PFI-TT: Interactive Software for Hyperspectral Image Analysis

Washington University, Saint Louis MO

Investigators

Abstract

The broader impact/commercial potential of this PFI project is to advance, facilitate and streamline the analysis of hyperspectral imaging data. In contrast to two-dimensional images produced by conventional color cameras usually with three-color channels, hyperspectral techniques generate three-dimensional datasets known as datacubes with both spatial and spectral dimensions with hundreds of color channels. Hyperspectral methods lead to a vastly improved ability to classify and differentiate objects based on their spectral features enabling small, otherwise unnoticeable, features to be amplified. The societal impact of hyperspectral imaging is rapidly expanding and includes the development of advanced medical diagnostic tools, systems for precision agriculture, methods to ensure food quality, discovery of new minerals, and the evolution of national defense projects. However, despite the progress in analysis method development and computational speed, efficient computational approaches suitable for rapid analysis of datacubes are still lacking. Three-dimensionality of data and large file size presents a significant challenge for isolation of useful data, and consequently individual investigators developing their own tools. The development of an advanced and commercially available software package for hyperspectral data analysis will facilitate this work and enable users to search for new optical signatures across multiple applications. The proposed project sets a new software architecture that is capable of analyzing complex and computationally challenging datasets recorded by hyperspectral imaging systems. The proposed computational platform for data analysis of hyperspectral data will provide a unique and powerful tool to process hyperspectral data for a variety of applications: from medicine and remote vision to forensics and plants science. The research objective is to integrate current and new analysis methods into an interactive, fast, and intuitive hyperspectral imaging software that will offer an unprecedented level of flexibility in hyperspectral data processing. The proposed software program will be designed to be operated without any prior programming skills allowing interactive sessions of raw and processed data to be easily shared, explored and evaluated by both peers and newcomers. The major feature of the software is an open architecture that provides a common framework for the implementation of the additional methods from the user community. The commercially available platform will have structural flexibility to allow users to integrate their novel computational methods with relatively little effort. With this feature, the software is expected to be a central part of hyperspectral technology and will be the preferred software of the majority of users. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.