deep learning over the past few years,it has become possible to improve the intuitiveness and accuracy of marine supervision technology.This project is based on this idea to propose a visualization method for 3D ocean scenes based on OSG.This method uses deep learning neural networks to detect and recognize 2D images.After acquiring the information in the image,it combines 3D max with osg Ocean and 3DS max for 3D modeling,the final 3D ocean scene will be presented on a Qt-based visual user interface.Compared with the traditional marine supervision technology,this method can more vividly and intuitively display real-time ocean scenes,which can bring certain convenience to the supervision of related departments.The main work of this thesis is as follows:On the one hand,the target detection algorithm R-CNN is used to detect the ship objects in the two-dimensional image,and then the fine-grained image classification network is used to specifically classify the identified ship objects,so that the number and types of all ships in the image can be obtained.Then based on osg Ocean,fast Fourier transform and skybox principle,a three-dimensional ocean model is established,and 3DS max is used to perform three-dimensional modeling for the identified ship.The data results of image recognition will be sent to the system user interface built by Qt in the form of XML files,and the interface will call the ocean scene and ship model accordingly according to the results of image recognition.This basically realizes the three-dimensional visualization of the ocean scene.On the other hand,in order to make the visualization method better applicable to the complex and changing ocean environment,this project continues to expand the functions and optimize the performance for different application scenarios.The function expansion part mainly focuses on creating three-dimensional ocean scenes under different time and weather conditions for the system to flexibly switch according to actual conditions.The sky box transformation method can be used to create a variety of atmospheres such as cloudless,cloudy and dusk;while the particles system of OSG can achieve rain & snow weather effects and explosion smoke effects with different concentrations.The performance optimization part is mainly to improve the rendering efficiency of the three-dimensional visual scene.The levels of detail model(LOD)is used to optimize the complex scene of multiple objects to improve the rendering performance of the entire scene.Finally,through experiments,the improvement of the rendering efficiency of the three-dimensional visualization scene with the LOD model is verified.