Advanced Hyperspectral Exploitation using 3D Spatial Information

Principal Investigator(s)

Emmett Ientilucci 

Research Team Members

Michael Gartley 

Project Description

There is a need to advance hyperspectral exploitation algorithms by incorporating 3D spatial information for improved target detection and identification.  Hyperspectral imaging (HSI) has demonstrated utility for material classification and target detection.  Using HSI alone can produce false alarms and possible low confidence in detects for certain target classes.  Additional 3D information can be used to help improve the separability of material and target classes, which can, ultimately, reduce the number of false alarms.   The overall focus of the effort is to improve and expand upon previous work with emphasis on target detection, rather than material classification. 

RIT is working with SSI to choose target shapes and spectra of interest to the sponsor to be placed throughout DIRSIG Megascene1. Many instances of the target of interest are placed within the scene in locations ranging from wide open to the sun with minimal adjacent surfaces to hidden within a building courtyard and within the shadow of a treeline. Emphasis is placed on radiometric accuracy of the DIRSIG simulations and include radiometric effects of adjacent surfaces, partial transmission through tree leaves, and a MODTRAN derived skydome. Additionally, confuser targets are also being placed around the scene with similar spectra and 3D shape in order to test the robustness of detection algorithms. Simulations will also be run for a variety of sun angles. Pixel geo-location truth (East-North-Up x,y,z coordinates) for each pixel will be provided to the sponsor for use in the detection algorithm.

Figures and Images

Figure 1

Example DIRSIG RGB rendering (left) and CAD view (right) of Megascene1 with T-72 tank targets placed in the open, along a tree-line, within an urban canyon, and next to a building side.