HyperPACS™
Application: Hyperspectral data Processing Algorithm Comparison Software

Comparison test bed for development of real-time HSI processing algorithms with user friendly interface. Supports several standard data formats and compares algorithm speed, probability of detection, and false alarm rate.

The application is a test bed to compare the performance of various hyperspectral imagery (HSI) processing algorithms with respect to probability of detection (Pd), false alarm rate (FAR), and efficiency. This tool will significantly enhance algorithm performance analysis by centralizing standard HSI algorithms in one user friendly application and streamlining the performance comparison workflow. Integration of user developed algorithms will indicate how the new algorithm compares to standard HSI algorithms. The graphical display of algorithm output and ROC curves for Pd and FAR allows the user to quickly determine an algorithm’s strengths and weaknesses.

Current hyperspectral algorithms do not provide the capability to perform near real-time target detection and classification. Techniques are needed to process hyperspectral data in near real-time, while ensuring reliable target detection in camouflaged, concealed and disguised battlefield conditions.

A SMDC funded Radiance research program (JWaTH) is working to develop advanced data processing algorithms that will speed data processing, increase target detection and reduce the false alarm rate. Part of the JWaTH focus is to develop benchmarks for current HSI algorithm performance. Currently, no published benchmarks for HSI processing algorithm performance exist. Such benchmarks are needed as a foundation upon which to base claims of algorithm improvements. The purpose of HyperPACS is to provide this performance benchmark for analysis of future algorithms developed toward the goal of real-time processing.

In addition to supporting claims of improvement by a newly developed algorithm, performance comparisons in HyperPACS will quickly illustrate algorithm strengths and weaknesses based on data variability such as terrain, signature data quality, and data resolution. This will aid decision makers when attacking a new HSI problem.

HyperPACS provides a user friendly interface, supports one or more data files and algorithms in a single session, and will automatically generate a ROC curve for visual performance analysis. Graphical displays of the selected data file (user option for which bands to display), truth mask, algorithm output, and performance metrics provide visual confirmation that data input and output are reasonable. HyperPACS includes several standard HSI processing algorithms: Spectral Angle Mapper, RX, Generalized Likelihood Ratio Test, Spectral Matched Filter, and Stochastic Expectation Maximization along with several variations of these.

Algorithm Output and Truth Mask Display

In addition to spectral based target detection and anomaly detection algorithms, HyperPACS supports change detection algorithm performance comparisons. This feature allows the user to quickly compare not only performance between change detection algorithms, but the affect image registration errors may have on algorithm performance.

		More HyperPACS Software Info
		Products Index Download HyperPACS Brochure Contact Radiance

Contact Radiance Technologies to inquire about HyperPACS.

Technologies | Services | Products | Contract Vehicles | Careers | Company | Home | Contact Radiance

©1999-2010 Radiance Technologies, Inc. All Rights Reserved.