Technical Summary: 

Precision agriculture is the use of detailed information within agricultural fields to adjust inputs on a spatially variable basis rather than to apply uniform applications across the entire field.  It will allow producers to place the proper amount of inputs such as pesticides, herbicides, fertilizers, etc., on each location in the field.  In general, to apply precision agriculture technology requires various kinds of information and variable-rate technology (VRT) for application.  The proposed system is an integration of information acquisition, information processing, and VRT.  The use of this system will optimize farm profit while minimizing environmental impact, because fewer chemical inputs are required overall. 

The proposed project is to develop a ground-based sensing system for diagnosing plant health and applying inputs based on plant needs.  Whereas remote-sensing data typically operate at spatial resolutions of 1.0 m or greater, a ground-based system can operate at “ultra-high resolutions” in the range of 1.0 cm.  The system will be comprised of optical sensors, ultrasonic sensors, a measurement unit, a diagnosing unit, and an actuator unit.  The system will be installed on a mobile device.  As the mobile device goes across the field, the ground-based sensing system scans the plant canopy, measures plant height, diagnoses plant health, determines plant needs, and applies inputs accordingly.  

The optical sensors will be designed and fabricated for measuring spectral reflectance of the plant canopy at various wavebands of interest.  Previous studies have shown that spectral reflectance characteristics of plant canopies have strong correlations with the plant health conditions such as nutrient status, diseases, water stress, etc.  Many factors are expected to exhibit their own spectral ‘signatures,’ which can be detected by the optical sensors.  Ultrasonic sensors will be employed to measure plant height.  Plant height is also a very good indicator of plant growth conditions.  The measurement unit processes the signals from optical and ultrasonic sensors and provides information to the diagnosing unit.  Based on the information obtained from the sensors and preset models, the diagnosing unit determines plant needs and triggers the actuator to provide appropriate inputs to the plant.

Objectives: 

The overall goal of this study is to develop a ground-based sensing system for diagnosing plant health conditions and applying treatments to the plant in situ in real-time. 

Specific objectives for year 1 are:

1. To develop optical sensors for measuring reflectance of plant canopy.
    
2. To develop electronic circuits to process signals from optical sensors and ultrasonic sensors.
    
3. To use fuzzy logic theory to develop a model for diagnosing nitrogen status in cotton plant.

Specific objectives for year 2 are:

1. To develop a control system for applying treatments.
    
2. To field-test the whole system.
    
3. To use neural networks and fuzzy logic to develop models for diagnosing P and K status in cotton plant.

Specific objectives for year 3 are:

1. To evaluate and refine the system.
    
2. To evaluate the diagnosis models.
    
3. To build another two prototypes of the system for test by users if the experimental results are promising and have potential for the system to be commercialized.

Procedures: 

1.  Optical sensor development for measuring reflectance of the plant canopy

The optical components of the proposed sensor consist of an optical window, focusing lens, optical filters, and a photodiode.  Light reflected by the cotton canopy travels through the optical window where it is focused by the lens onto the optical filters.  Each filter allows light within only a selected waveband to pass.  Measurements of reflectance at certain wavebands can be achieved by using different kinds of filters.  Four different wavebands may be selected for the optical sensor, they are blue band (350-500 nm), green band (520-580 nm), red band (600-740 nm), and NIR band (750-1100 nm) 

A silicon photodiode can be used for light detection.  Its spectral response range is from 320 to 1100 nm.  The lenses will be used to focus the light onto the photodiode.

    The electronic components of the sensor include a photodiode, an electronic filter, and an amplification circuit.  Each spectral band of the sensor will have a single analog output. 

2. Ultrasonic sensor implementation for plant height measurement
 

    An ultrasonic sensor will be used for plant height measurement.  The ultrasonic sensor will be driven by an ultrasonic ranging module.  As the module is triggered by an initiation signal provided by a computer, the module generates a set of pulses and sends them to an ultrasonic sensor.  Through that sensor a set of ultrasonic pulses are transmitted toward the plant canopy at a speed of about 350 meters per second.  When the first ultrasonic pulse is echoed back to the sensor, the module detects the returning echo and sends an echo signal to the computer.  The difference in time between initial and echo signal is a measure of the distance from the ultrasonic sensor to the plant canopy.  Plant height can be calculated by subtracting the distance measured by ultrasonic from the distance between the ground and the sensor.

3. Measurement unit development

    The measurement unit will be a based on a SBC (single-board-computer) electronic system. The system will include a multi-channel analog to digital converter for interfacing output of the sensors to the computer, amplification circuits, an interface for actuator control, and power supplies.

4. Diagnosis model development

It has been found that spectral reflectance characteristics of plant canopy have strong correlations with plant health conditions.  A model to predict plant needs can be based on spectral data and plant height measurements.  Fuzzy logic and artificial neural networks (ANN) are powerful tools for system control and for analyzing nonlinear systems with classification and pattern recognition techniques.  Spectral reflectance at various wavebands and plant height can be used to construct a fuzzy set or to train an ANN for diagnosing plant health conditions and determining plant needs.  Spectral signatures differ among plant types, varieties, growth stage, weather conditions, etc.  Thus, various models must be developed to meet different conditions.

5. System installation and operation

The system will be installed on a mobile device, such as a high-clearance spray tractor.  The optical and ultrasonic sensors can be mounted on a frame placed at the front of the tractor.  Sensors will face down toward the plant canopy.  As the tractor is driven across the field, the sensors collect reflectance spectra and measure plant height.  The SBC-based electronic system diagnoses plant health and determines plant needs and controls the actuator to apply the inputs accordingly.  For instance, if the system diagnoses nitrogen deficiency in a plant based on its spectral signature and height, it will control the actuator to apply a certain amount of nitrogen at that location.

PIs will pursue patenting of significant technological innovations. 

Justification: 

Precision agriculture strives to optimize profit and minimize environmental impact by adjusting production inputs according to needs of individual areas within a field, rather than applying uniform applications across the entire field.  Two types of data, historical and real-time, can be used in precision agriculture.  With the use of a GPS receiver, precisely determined inputs can be applied at precise field locations based on pre-determined site-specific historical data within the field.  However, it would be more desirable if inputs could be varied by with a real-time, in-situ, ground-based sensing and control system.  Such a system can have extremely high spatial resolution and can be very responsive to the immediate needs of plants.  The ultimate deliverable of this project is a real-time, in situ, ground-based sensing and control system for agricultural inputs.  Such a system will bring about a step change in the accuracy and usefulness of variable-rate application. 

Taking nitrogen (N) as an example, a proper N fertilizer application usually must be provided to maximize yields.  Both under-fertilization and over-fertilization with N can adversely affect yield and/or profit.  Additionally, over-fertilization of N can have a negative environmental impact.  Although soil and plant tissue testing can be used to predict N requirements of plants, soil and plant N analyses often require considerable time and expense.  The proposed ground-based sensing system would be able to diagnose the N status of plants and apply proper amounts of N according to plant needs.  

Theoretically, the proposed system will have the potential to deal with various nutrients including N, P, and K, and various kinds of plant diseases, because all those factors have effects on plant reflectance and height.  At first, N will be the nutrient considered for cotton.  Then, other cotton nutrient such as P and K will be considered.  Further, the potential of the proposed system for other crops and for plant diseases may be explored when the immediate goals are reached. 

Literature Review: 

Leaf color is usually the most sensitive indicator of deficient nutrient levels (Blinn et al., 1988).  The main pigment responsible for leaf color is chlorophyll.  Because N is a major component of the chlorophyll molecule (Tracy et. al., 1992), chlorophyll and leaf N are usually correlated.  Instruments commonly used in the field to evaluate leaf color are spectrometers, radiometers, and imaging systems.  These instruments measure the reflectance spectra of plant leaves non-intrusively.  By analyzing the reflectance spectra, relationships between the characteristics of the reflectance spectra and N concentration of the plant leaves can be obtained. 

Ma et al. (1996) used a radiometer to measure canopy reflectance of six maize hybrids and found that light reflectance measurements prior to anthesis may provide an in-season indication of N deficiency.  Sembiring et al. (1998) detected N and phosphorus nutrient status in winter wheat using a fiber optic spectrometer.  Results demonstrated that normalized difference vegetation index (NDVI) was a good index to predict N and phosphorus uptake for winter wheat.  Furuya (1987) used a Standard Rice Leaf Color Scale to conduct a series of studies on rice growth diagnosis by leaf color and found that the rice leaf color value was significantly correlated with average N concentration in all the living leaves. Wood et al. (1992) found that the chlorophyll meters were as reliable as leaf N concentration for predicting supplemental N fertilization requirements of cotton. 

Stone et al. (1998) developed optical sensors for detection of N availability in winter wheat. They found a good correlation between vegetative N mass per unit area and the spectral index, and that the reflectance rises in the NIR band and falls in the red band for increasing N availability.  Beck and Vyse (1995) patented an apparatus and a method that can be used to selectively eliminate weeds in agriculture operations.  The spectral reflectance characteristics of plants and soil were used to differentiate between plant and soil. 

Sui et al. (1998) developed a non-intrusive optical sensing system for measuring nitrogen status in cotton.  It was found in that study that nitrogen status of cotton plants was well correlated with spectral reflectance characteristics of the cotton canopy at various growth stages.  A multi-band optical sensor was designed and fabricated to measure reflectance from a cotton canopy.  An artificial neural network was implemented to classify the nitrogen status of a cotton plant based on spectral data from the sensor.  Preliminary test results showed that the in-situ sensing system could accurately diagnose nitrogen status in cotton. 

Jennifer and Varco (2000) evaluated the effects of varying N and k nutrition on relative leaf reflectance, red edge shift, and NDVI in relation to detecting N and K deficiencies. It was found that the greatest separation between N treatments occurred at the 550 nm waveband given an adequate supply of K. the red edge shift significantly moved towards longer wavelengths when applied N increased and K was sufficient. 

Tumbo et al. (2000) studied a hyperspectral-input based neural network model for predicting chlorophyll status in corn.  Spectral reflectance response patterns (SRRPs) from individual corn plants were collected under variable cloud cover and solar angles with a fiber optic spectrometer.  Chlorophyll levels were also measured on each corn plant with a SPAD meter. The back-propagation neural-network model was trained with spectral channels of SRRPs as inputs and chlorophyll readings as outputs.  The model showed strong correlation (r²=0.91) between predicted and actual chlorophyll meter readings.  

Haberland et al. (2001) developed a ground-based sensing system to provide real-time management information for agriculture and to serve as a research tool for remote sensing research.  The sensing unit collected reflectance data in the visible, near-infrared, and thermal bands.  The resulting field maps included the reflectances at 5 wavelengths, derived vegetation index maps, or nitrogen and water status maps.  The ground-based sensor data provided higher temporal and spatial resolution than satellite images.  The system was used to develop nitrogen and water status indices in cotton and broccoli in Arizona.

Current Research: 

While this is a new research program, the team has been conducting research on developing a mobile sensor platform for soil texture and surface roughness measurements.   

References: 

Beck, J. and T. Vyse.  1995. Structure and method usable for differentiating a plant from soil in a field.  United States Patent No. 5389781. 

Blinn, Charles R., A. Lyons, and E. R. Buckner. 1988. Color aerial photography for assessing the need for fertilizers in loblolly pine plantations.  South J. Appl. For. 12(4):270-273. 

Furuya, Shoji.  1987. Growth diagnosis of rice plants by means of leaf color. JARQ. 20(3):147-153. 

Haberland, J. A., P. D. Colaizzi, M. A. Kostrzewski, P. M. Waller, C. Y. Choi, F. E. Eaton, E. M. Barnes, and T. R. Clarke. 2002. AgIIS, Agricultural irrigation imaging system. Applied Engineering in Agriculture.(in press) 

Lough, L. J., J. J. Varco. 2000. Effects of varying N and k nutrition on the spectral reflectance properties of cotton. In proc. The fifth International Conference on Precision Agriculture. ASA_CSSA_SSSA, 677 South Segoe Road, Madison, WI 53711. 

Ma, B. L., M. J. Morrison, and L. M. Dwyer. 1996. Canopy light reflectance and field greenness to assess nitrogen fertilization and yield maize. Agron. Journal. American Society of Agronomy. 88(6):915-920. 

Sembiring, H., W. R. Raun, G. V. Johnson, M. L. Stone, J. B. Solie, and S. B. Phillips. 1998. Detection of nitrogen and phosphorus nutrient status in winter wheat using spectral radiance. Journal of Plant Nutrition. Monticello, N.Y.: Marcel Dekker Inc. 21(6):1207-1232. 

Stone, M. L., J. B. Solie, R. W. Whitney, W. R. Raun, and H. L. Lees. 1998. Sensor for detection of nitrogen in winter wheat. http://bioen.okstate.edu/Home/mstone/N-Sens.htm. 

Sui, R., J. B. Wilkerson, W. E. Hart, and D. D. Howard. 1998. Integration of neural network with a spectral reflectance sensor to detect nitrogen deficiency in cotton. ASAE Paper No. 98-3104.  St. Joseph, Mich.: ASAE. 

Tracy, P.W., S. G. Hefner, C. W. Wood, and K. L. Edmisten. 1992. Theory behind the use of instantaneous leaf chlorophyll measurement for determining mid-season cotton nitrogen recommendations. In proc. 1992 Beltwide Cotton Conferences. 1099-1100. 

Tumbo, S. D., D. G. Wagner, and P. H. Heinemann. 2000. Hyperspectral input-based neural network model for predicting chlorophyll status in corn. ASAE Paper No. 00-1009.  St. Joseph, Mich.: ASAE. 

Wood, C. W., P. W. Tracy, D. W. Reeves, and K. L. Edmisten. 1992. Determination of cotton nitrogen status with a hand-held chlorophyll meter. Journal of Plant Nutrition. 15(9):1435-1448. 

Back to Top

Back to Menu