DynoViz: Dynamic Visualization of Large Scale Satellite Data

With the rapid increase in publicly available satellite data with high-resolution, so did the demand on interactive visualization of this data on a web map. This data is often high-resolution, with up-to daily three-meter resolution data, and multi-spectral with up-to 15 bands per datasets. Users in various fields need to interactively explore terabytes of this data via a web-based interface to choose the right data for their projects. Unfortunately, existing systems are either single-machine with limited scalability, or they do have limited support for dynamic visualization. Moreover, most systems pre-render visible bands only, i.e., RGB, and ignore other bands even though many scientific domains are more interested in other bands, e.g., infrared. This work introduces DynoViz, a novel system for dynamic web-based scalable visualization of satellite data. It visualize big satellite data on a regular web-based interface through multilevel dynamic-resolution visualization. The design consists of three main parts. First, a pre-generation process produces a limited set of select static tiles stored on disk. This process is controlled with a parameter to balance interactivity and disk usage depending on the application needs. Second, a dynamic on-the-fly generation technique uses a raster index to provide real-time visualization of high-resolution regions of the map. Third, a web-based interface provides client-side rendering of tiles according to the user requirements and can handle multi-spectral data with no additional overhead on the server. Experiments with terabyte-scale datasets show that DynoViz is up-to an order of magnitude faster than other distributed systems in the pre-generation phase and uses 60 times less disk storage without sacrificing the interactivity.