Hyperspectral signatures collected in situ are important in investigating relationships between the physical and biochemical properties of natural and manmade objects and the electromagnetic radiation they emit. These spectral data are also used to calibrate, validate, and simulate remote-sensing imagery and its information products. Signature gathering can take place under both field and laboratory conditions. Either invariably produces innumerable files and, depending on the instrument, ‘metadata’ detailing instrument settings. Long-term usability requires even more metadata that describe the circumstances of capture, such as illumination and sensor geometries, environmental factors, features about the object observed, and the spatial location of the sampling site. Not least, this supplementary material supports sharing between research groups, as it allows search and selection of spectral information and assessment of its suitability for a specific task.
Having an organized, accessible, and nonredundant means of storing spectral signatures and their associated metadata would be an important step toward better data quality, long-term usability, and information exchange between researchers. To this end, our group in the Remote Sensing Laboratories at the University of Zurich, Switzerland, has implemented an approach called SPECCHIO. We have taken special care to ensure that data input is highly automated, as overly complicated procedures have proven to deter users from entering their spectral collections.
The core of the SPECCHIO system is a MySQL1 database that is hosted on a database server (see Figure 1). The SPECCHIO application was implemented as a Java22 application that permits full access to the local file system without the kinds of security restrictions imposed by other Java technologies. Basing the software on Java guarantees operating system independence, which is essential in a heterogeneous computing environment. The application thus runs on any machine with a Java Virtual Machine (VM) installation and connects to the database via TCP/IP on a configurable port. That makes it possible to connect to the SPECCHIO database from anywhere via the Internet, enabling research groups worldwide to send data back and forth. Database user creation is automated via dynamic web pages written in PHP (hypertext preprocessor).3
Figure 1. SPECCHIO system architecture.
Two database schemas (collections) are currently available to the public on the database server: specchio and specchio_test.4 The former holds data intended for research and sharing. The latter is intended for users who wish to work through the tutorial provided or run tests of their own. The multiuser concept enables the creation of `campaigns’ (activities that result in data sets) that can only be edited or deleted by the creator but are available for browsing and downloading by all registered users. Research institutes that desire to use the tool but require total control over their data may choose to install a private version of the SPECCHIO database on an in-house database server.
The current version of SPECCHIO supports the following spectral signature files as data input formats: ASD binary,5 GER signature,6 ENVI Spectral Library,7 ASCII tab separated, and MFR7 OUT.8 Two output formats are implemented: comma separated value (CSV) files that can be read by statistical and spreadsheet applications, and ENVI Spectral Library files that are primarily a data format used by ENVI7 but can be read by other remote-sensing packages as well.
We hope that the remote-sensing community will make use of the SPECCHIO system as an important step toward long-term data usability, information sharing between scientists, and future development of information management for spectroscopy data. Further details about the system are available at the SPECCHIO web site, including a user guide and published articles.9,10 A few of the many challenges for the future include tracking so-called white reference panel performance against national standards, using spectroradiometer calibration data to provide error estimates of signatures, and defining generic data-exchange formats.
Andreas Hüni, Mathias Kneubühler
Remote Sensing Laboratories
University of Zurich
Andreas Hüni is currently a research assistant in the Remote Sensing Laboratories at the University of Zurich. He obtained an MPhil(Sc) in earth science from Massey University, New Zealand, before commencing his current PhD studies.
Mathias Kneubühler is a senior scientist at the Remote Sensing Laboratories and group leader of the SpectroLab group. He obtained his PhD from the University of Zurich.