Geographic Information Systems


OpenCL Studio was used to develop an application to geo-reference hyperspectral images collected by an airborne remove sensing platform. In addition to a hyperspectral imager the aircraft also carries LIDAR, GPS and gyroscopic sensors designed to record the orientation and location of the craft as well as the topography of the underlying terrain. All of this data must then be fused in order to determine the correct geo-spatial coordinates of every pixel recorded by the camera. Since the algorithm is the same for each pixel, the parallel processing hardware of a graphics card brought considerable performance improvements to the geo-referencing procedure.

Hyperspectral Imaging


Hyperspectral images record hundreds of wavelengths of light per pixel. While this relatively new type of imagery opens new possibilities for remote sensing, the accompanying processing and analysis algorithms are often computationally expensive. OpenCL Studio was used to develop a high performance classification system for hyperspectral images consisting of parallelized versions of standard classification algorithms including K-Means clustering, the Spectral Angle Mapper (SAM) algorithm and Pixel Purity Index (PPI). The application, while still work in progress, allows users to interactively select reference pixel, zoom into selected areas and control various variables.