Image processing for NASA applications
Software is being developed to improve geolocating Corona spy satellite photos from the 1960s. The winter and spring (2014) studies were done of how the accuracy of mapping depended on the starting point for optimization software. Five hundred good starting points were identified for each frame of a Corona image, and automated searches were done to minimize mapping error. It was expected that these searches would identify a few local minima where the search software would get stuck and some of the searches would find a true minimum error state. It was discovered that the lowest minimum error corresponded to the 30m resolution of the images used for calibrating the mapping. Other minima were scattered over the parameter space instead of clustering at a few points. At the suggestion of a National Aeronautic Space Administration colleague, mapping points were recalibrated on higher resolution imagery over several months. The starting point studies need to be repeated. Work continues on providing and updating data resources for the www.oceanmotion.org educational website. The updates are typically done twice a year.