Graduate STEM Fellow Profile
Matthew A. Lee
Initiating New Science Partnerships in Rural Education (INSPIRE)
Thesis: Exploiting Spatial and Spectral Information in Hyperdimensional Imagery
College/University: Mississippi State University
Research Advisor: Lori Mann Bruce
Degree Sought: PhD, Computer Engineering
Department: Computer Engineering
Research Focus: Image and signal processing, remote sensing, and artificial intelligence
Teaching Partner(s): William Funderburk
Description of Research
Hyperspectral images are much like the images from standard cameras except they have a lot more than three spectral bands (red, green, and blue). Typical hyperspectral images contain hundreds of narrow bandwidth spectral bands that are evenly spaced in a range from ultraviolet to infrared and perhaps into thermal regions of the electromagnetic spectrum. The fine sampling of the electromagnetic spectrum in hyperspectral images permits one to infer the chemical makeup of surfaces in the scene. However, the narrow bandwidth of the spectral bands means that much of the energy in the electromagnetic spectrum is filtered out, which, if not compensated for, can make the images dark or even blank. One common way to compensate for this problem is to increase the available energy by increasing the size of each pixel in the image (collect electromagnetic energy from a larger area). Until recently, the use of this technique has led to poor spatial resolution in remotely sensed images of the Earth. Commonly, pixels were approximately 30×30 meters squared (most houses would fit in less than 4 pixels of this size). However, advances in technology and the proliferation of mounting hyperspectral cameras on a greater diversity of vehicles have made hyperspectral remotely sensed Earth images with greater spatial resolution possible (1×1 meter squared or less is now common). My research focuses on developing ways to use the increased spatial information (available due to decreased pixel sizes) together with the information contained in the spectrum to automatically detect features and objects in hyperspectral images. My research has practical applications in national defense, disaster relief resource allocation, environmental monitoring, agriculture, land cover estimation, land use planning, and artificial intelligence.
Example of how my research is integrated into my GK-12 experience
Mississippi School for Mathematics and Science, the high school that I have been assigned to, is not a typical high school. It is a public boarding school where the state of Mississippi has concentrated students who possess enough aptitude to pass the entrance exam. Thus, some of the students are very advanced for high school. My research, combined with the advanced knowledge of the students, has allowed me to bring in hyperspectral equipment such as a handheld Analytical Spectral Device (ASD), which I used to make graphs of the electromagnetic spectrum when I talked about the Balmer series and the electromagnetic spectrum with a modern physics class. Although I do not get to use equipment such as the ASD with my physics students every day, I very often use math that I have learned through my college and research to understand complex physics models, which I often simplify and explain to my students. I also spend a great deal of time teaching the students how to design experiments and how to interpret the data they get from the results. I believe that skills in designing experiments and interpreting data are perhaps the most useful skills I can teach the students since those skills are common in most sciences and can be useful life skills to have.