Hyperspectral Data Cube
A hyperspectral image is made up of pixels—each of which is an entire spectrum of the material imaged at that point. The imagery is comprised of many contiguous bands of up to hundreds of frequency signals, which may be visible or infrared depending on the sensor.
The datasets are information-rich but somewhat challenging to interpret. For satellite-borne sensors, an important issue has been poor spatial resolution. If each pixel covers 20 or more square meters on the ground, it is hard to derive useful information for many of the potential applications. Airborne sensors with resolutions of about one square meter are becoming more widely used, and appreciation of the value of the datasets is growing in many fields, including defense and intelligence, mining, agriculture and the environment.
Future satellite systems with higher data transmission rates will provide better spatial resolution and enable applications to many fields, including remote mineral exploration and the assessment of crop yields and other civil and military issues. The data can be used to identify the chemical composition of rocks, vegetation type, soil or water pollution, and other attributes that can be characterized in terms of spectral reflectance. The imagery can be applied to resource management, agriculture, mineral exploration and environmental monitoring.
In typical color imagery, three broad frequency bands (usually, red, green and blue) are used to form each pixel. The signal at a given point on the image is comprised of three numbers—intensities in the red, green and blue bands. Typical multispectral satellite sensors will have three to eight or so bands, each one fairly broad and usually with gaps in the frequencies covered. Well-designed sensors choose bands where information is expected to be maximized.
Typical bandwidths in hyperspectral sensors are a few nanometers, and there are no gaps between the individual frequency bands. A typical hyperspectral sensor simultaneously acquires hundreds of optical images, each from a different frequency, thus enabling a “spectral assessment” from distances high in the air via airplanes and in orbit using satellites.
The resulting data can be thought of as a “data cube” or “image cube” with two dimensions in space and one in frequency space.
(Top) Color imagery provides just three reflectance. (Bottom) Hyperspectral pixel values at each pixel.
Why is the imagery useful?
At each pixel, or point, in the picture, a complete spectrum of the substance being imaged is obtained—much like a spectrogram. Hyperspectral datasets are characterized by high spectral resolution. In other words, small differences in frequency (or wavelength) can be readily discerned. The data can be used to detect differences in the composition of objects. If there are images of a scene that appear to contain military tanks, for example, hyperspectral imaging can reveal if they are in fact tanks or instead wooden mockups.
The data are useful for measuring water pollution and monitoring progress in cleanups of oil spills and other environmental accidents. Hyperspectral instruments are being used to detect and monitor gas pipe leakages. The data are also used in agriculture to estimate the moisture content of the soil, a technique which can be used to increase the efficiency of irrigation and conserve water.
There is a difficulty—a spaceborne (or airborne) imager will generally not collect so-called pure pixels with each pixel imaging a single substance. Having the ability to “unmix” such pixels and determine which substances are part of the picture is critical in the analysis of hyperspectral imagery. The stored spectra of substances that are likely to be in the image are used in unmixing algorithms. Another issue is the effect of the atmosphere through which the signal passes, and much effort has gone into the development of atmospheric correction packages.
Development of this technology has been slow, especially in satellite-borne instruments. But that is changing now: More and more airborne instruments are available, so that the utility of the data is becoming more widely known.
Current space-based hyperspectral instruments
Hyperion, on NASA’s EO-1 satellite, launched in December 2000, covers the range from 0.4 to 2.5 µm in 220 spectral bands. The spatial resolution is 30 m with a typical image covering 7.5 x 100 km. The EO-1 orbit is about 50 km (30 miles) behind Landsat 7, which allows similar images acquired at almost the same time to be compared for performance evaluation. The imager, which was originally planned for a much shorter lifetime, is still in use. The fine spectral resolution enables imaging and classification of complex land ecosystems. These classifications enable more accurate remote mineral exploration and better predictions of crop yield and assessments.
The CHRIS (Compact High Resolution Imaging Spectrometer) hyperspectral sensor operates on Proba (Project for Onboard Autonomy), an experimental European Satellite launched in October 2001. Though intended for a one-year mission, it is still in use. Developed by the U.K.-based Sira Electro-Optics Ltd, CHRIS is one of the main payloads on the 100-kg spacecraft. The system has 200 narrow bands between 415 and 1,050 nm and a spatial resolution of 30 m (or as good as 20 m at nadir).
However, at full resolution, only 19 bands are available simultaneously. CHRIS, both by itself or in combination with other satellite sensors, is used to study land vegetation and forests. Important biophysical and biochemical properties can be gathered, including chlorophyll and water content, leaf area index and overall biomass and health. CHRIS images of agricultural crops were being used to estimate crop type and state.
Hyperspectral technology offers enormous potential for defense, commercial and societal application areas. To date, difficulties in processing and interpretation have limited the apparent usefulness of the data. Numerous operational airborne instruments offer solutions to many customers. This market is growing; in the future, the airborne will inform the spaceborne. Advances in communications technologies and data-processing techniques will make it practical to fly satellites with instruments comparable, in spatial resolution, to those now flown on planes. Many more applications will then be addressable from space.
The PRISMA (PRecursore IperSpettrale della Missione Applicativa) satellite is scheduled for launch by 2012. PRISMA is described by the Italian Space Agency as a hyperspectral precursor. This is a medium-resolution hyperspectral imaging mission under development as of 2008. PRISMA is a technology-demonstrator mission focused on the development and delivery of hyperspectral products and the qualification of the hyperspectral payload in space.
The mission objectives include the monitoring of natural resources and atmospheric characteristics. The goal is to provide standard data products with short delays to address applications, such as those related to quality and protection of the environment, sustainable development, climate change, etc.
The instrumentation combines a hyperspectral sensor with a panchromatic, medium-resolution camera. The advantages of this combination are that, in addition to the classical capability of observation based on the recognition of the geometrical characteristics of the scene, there is the one offered by hyperspectral sensors which can determine the chemical-physical composition of objects present on the scene. The combined payload specs are
Spatial resolution: 20-30 m (Hyp) / 2.5-5m (PAN)
Swath width: 30-60 km;
Spectral range: 0.4-2.5 mm (Hyp) / 0.4-0.7 m (PAN)
Continuous coverage of spectral ranges with 10 nm bands.
This offers the scientific community and users many applications in the field of environmental monitoring, resource management, crop classification, pollution control, etc. National security applications are also possible.
The Japanese have purchased an advanced hyperpectral instrument from Headwall, a U.S. photonics company, for use on a future satellite, possibly ALOS-3. Although not specified, it is expected that the spatial resolution will be much better than current hyperspectral satellites.
Less certain is the Shalom project, a joint Italian-Israeli project to fly a high performance hyperspectral instrument on a satellite. The instrument will, according to the president of the Italian Space Agency, improve upon capabilities developed for Prisma. The joint effort, he said, is expected to improve resolution to “1 meter or below” while multiplying the rate of data transmission to about 300 megabytes per second. A total of 200 to 250 spectral bands will be collected.
In the United States, there is interest within NASA in a future hyperspectral satellite. This system would offer user-selectable hyperspectral imaging in the range 380 to 1700 nm but the spatial resolution would be about 15 m. Prospective applications include climate change and disaster monitoring, geological surveys, vegetation, mineral mapping, crop monitoring and others.
Joan Lurie is an imagery systems consultant who specializes in remote sensing and technology transfer.