Low-Latency Web-Based Seismic Data Rendering: A Collaborative Optimization Study on Storage, Transmission, and Visualization

In this paper, we propose an end-to-end co-optimization scheme for the three major technical bottlenecks faced by real-time visualization of large-scale seismic data on the Web side, namely, inefficient storage and computation, overloaded data transmission, and insufficient rendering performance. The main research contents are as follows: (1) We propose a four-dimensional spacetime indexing mechanism based on improved OpenVDS, enhance the spatial locality feature by Hilbert curve coding, and combine the hybrid compression strategy of LZ4 and Zstd algorithm to realize the synergistic improvement of compression efficiency and decompression speed; (2) We construct a multi-level adaptive tile segmentation algorithm based on Z-order octree, establish a dynamic adaptation mechanism between transmission granularity and network bandwidth, and build a parallel computing architecture based on WebGPU Compute Shader to realize real-time invalid data rejection and rasterization acceleration; (3) We design a Canvas double-buffered rendering architecture, innovate a perception-driven dynamic resolution regulation mechanism, and reach a dynamic balance between system load and visual quality through dirty rectangle differential updating and visually salient graded rendering strategies. The experimental results show that compared with the traditional scheme, the slicing query latency of this method is reduced from 800 ms to 320 ms; the first screen rendering time is optimized from 14.2 s to 4.1 s; and the peak occupancy rate of decompression CPU is reduced from 75% to 32%. The technical system in this paper provides a solution with universal reference value for real-time visualization of large-scale scientific data in Web environments.