SBIR Phase I: GPU-based Raster Processing Algorithms

This Small Business Innovation Research (SBIR) Phase I project will test the feasibility of using Graphics Processing Units (GPU's) for substantially increasing the performance of raster Geographic Information Systems (GIS) software operations. Map Algebra is a linguistic and algorithm framework for combining and transforming raster data and is the basis for raster-based analysis in all contemporary desktop GIS products. It is used for many types of government and business needs including ecology, land conservation, business siting, watershed analysis, transportation planning, hazard risk assessment, wildfire propagation and many other domains. However, particularly when working with large data sets, raster processing is usually too slow to support real-time calculation for web-based applications, even when the work can be distributed across multiple machines. It is therefore mostly consigned to workstation environments and people trained in the techniques. This project will focus its efforts on raster processing for server-based applications, where it is believed that both the greatest improvements can be made, there is a substantial market opportunity and, given the types of applications to which these techniques are applied, great potential for positive social impact. Most contemporary work in geographic information systems is directed at three types of activity: a) database development; b) desktop analysis and map production; and c) web-based display of maps for a variety of purposes. This project is directed at enabling a far broader audience, through use in web-based software applications, to perform sophisticated analysis with real-time results. If successful, this will enable real-time, online calculation of open space acquisition models, sea level inundation impact, natural hazard assessment, military scenario planning and travel cost calculation, to name only a few applications. These applications are currently all tied to desktop systems due to the significant amount of time, memory and processing power required to execute the operations. By moving these calculations to the GPU, it is believed that at least an order of magnitude improvement in the performance can be made and such applications could move to from the desktop to a web-based platform.