A free online reference library for hyperspectral reflectance signatures

A Web-based interface for reference spectra of natural surfaces facilitates data sharing and increases access to reference material for classification.
13 December 2006
Jelle G Ferwerda, Simon D. Jones and Marcus Reston

Field spectroscopy is a nondestructive way to collect information on natural resources and is becoming a common exercise in remote sensing field campaigns. These signatures are collected for various tasks. In some cases the signatures are used as a reference signature in spectral matching routines. In other cases these spectra themselves are the aim of the study, being used to predict the presence of minerals in the soil or the chemical composition of foliage. Invariably, field campaigns result in a large number of signature files. This data is useful as reference material for future study of similar environments. However, methods for sharing the information are scarce.

There is a need for a unified repository of spectral signatures, and, just as important, their metadata. Using the grid-computing paradigm, our aim is to connect hyperspectral experts to facilitate enhanced research. To this end we propose an information repository that exceeds that of an individual researcher or team. To facilitate the sharing of data with collaborative research peers, we designed a database for storing and distributing hyperspectral signatures and their metadata. Use of the database is free to all Web users. However, for any operation other than browsing the database, users must register. The database can be accessed through http://www.hyperspectral.info.

Web-based services consist of two separate software components, linked through the Internet. The client side of the software (the Web pages and scripts included in them) runs on the user's computer. The second half, the server side component of the software (all the routines required to process the client-side requests), runs on the Web server. The database and all database-querying routines for the Web service presented in this paper are processed server side. The database uses a commonly found combination of freeware software packages: the Apache Web server, MySQL database engine, and Hypertext Preprocessor (PHP).

The database currently supports import of all GER formats (ASCII and SIG formatted files), ENVI spectral library files, ASD binary files, and ASD ASCII export files. That metadata is an important feature of any dataset is beyond dispute. However, the matter of which metadata is required is not. What is considered important information in one application field may be completely irrelevant for other fields. Therefore, specific attention was paid to the metadata storage format. Instead of storing metadata in columns dedicated to one metadata type, this database stores metadata in one table, with a metadata identifier column. This column is linked to a metadata lookup table, in which the label and the units for the specific metadata are defined. This makes the database very flexible with regard to storable metadata, with a basic set of metadata collected for each sample, while maintaining the option to add additional metadata.

Data can be retrieved through a browse routine on the website. After selecting the files of interest, the user can download them in one of several formats (currently ASCII flat file, JCAMP, and ENVI spectral library formats are available). If the user does not have access rights to the data, he or she can decide to request access, after which the data owner will receive a request for access from the website. At no time during this process will the data sharer's e-mail address be displayed to the user.

A recent addition to the database is the option to browse the database through a Google Earth interface. To do this, start Google Earth. Click ‘Add’ and select ‘Network Link’. Enter http://database.hyperspectral.info/map/specmap.kml as the network link to add the database records to ‘My Places’ in the Google Earth navigation menu. This allows you to browse the globe, while viewing reflectance signatures (see Figure 1). A link (‘More information’) connects directly to the record for the selected signature in the database.

Figure 1. The Google Earth-enabled browse interface allows access to the hyperspectral reference library.

This database was constructed to facilitate data sharing. However, the database itself is only a piece of software. To render it useful, it needs to be filled with information. We are negotiating with a number of research institutes regarding access to their data. Ideally, several different research groups will include their data, and we invite everybody to register and share their data. A forum has been placed online to discuss features and future additions. We hope that the tool as it currently stands will serve as a good starting point from which to further develop the toolkit. We are interested in contributions and ideas by people who are not in the working group to make this a general tool for the spectroscopy community.

This project was supported by the Victorian Partnership for Advanced Computing, which provided the Web hosting. The Spatial Sciences Institute fully endorsed the project, and publicizes the project through its user lists.

Jelle G. Ferwerda
School of Mathematical and Geospatial Sciences, Royal Melbourne Institute of Technology
Melbourne, Victoria, Australia
Department of Zoology, University of Oxford
Oxford, United Kingdom

Jelle G. Ferwerda is currently a research associate at the Department of Zoology of the University of Oxford working on the integration of spatial ecology and remote sensing. He obtained a PhD from ITC and Wageningen University, both in The Netherlands, for his work on mapping the chemical composition of plants using hyperspectral remote sensing. The database was developed during a postdoctoral position at the Royal Melbourne Institute of Technology.

Simon D. Jones, Marcus Reston
School of Mathematical and Geospatial Sciences, Royal Melbourne Institute of Technology
Melbourne, Victoria, Australia

Recent News