The resolution of hyperspectral sensors is typically much less than that achieved by panchromatic and multi-spectral sensors due to fundamental noise limitations. The proposed work will demonstrate a new and innovative technique for improving the spatial resolution of hyperspectral image data and develop software implementations of the algorithm and the required multi-sensor registration. This technique, called Color Sharpening, combines a multi-spectral image with a lower spatial resolution hyperspectral image to produce a product that has the spectral properties of the hyperspectral image at a spatial resolution approaching that of the multi-spectral image. This approach is contrasted to panchromatic sharpening that is to combine a single panchromatic band with several lower resolution color bands. Color Sharpening is a departure from procedures that have been implemented before, and offers the benefits of high spatial and spectral resolution for subsequent hyperspectral processing such as terrain classification and target detection. Under Phase I, the resolution enhancement algorithm and required high quality registration procedures were developed. Tests were conducted on image data collected from airborne and satellite platforms with both high-resolution multi-spectral sensors and hyperspectral sensors. Under Phase II software tools for both registration and resolution enhancement will be developed and delivered to the government.
Benefits: Hyperspectral imaging systems are assuming a greater importance for a wide variety of commercial and military systems. The reason for this increased interest is the fact that a hyperspectral sensor of a given spatial resolution or pixel size will reveal information on the scene that can not be obtained by single band or multi-spectral sensors. For commercial geological remote sensing, the spectral properties of the surface tell the existence of minerals of potential commercial value. For military surveillance systems, a hyperspectral system can often be used to detect and identify a military target, even though the target may occupy less than a single pixel. The ability of the hyperspectral sensor to behave as a sensor with higher spatial resolution does not mean that there is not a place for high-resolution imagery. In fact, many operational and planned hyperspectral sensors are coupled with a high-resolution instrument. In some concepts, this high-resolution sensor is used as part of a detection process. The Adaptive Spectral Reconnaissance Program (ASRP) sensor demonstration is one such example where the hyperspectral sensor was used for wide area search and each detection was then confirmed by a panchromatic imager. The same combination of sensors could be used together to optimize detection by using the panchromatic imager to improve the resolution of the hyperspectral and then running detection algorithms on that data. This method of pre-detection fusion can be further extended if instead of a panchromatic line scanner a color line scanner is used. Our technology, called Color Sharpening, provides a means to optimally combine a high-resolution multi-spectral sensor with a lower resolution hyperspectral sensor. There are many applications for a technology that can optimally combine the data from these two types of sensors. An existing procedure often called 'sharpening' combines the output of the analysis of the hyperspectral data with the high-resolution image. The Color Sharpening procedure is superior to that approach since the result is a hyperspectral product that can be further analyzed. This means that data from existing sensors in space and on aircraft can be combined to give data that is nearly equivalent to data from a high-resolution hyperspectral imager. Since there is no current high-resolution space hyperspectral sensor, this is a unique capability. For example, data from the commercial multi-spectral sensor on the IKONOS satellite and the space-based hyperspectral sensor, Hyperion, can be combined to aid in the identification of vegetation types and crop health. There are two military applications: target detection and scene classification. Military sensors with both a high-resolution imager and a hyperspectral sensor include ASRP, HYCAS, and SPIRITT. These sensors currently plan on using the sensors serially: initial detection decision on hyperspectral followed by confirmation with the panchromatic. Processing of the data from the two sensors together will result in higher probabilities of detection and lower false alarm rates. For the scene classification application, which is used for terrain trafficability, crop assessment, damage assessment, detection of non-isolated 'target' materials, as well as intelligence, the development of image products by merging multi-spectral data with hyperspectral data to achieved higher resolution will aid the work of the Image Analyst. TRA has already investigated many of these applications in Phase I and will be investigating these areas for potential future application The market timing for this effort is ideal because: 1). New space-based hyperspectral sensors are in the development phase. 2). With the shortage of space-based hyperspectral assets, software that can better exploit combinations of data from different sources will be in demand. 3). Military surveillance technology employing hyperspectral sensors is being investigated in all three branches of the military for a variety of applications. This proposed program represents a key opportunity to develop a new technology with high likelihood potential for commercial success that also has high promise for the United States military.
Keywords: Hyperspectral, Sharpening, Multi-spectral, Registration, Satellite, Sensor
