When dealing with remotely sensed images, a parameter that aids in material identification is its emissivity. By extracting a materials emissivity spectrum, the probability of correctly identifying it greatly increases. The Alpha Residual algorithm was developed for just such a process. The Alpha Residuals emissivity extraction technique was originally developed for six channel Thermal Infrared Multispectral Scanner (TIMS) data. In dealing with TIMS data, it is known that the error at 300K is 1%. In recent years, detector technology has led to sensors with higher spectral resolution and an increased number of bands. The goal of this project is to determine if error is reduced due to increased spectral resolution and to produce an error plot that is temperature dependent. An error characterization algorithm was written in C++ to simulate one pixel in a 128-channel Spatially Enhanced Broadband Array Spectrograph System (SEBASS) image with known output. After the results for this were validated, the algorithm was applied to real data. Spectral response and the number of sample points per channel were shown to contribute to the resulting amount of error. As the number of sample points per channel increased, the error between the two alpha residual spectra decreased. When spectral response width increased, so did the error. When actual image data was used, the error had significant spectral structure, but was still very small. This technique is a pixel by pixel operation and thus, outputs an error spectrum for each pixel.
Baglivio, Michael, "Error characterization of the alpha residuals emissivity extraction technique" (1998). Thesis. Rochester Institute of Technology. Accessed from
RIT – Main Campus