Hyperspectral Imaging Offers Huge Potential to Agriculture

Hyperspectral images acquired by aircraft and satellites have the potential to detect crop stress and diagnose its cause long before a farmer can spot the problem in his field with the naked eye.

Farmers have heard the remote sensing oversell before, so let's make one point clear: Hyperspectral imaging will not revolutionize precision farming overnight. But it is a technology that will have a major impact.

Much research still must be completed, and the practical evolution of hyperspectral imaging in agriculture will occur much faster than it did with multispectral imagery. For this reason, some precision farming practitioners are already gearing up to handle the new data, and they will be the first to benefit from it.

Numerous factors are driving the rapid development of hyperspectral imaging applications. For example, NASA's Lewis satellite will acquire 384 bands of data which will be provided at no charge to anyone who wants it. Unrestricted distribution will enable precision farming researchers to explore the data's potential at relatively small expense.

More importantly, many agriculture researchers are already using hyperspectral airborne imagery today. Many systems, such as the Jet Propulsion Laboratory's AVIRIS and ITRES Corp.'s casi sensors, are available for private and public sector use. Increased competition is driving down the cost to purchase or lease these instruments.

New satellites with one-meter spatial resolution in a single band (i.e. panchromatic) will also enhance the utility of hyperspectral and multispectral imagery. Media publicity portrays these high-resolution satellites as the remote sensing tool for precision farming. This is an exaggeration. Spatial resolution offers great value in monitoring crop appearance, but it is spectral data that reveals the most information about plant stress and health.

An important point to keep in mind is that hyperspectral, multispectral, panchromatic, and even radar images are each individual tools that provide a complete picture of crop condition when they are used together. This is why Research Systems, for example, has integrated the ability to process all these remote sensing data types into one software package.

Hyperspectral Benefits

The objective of satellite and airborne remote sensing in precision farming is to provide farmers with detailed crop condition information that can be used to program variable-rate technology equipment. Application of water, pesticides or fertilizer can be tailored to the needs of the crop, based on conditions seen in the imagery.

In multispectral and hyperspectral imagery, plant conditions are assessed by their spectral "signature" -- a measure of electromagnetic energy reflected from the crop. Plant health is a major factor that dictates the amount of energy reflected.

The difference between multispectral and hyperspectral imagery is the detail of the spectral signature. Most multispectral sensors take one measurement in a wide portion of each major wavelength band, such as visible blue, near infrared, etc. Hyperspectral, on the other hand, measures energy in numerous narrow units of each band.

As a result of this higher number of measurements, the hyperspectral signature is more detailed and contains more specific information about crop health.

"The signature from a few wide bands of multispectral imagery can tell a farmer that his crop is stressed," explained Dr. Joe Boardman, chief geophysicist at Analytical Imaging and Geophysics (AIG) in Boulder, Colorado. "But, the many narrow bands in hyperspectral imagery can potentially indicate whether stress in caused by water depletion, insect infestation, or poor fertilization."

The narrower bandwidths of a hyperspectral sensor are more sensitive to subtle variations in energy reflection, which could indicate stress, Boardman added. As a result, hyperspectral imagery can detect crop stress well in advance of multispectral imagery.

NASA's Stennis Space Center in Mississippi is spearheading the Lewis program and has completed some evaluation of hyperspectral applications in farming. Initial results indicate the most significant benefits will be found in high-gross crops such as melons, strawberries, broccoli and citrus fruits.

Hyperspectral Concerns

Buying, processing and storing 384 bands of digital hyperspectral data is an intimidating prospect, but advanced image processing tools will make it easier for practitioners of precision farming to handle.

Just as agriculture professionals now know to purchase three specific bands of Landsat multispectral data instead of all seven, so too will precision farming experts choose solutions-oriented products derived from the hyperspectral data set. Researchers at AIG, NASA and other remote sensing centers are now developing and testing the methods for creating the products that are best for assessing water content in corn, insect infestation in citrus, and a variety of other ailments in specific crops.

Users will purchase prepackaged data sets called "Corn Water Content" or "Equivalent Biomass." The current trend in remote sensing is to provide total solutions instead of data. The developer of the casi airborne scanner, ITRES Research Limited of Clagary, Alberta, plans to begin offering mapping services in addition to selling and leasing its systems.

"Providing answers is the key to these services," said ITRES' Robert Price. "We will supply customers with a bitmap rather than raw data."

Of course, many large agriculture supply companies already perform image processing in-house for their clients and will continue to do so. Advanced algorithms in some image processing packages will assist them in handling large data files. For instance, Research Systems has included Principal Component Analysis and Minimum Noise Fraction routines in ENVI to reduce data file sizes.

What You'll Need

Agriculture supply companies, agronomists and individual farmers will be surprized at how little computer equipment they need to beging processing hyperspectral, multispectral or panchromatic imagery. To run a powerful image processing package like ENVI, all that is required is a Pentium PC or Power Macintosh. We recommend a CD-ROM drive and at least one gigabyte of storage. Access to the World Wide Web is necessary for perusing data availability and eventually downloading images.

As mentioned, hyperspectral imaging techniques are being developed and refined on a daily basis. We recommend precision farming experts take at least one short course on the subject. The University of Colorado and AIG, both in Boulder, offer excellent resources in this area.