Fast 3D Visualization Of Large-scale Seismic Data Based On Deep Clustering Index

Nowadays,the volume of data is growing with the accumulation of precision,and its structure has become more complex.In order to bring users a more intuitive experience,the data needs to be processed and presented in another representation different from the one-dimensional one,so the use of 3D visualization technology to present data objects in multiple directions is an effective means.As an important technology in computer graphics,3D visualization technology has been applied in various industries,including the medical field with image technology to the human body’s tissue structure and pathology for more intuitive and clear observation,in the field of architecture using technology to render the building,in the marine atmosphere can be used to simulate the dynamic environment,in the oil and gas geological exploration can clearly analyze the terrain structure,as well as machinery manufacturing,transportation,biological fields,etc.,can use 3D visualization technology to play a good role in guiding people to clearly perceive the object objects and research exploration.However,the development of 3D visualization technology is surrounded by many problems in the face of a wide variety of data objects.On the one hand,the traditional volume rendering algorithm has a high complexity and requires a large memory space,which requires high requirements on computer hardware.On the other hand,for the data with a certain size,the calculation speed is slow,and it is not possible to load a large volume of data into memory at one time based on traditional drawing algorithms,and there will be problems such as picture delay and image jump lag in interactive display.Therefore,many 3D visualization schemes have certain limitations.In this paper,in order to solve the performance bottleneck of large-scale seismic data processing,optimizing the index structure algorithm and the smooth loading and rendering modeling of data are the key elements to improve the efficiency of rapid 3D visualization of large-scale seismic data,fully considering the needs of system browsing speed and fluency of large-scale seismic data,studying the index structure algorithm,the viewpoint trajectory prediction model and optimizing the scheduling and loading scheme of seismic data,focusing on solving the problems of low efficiency of index structure and the delay and lag of modeling screen caused by concentrated loading of large-scale data.This paper proposes a fast 3D visualization technique under deep clustering:(1)To address the problem of inefficient indexing of large-scale data due to computer memory limitations,we build a data structure algorithm that combines a deep clustering model with spatial curve encoding to further optimize the indexing capability of the data structure while learning the spatial characteristics of data hidden variables.(2)In view of the problem of picture lag caused by a large amount of data loading in a centralized manner,a viewpoint trajectory prediction model based on temporal convolutional network is proposed to improve the accuracy of prediction according to the location information of historical viewpoints and current viewpoints,which is the basis for scheduling loading in the next step.(3)To solve the problem of slow data rendering caused by system load,a data scheduling loading algorithm based on visibility judgment is proposed to dynamically load and release data in different regions,thus reducing system load and speeding up data rendering efficiency.This paper aims to solve the application problems of rapid 3D visualization of large-scale seismic data,as well as to provide a theoretical basis for research in related fields.The experiments use real geological data as the experimental dataset to test and evaluate the performance of this paper’s scheme: the dataset is divided into three groups of different sizes and proportions for comparison: Group A with small data volume,Group B with moderate data volume and relatively uniform distribution,and Group C with larger and more dispersed data volume,the data information includes burial depth,extent,thickness and extension trends.The experimental test section evaluates the efficiency of the indexing structure,the effect of viewpoint prediction,the data scheduling and loading module,and the actual system interaction performance of this paper,respectively.The results show that the index time of VDEC-HRT is reduced to 55.89%-72.22% compared with other methods.Compared with the lagrangian interpolation algorithm the correctness of TCN model predicting viewpoints under different time durations is increased by 12.08%～22.7%,and the real-time frame rate of the interactive display is also stable on larger data sets,and the overall rendering performance and quality of the system can achieve the expected results.