It has been said that as-tronomy is the surest proofof human mind because

we can never expect to bene-fit materially from it. On thecontrary, the spectrophoto-metric devices that reckon theatmospheric chemistry of dis-tant planets and stars arefinding applications here onEarth. Sophisticated airborneand space-based imagersfrom NASA, such as Aviris,Hyperion and Advanced LandImager, are enabling preci-sion agriculture.

“Precision agriculture isfarming by the foot and notby the field,” said RodneyMcKellip, Ag20/20 managerin NASA’s Earth ScienceApplications Directorate atStennis Space Center in Mis-sissippi. Using hyperspectral,multispectral and infraredimaging, he explained, farm-ers can identify areas that areexperiencing water, nitrogen or mi-cronutrient stress, and can targettheir relief.

Different techniquesThe terms “multispectral imaging”

and “hyperspectral imaging” some-times are used interchangeably, butthe techniques are quite different. Amultispectral imager produces animage for each of a number of dis-crete and relatively wide bands, anda hyperspectral imager is more likea laboratory spectrometer. It pro-duces images over a contiguousrange of narrow spectral bands andenables the spectroscopic analysisof data.

For example, the SAMRSS airborne

multispectral imager from Opto-Knowledge Systems Inc. of Torrance,Calif., covers the 550-, 660- and 850-nm, and 7- to 14-µm bands, simu-lating Landsat bands 2, 3 and 4. Thissystem sports three 1k 3 1k-pixelcameras from Dalsa of Waterloo,Ontario, Canada, and two frame grab-bers from BitFlow Inc. of Woburn,Mass.

The trick, said Opto-Knowledge’spresident, Nahum Gat, is to syn-chronize the cameras and the framegrabbers down to ±0.5 of a pixel sothat the fields of view overlap. Withoutsynchronization, one camera couldbe in the middle of exposing a framewhile another is finishing, resultingin three unregistered pictures of the

same target area because of aircraftmotion.

Software corrects any remainingmisregistration due to alignment ofthe optical axes of the camera. A mid-IR camera from Indigo Systems Corp.of Santa Barbara, Calif., also hasbeen registered to the multispectralimages, using software that accountsfor the different pixel size of the ther-mal imager.

Opto-Knowledge’s Avnir hyper-spectral imager measures 60 bandsin 10-nm increments between 430and 1012 nm. It essentially is a high-speed, thermoelectrically cooled,back-illuminated CCD camera withan imaging spectrograph mountedon the front. Apochromatic optics,

by Brent D. Johnson, Senior News Editor

With the lower cost of components — and the increased pressures of globalism — remote spectral imaging is becoming a viable agricultural tool.

Spectral Imaging Finds a Place on the Farm

Remote Sensing

Remote spectral imaging enables farmers to target specific areas that are experiencing stress, potentially boosting yields while lowering costs.

Reprinted from the January 2002 issue of PHOTONICS SPECTRA © Laurin Publishing

corrected for the full visible and near-IR spectral range, eliminate chro-matic aberration.

Gat said that the hyperspectralsystem is critical because it can pro-vide information about plant physi-ology. When coupled with radiativetransfer models, Avnir allows theuser to characterize the spectral sig-nature of plants and to determinewhether they are healthy or understress. The company also has devel-oped a 900- to 1700-nm sensor, butit has not yet been used in precisionagriculture.

SAMRSS and Avnir fly together ona Cessna at 5000 to 9500 feet. Theairborne approach gives them theflexibility to image at specific timeswhen a crop is at key growth stages.“Timing is everything,” McKellip said.The spatial resolution for the air-borne system is 1 meter, and the re-sults of the flight screening are avail-able within 24 to 48 hours, enablingnearly real time decisions aboutgrowing conditions.

Extracting plant signatures from a

signal poses a significant challenge.The sensors receive both reflectedand emitted light from the ground.Signal processing is required to ex-tract the desired signature, includ-ing sensor calibration, atmosphericcompensation and aircraft attitudecorrections.

Sensor calibration converts the rawsignal into engineering radianceunits. Atmospheric compensationaccounts for the solar spectrum,solar slant angle, atmospheric con-stituents, such as CO2, humidity,aerosols and other species that af-fect light passing through the at-mosphere. Aircraft attitude correc-tion registers the imagery with thegeographical coordinate system.Finally, the data are converted intoa form that the farmer or the cropadviser can use to make decisionson crop management.

Many of these steps currently re-quire manual intervention, but thetrend at Opto-Knowledge is towardfull automation of the process. “Froman engineering standpoint,” Gat said,

“there is a tremendous amount of ef-fort that goes into this.”

The hyperspectral images obtainedgive growers a tool that allows themto manage field conditions. The in-formation can be fed into a digitalcontroller and, using the global po-sitioning system for navigation, trac-tors can be guided to problem areas— where they can distribute nutri-ents or pesticides within 3/4 in. ofwhere they are needed.

Precision on Sheely farmThe Sheely farm in California’s San

Joaquin Valley, the largest and mostintensively used agricultural land inthe country, is one of a number of testsites for the Ag20/20 program, sup-ported by NASA, the US Departmentof Agriculture and the National CottonCouncil. Several precision agricultureapplications — geared mainly to cot-ton — are being explored on the farm,including irrigation scheduling, nu-trient application, soil remediation,insect detection, growth control andyield predictions.

Ted Sheely has been usingthe Opto-Knowledge hyper-spectral/multispectral sys-tem for two growing seasonsto farm a seven-acre plot ofcotton, tomatoes, garlic,wheat and pistachios. Multi-spectral, thermal and hyper-spectral systems image thefields weekly during thegrowth season. Flight plan-ning affords the flexibility toreschedule if cloud cover is aconcern on a particular day.

One problem in Californiais an increased level of soilsalinity because of irrigationpractices, eventually causingthe soil to become barren.The application of gypsumcompensates for salinity, pre-venting damage. On Sheely’sfarm, remote sensing detectsspecific areas of high salin-ity in the field, enabling thetargeted distribution of gyp-sum. Without this informa-tion, he would have to applythe mineral uniformly overthe field — depositing too lit-tle where it is most needed

Remote Sensing

The Sheely farm is one of several sites testing precision agriculture techniques as part ofthe Ag20/20 program. Here, Gordon Scriven of Opto-Knowledge Systems Inc. collectshyperspectral data on the ground that will be used to validate airborne measurements of the cotton fields.

and too much where it is not. The technique has also al-

lowed the farmer to cut backon the use of nitrogen fertilizerand growth regulators. For ex-ample, he used to apply 8 ozof plant growth regulator peracre. Based on the remotesensing data, he now uses 1lb in some places and nonein others, producing highercrop yields at a lower cost.

Pest controlResearchers hope that re-

mote sensing also will helpthem to identify the conditionsthat make crops susceptibleto insect infestation. InCalifornia, spider mites turnplant leaves a bronze color.The spectra of the afflictedplants should indicate these“hot spots.”

But experiments have suc-cessfully predicted even thepotential danger of spidermites, identifying where cot-ton is most likely to producean infestation before it does.Dust catching in the mites’webs, for example, causes aminute change in the spec-tral reflectance of a plant.

Early indications show that remotesensing — hyperspectral imaging inparticular — can distinguish the sig-natures of healthy and infestedplants to allow intervention beforethere is significant damage, promis-ing a savings of up to 20 percent inlabor and pesticides.

A Mississippi team supported byNASA uses remote sensing to detectlygus bugs. These pests also feed oncotton, causing damage to the budsand the bulbs of developing flowers.Airborne or satellite imagery canidentify the lush areas in a field thatare preferred by lygus bugs to predictwhere the pests may be. The teamhopes to use the technique to iden-tify aphids, white flies and thrips,which damage seedlings.

Water conservationIrrigation management has been

a major focus of the program. Theentire San Joaquin Valley relies heav-

ily on irrigation. Because one or twocycles of irrigation can cost tens ofthousands of dollars, farmers andagrobusinesses are keen on a differ-ent approach. “If you can time irri-gation to optimize water use, thathas tremendous implications,”McKellip said.

Steve Maas of Texas TechUniversity in Lubbock was involvedwith Opto-Knowledge as a projectleader at the Sheely farm. An aero-space engineer who has found ahome in agricultural and environ-mental studies, he is building a mul-tispectral imager for thermal wave-lengths that will indicate the waterstatus of the plant canopy by mea-suring the temperature of the field.The bands between 1500 and 1900nm are good for detecting water ab-sorption, which would help deter-mine the status of the vegetablecanopy.

Maas said that the price of high-

performance CCDs and research-grade digital cameras has fallen sig-nificantly, making them suitable forthis multispectral work. Faster com-puters with PCI buses can now han-dle the throughput from these de-vices, which is typically in the giga-byte range for a single scan.

Competitive edgeAs improvements in software, au-

tomation and sensor technology con-tinue to make remote sensing moreaccessible, multispectral and hyper-spectral remote imaging will play agreater role in agriculture. This issignificant for a farmer seeking tocompete in today’s global markets,Sheely said.

“I do not have cheap land or cheaplabor, and I have the strictest envi-ronmental regulations in the world.The only edge that the Americanfarmer has against the world econ-omy is technology.” G

Remote Sensing

A hyperspectral image of a field indicates the relative health, or vigor, ofthe plants, based on their normalized difference vegetation index (NDVI).Healthy plants absorb more visible light and reflect more infrared than dounhealthy plants.