design and development of an automated and non–contact sensing ...

DESIGN AND DEVELOPMENT OF AN AUTOMATED AND NON–CONTACT SENSING SYSTEM FOR CONTINUOUS MONITORING OF PLANT HEALTH AND GROWTH M. Kacira, P. P. Ling

ABSTRACT. An automated system was designed and built to continuously monitor plant health and growth in a controlled environment using a distributed system approach for operational control and data collection. The computer–controlled system consisted of a motorized turntable to present the plants to the stationary sensors and reduce microclimate variability among the plants. Major sensing capabilities of the system included machine vision, infrared thermometry, time domain reflectometry, and micro–lysimeters. The system also maintained precise growth–medium moisture levels through a computer–controlled drip irrigation system. The system was capable of collecting required data continuously to monitor and to evaluate the plant health and growth. Keywords. Plant health monitoring, Data acquisition, Image processing, Design, Instrumentation.

A

n automated plant monitoring system should be capable of providing sufficient and accurate information that reflects the true physiological state of the plants under observation. It is important to characterize the interactions between plants and their environment if an automated plant monitoring system is to be established. A well–designed plant monitoring system should be capable of dealing with environmental variability, monitoring both plant microclimate and soil conditions, and providing non–contact sensing. The effect of environmental variability is one of the major concerns in experimental design for plant systems. Microclimates surrounding plants are not usually uniform. Therefore, many samples and sensors are required to obtain a true representation of the plant population. A plant monitoring system capable of reducing the required number of samples by reducing environmental variability would be advantageous. To better understand plant–environment interaction, it is essential to study both the microclimate surrounding the plants and the growth media. For instance, plant–water relationships should be studied by monitoring both supply (water reservoir) and demand (transpiration) conditions. To achieve this, the system must be equipped with proper instrumentation to continually monitor soil moisture and

Article was submitted for review in May 2000; approved for publication by the Information & Electrical Technologies Division of ASAE in May 2001 The authors are Murat Kacira, ASAE Member Engineer, Assistant Professor, Department of Agricultural Machinery, Harran University, Sanliurfa, Turkey; and Peter P. Ling, ASAE Member Engineer, Assistant Professor, Department of Food, Agricultural, and Biological Engineering, Ohio Agricultural Research and Development Center, Ohio State University, Wooster, Ohio. Corresponding author: Murat Kacira, Dept. of Agricultural Machinery, Harran University, Sanliurfa, Turkey; phone: 90–414–2470386; fax: 90–414–2474480; e–mail: [email protected].

Vol. 44(4): 989–996

plant microclimate conditions such as air temperature, humidity, radiation, air velocity, and CO2 level. Non–contact and non–destructive sensing techniques can continuously monitor plants and enable automated sensing and control capabilities (Ling et al., 1996). Traditional sensors used in plant sensing to assess growth and health (such as pressure chambers to measure leaf water potential, sap flow meters to measure sap flow, and dendrometers to measure stem diameter change) are often destructive and require contact measurements. Tensiometers that measure soil water potential require contact with the soil. Such measurements may not be feasible for real–time monitoring and control of plant health and growth. Therefore, it is desirable to develop non–contact and non–destructive sensing techniques to replace conventional methods, such as infrared thermocouples for non–contact temperature measurement. Canopy temperature is an important indicator of plant response to environmental influences such as solar radiation, air temperature, air movement (Pang, 1992), and water availability (Ehrler et al., 1978). In addition, plant leaf temperatures have long been related to plant transpiration (Gates, 1964), leaf water potential (Clark and Hiler, 1973), and soil water deficit (Choudhury, 1983; Choudhury and Idso, 1984). Advances in infrared thermometry have promoted research in monitoring plant health and evaluating plant water stress using remotely sensed plant canopy temperature. Infrared thermometers are frequently used, and the plant’s response to environmental changes can be monitored (Idso et al. 1981; Jackson et al., 1981; Clawson et al., 1989; Stanghellini et al., 1992; De Lorenzi et al., 1993; Kacira, 2000). Machine vision is a non–invasive, non–contact sensing technology that enables multi–dimensional sensing capabilities. Machine vision can be used to extract various information from a targeted object including morphological (size, shape, texture), spectral (color, temperature, moisture), and temporal data (growth rate, development, dynamic change of spectral and morphological states). For example,

Transactions of the ASAE E 2001 American Society of Agricultural Engineers ISSN 0001–2351

989

Oosterhuis et al. (1985) studied soybean leaflet movement as an indicator of crop water stress. Tracking leaf tips for tomato plant wilt detection using machine vision was evaluated by Seginer et al. (1992). Ling et al. (1996) studied the spectral properties of water– and nutrient–stressed plants for monitoring the development of lettuce. The objectives of this research were to design an automated plant monitoring system to study plant and environment interactions closely for a plant factory or greenhouse production system, to demonstrate the system’s capability, and to evaluate its performance. A case study using infrared thermometry and imaging for water stress detection is presented to illustrate the system’s capabilities.

MATERIALS AND METHODS A distributed signal conditioning, sensing, and operational control approach was used for design of the plant monitoring system, as illustrated in figure 1. Operation of the monitoring system is centrally coordinated but locally

executed. A host computer was used to coordinate operation of the plant monitoring system, which includes five subsystems: (1) a stepper motor–driven turntable, (2) a drip irrigation system, (3) six load cell–based micro–lysimeters, (4) an environmental monitoring system, and (5) two plant monitoring systems. Signal conditioning, data acquisition, and data preprocessing were carried out locally using data loggers and centrally using the host computer. Figure 2 shows the complete schematic of the system developed for plant sensing and monitoring. The host computer including plug–in boards such as frame grabber and data acquisition boards, the stepper motor control box, and a data logger were located outside of the environmental chamber. All other sub–systems were located inside the chamber. The plant monitoring system is housed in a walk–in environmental chamber. To assure fully automated, accurate monitoring of plants, design of the system took into account material flow control, signal conditioning, and sensor placement and data collection. A detailed description of the growth chamber can be found in Kacira (2000).

load cell readings

excitation voltage

Load cell

Data logger IR temperature sensing –temperature referencing –signal conditioning –backup data storage – data acquisition and transmission

Data logger with multiplex sensor inputs –temperature referencing –signal conditioning –backup data storage –excitation voltage for load cells –data acquisition and transmission Host Computer Plant evapotranspiration data –image acquisition –data acquisition –irrigation control –turntable motion control

Plant canopy and environmental data

Frame grabber Irrigation pump

Stepper motor controller

Drip emitters

Turntable operations

Image acquisition

Figure 1. Block diagram of the distributed signal conditioning, sensing, and control system.

TURNTABLE DESIGN AND MOTION CONTROL Radiative spectral quality and quantity in a growth chamber can be affected by factors such as degradation of lamp output and barrier discoloration. In addition, airflow may not be uniform along the plant canopy. Thus, a turntable was designed and custom built to reduce plant–to–plant

990

variability in the experiment (fig. 3). The turntable system, consisting of a circular plywood platform and associated drive mechanism, was used to support six plants mounted on six micro–lysimeter stations. The diameter and the thickness of the circular platform were 1.22 m and 2.54 cm, respectively. The circular platform was mounted on a rotary

TRANSACTIONS OF THE ASAE

15 13 12

11 2

14 1. Infrared thermocouple. 2. Monochrome CCD camera 3. RH sensor 4. Hot–wire anemometer 5. Pyranometer 6. Air temperature sensor 7. Soil moisture sensor 8. Pump and nutrient tank 9. Stepper motor 10. Load cell 11. Data logger 12. Data acquisition board 13. Stepper motor control box 14. Frame grabber board 15. Computer 16. Turn–table 17. Drip irrigation emitter 18. Irrigation line

5

1

4

6 7

17

3

11 10 16

9

18

8

Instrument control Data acquisition Irrigation line Figure 2. Diagram of the plant health monitoring system. The actual position of pyranometer, hot wire anemometer, and relative humidity sensors were next to the plant whose temperature was being measured with the infrared sensors. 60O AA

2.54 mm 2.54cm

0.45m

1.22 m

∅ 11.2cm 1.8cm

2.5mm

8.2cm

1.3cm 2.0cm

∅ 0.81mm

WEIGHING PLATFORM

LOAD CELL AA

Figure 3. Structural dimensions of the turntable and load cell with weighing platform (designated AA in the figure).

Vol. 44(4): 989–996

positioning table (RT–12, Arrick Robotics Inc., Hurst, Tex.), which was powered by a stepper motor (MD2–a, Arrick Robotics Inc.) through a pulley assembly (PR–23, Arrick Robotics Inc.). The turntable’s operation was controlled by a computer system. The table presented plants individually to a camera and to other stationary sensors in the system. A stepper motor driving system was used for good indexing accuracy. The accuracy of the rotary table used was specified as 0.1 degree, which gives an approximate error of 1 mm at the circular platform level. A limit switch was installed beneath the edge of the platform to define the reference position of the table. Positioning the table at the reference position before sampling eliminated accumulated positioning error. Electrical cables were run from the host computer to a data logger mounted at the center of the table for data flow between the computer and the data logger. To avoid tangling the electrical cables, rotation of the table was limited to less than 360o before rotation direction was reversed. A slip ring option was considered for continuous unidirectional turntable operation but was ruled out due to the potential for electrical noise and cost concerns. Acceleration of the stepper motor was managed using a ramp function that gradually accelerates the motor from a slow speed to a fast speed and then decelerates when travel is near completion. The torque of the stepper motor decreases as the speed increases, which could cause missteps at faster speeds if higher–than–available torque is required. Using the ramp function for motion control allowed maximum torque to reduce missteps at the startup stage, enough speed to complete the desired motion in a timely fashion, and a “soft landing” at the end of motion to reduce overshoot error.

991

IRRIGATION CONTROL The system was also designed to perform closed–loop automatic irrigation. The irrigation control used feedback from soil moisture sensor readings to trigger watering events. A soil moisture retention curve for the growing medium was determined by a laboratory experiment. This curve provided enough information for selection of a set point to trigger the irrigation event. When the measured soil moisture reading was below a pre–selected set point, an analog voltage signal was sent from the host computer to a relay through the data acquisition board to initiate drip irrigation. The bottoms of the pots were sealed to prevent water spill over the electronics. A detailed description of the soil moisture sensor is provided in a subsequent section of this article. The reader is referred to Kacira (2000) for information on the procedures used to determine the soil moisture release curve and select the irrigation set point, nutrient solutions, and growth medium mixture used in the experiments. MICRO–LYSIMETERS Six micro–lysimeters with load cells (Model 355, Tedea–Huntleigh, Calif.) and custom–built weighing platforms were positioned 60° apart (for dynamically balanced loading of the table and proper spacing for image acquisition) and 0.45 m from the center of the table to hold plant pots (15.2 cm in diameter) and to monitor the weight change of the plants due to evapotranspirational water loss (fig. 3). The load cells had a resolution of ±1 g and a maximum load capacity of 5000 g. The load cells were supported with a Plexiglas material on the turntable platform. A test was conducted to determine the effect of offset loading on load–cell readings due to soil moisture sensor placement on the pots. An offset load of 50 g was applied at nine different locations with five different center loads (2000, 1000, 500, 100, and 50 g) on the weighing platform (fig. 7). A standard set of brass weights was used. INSTRUMENTATION AND SIGNAL CONDITIONING The system was furnished with instrumentation to monitor both plant microclimate and soil water condition. The microclimate parameters monitored included leaf temperature with two infrared thermocouples (IRTS–P, Apogee Instruments, Logan, Utah), air temperature measurement using one type–K thermocouple, ambient air velocity using a hot–wire anemometer (TSI 8455–12, TSI Inc., St. Paul, Minn.), light intensity using a pyranometer (PY 8017, Li–Cor Inc., Lincoln, Neb.), relative humidity (H3V–200, Rotronic Instrument Corp., Huntington, N.Y.), CO2 concentration (CMP111, Vaisala Inc., Columbus, Ohio), and evapotranspirational water loss using load cells (Model 355, Tedea–Huntleigh, Calif.). Soil moisture measurement was accomplished using a time domain reflectometry device (Model ML2, Dynamax, Houston, Texas). Two Campbell Scientific data loggers (Model 21X and 23X) were used. The 21X with a multiplexer was mounted at the center of the turntable, while the 23X was located outside of the chamber. In addition to data storage, analog channels of the data loggers sent data to the host computer via a data acquisition board (DAQ AT–MIO–16XE, National Instruments, Austin, Texas). Data loggers were also used as signal conditioning devices. The Campbell 21X data logger provided required excitation voltage for load cell operation 992

and was used for temperature referencing and signal conditioning required for air temperature measurement using thermocouples. Sensor body temperature compensation required for accurate infrared thermometry temperature measurement was accomplished using a program with a Campbell 23X data logger, and the relative humidity sensor and hot wire anemometer were connected to this logger. Soil moisture sensors were connected directly to the data acquisition board. Analog output channels of both data loggers sent data to the host computer for data storage. INFRARED TEMPERATURE SENSING Infrared sensors were mounted at a stationary location over the path of the plants to enable measurements of plant canopy temperatures. Leaf temperature of the plants was measured when they were beneath the infrared sensors. Readings from two sensors were averaged to obtain a representative canopy temperature. The accuracy of the infrared thermocouple sensors was ±0.2°C within the 15° to 40°C operating range specified by the manufacturer. Use of non–contact, infrared temperature sensing provides rapid readings of surface temperatures for accurate plant temperature readings. However, to reduce measurement errors, readings must be taken carefully so that only the target object is within the sensor’s field of view. In order to determine the field of view, one of the infrared sensors was positioned vertically at a fixed height. A small LED point light source, painted flat black, was moved slowly towards the field of view of the infrared sensor while the sensor continuously read the temperature of a flat white background. The LED light source was moved until a sudden increase in the temperature reading was observed. The LED’s location was then marked. This procedure was repeated several times with the LED approaching from different directions. A circle was drawn passing through marked positions to define the field of view (FOV) of the sensor. Based on measured FOV and the distance between the infrared thermocouple and its target plane, the ratio of the sensor’s distance to the FOV was found to be 3:1; that is, at a distance of 0.03 m the sensor has a circular FOV of 0.01 m diameter. To properly position sensors so that the sensors did not shade the leaves, the sensors were located 0.05 m above the center of the plants. A diagram of this procedure and the FOV of the sensor are shown in figure 4. The output of the infrared thermocouple sensor is affected by both target temperature and the temperature of the sensor itself. Uncompensated variation in the sensor body temperature results in an erroneous target temperature reading. Bugbee et al. (1998) minimized this error by adding thermal mass around the detector to prevent rapid temperature changes and to keep all parts of the sensor at the same temperature. The sensor used in this study also had a thermal mass around the detector to prevent rapid temperature changes and their effect on temperature measurement errors. Accuracy of the infrared sensors was tested using a laboratory procedure. The temperatures of a blackbody reference (painted flat black), measured with a thermometer, were compared against the infrared sensor readings. Figure 5 illustrates the resulting curve obtained from the sensor calibration.

TRANSACTIONS OF THE ASAE

Infrared thermocouple sensor

25.7 oC

Data logger

Power supply LED light source

Flat white background

Blackbody temperature ( C)

Figure 4. Illustration of procedure used to determine field of view of the infrared thermocouple.

5

45 40

y = 1.0284x – 1.2221 R 2 = 0.998

35 30 25 20 15 10

were acquired sequentially, two seconds apart, five seconds after a plant was brought to a new position. The top projected plant canopy area (TPCA) was extracted from each acquired image before the average of four TPCAs was used to represent the plant canopy size. SOIL MOISTURE SENSING In order to measure soil moisture in the pot, a soil moisture sensor (Model ML2, Dynamax, Houston, Tex.) was used (fig. 6a). The working range of the sensor was 0% to 100% volumetric soil moisture with an accuracy of ±2%. The sensor outputted a voltage signal, and this voltage was converted to volumetric soil moisture readings using calibration constants that were obtained for each sensor (Kacira, 2000). The response time of the probes was less than 0.5 s. The sensors were inserted directly into root zone from the side of the pots (fig. 6b, c, and d). The sensing principle of the soil moisture sensor is based on time domain reflectometry, which measures the changes in apparent dielectric constant as a function of soil moisture content. This sensor measures soil parameters via a specially designed transmission line whose impedance is changed as the impedance of the soil changes. (ML2 User Manual, 1998). Attention should be paid to removing and re–inserting the probe into a previously inserted soil location, which may create air pockets around measuring rods. Air pockets between the rods and soil may cause errors in moisture readings (reducing the value of soil moisture content measured). The sensors in this design were inserted and sealed to the pots; thus, they were stationary in their original position. This sensor requires one–time calibration for the soil medium being used (ML2 User Manual, 1998).

5 0 5

10 15 20 25 30 35 40 45

Infrared thermocouple temperature ( 5C Figure 5. Curve obtained from the infrared sensor calibration.

IMAGE ACQUISITION AND TPCA MEASUREMENT The image acquisition system consisted of two main components: a monochrome CCD camera (Pulnix TM–200, Pulnix America Inc., Sunnyvale, Calif.) and a 640 × 480 × 8 frame grabber board (Matrox Meteor II Standard, Matrox Electronic Systems Ltd., Quebec, Canada) installed in the host computer. The camera was mounted perpendicular to the horizontal plane at a height of 1.0 m above the turntable. The system was capable of acquiring images of each plant at a given time interval. The image seen by the camera was the top view of the plant canopy. The system was designed for light–period image acquisition only. A 32 mm double acrylic diffuser barrier was installed between the lights of the chamber and the growing area. The diffuser barrier made shadow–free lighting possible. For more detail on growth chamber configuration, the reader is referred to Kacira (2000). Many steps were taken to improve plant canopy image quality, which can be adversely affected by plant movements caused by turntable movement, vibration of the chamber, and air movement in the chamber. Four images of the same plant

Vol. 44(4): 989–996

RESULTS AND DISCUSSION PERFORMANCE OF THE SYSTEM Figure 7b illustrates the effect of the offset loading on load–cell reading errors. Weights of the center–loads applied are shown on the x–axis, while percent measurement errors are illustrated on the y–axis. The offset load used was 50 g 15.2 cm 6.0 cm

a

c

11.2 cm

9.8 cm

0

6.5 cm

b

d

Figure 6. Location of soil moisture sensor in the pot: (a) the soil moisture sensor, (b) design of the pot, (c) insertion of the sensor to the pot, and (d) sensor location in the soil medium.

993

Top view

Center load

8 4

1

2

Offset load

3

6 Load cell

7 Weighing platform

a

Measurement Error (%)

1.0 0.8 0.6 0.4 0.2 0.0 –0.2 –0.4 –0.6 –0.8 –1.0 2000

1000

500

100

50

Weights (gr) 1

2

3

4

6

7

8

9

5

b

Figure 7. (a) Diagram of the laboratory platform used to test the effect of offset loading on weight measurements. Numbers indicate positions of the offset loads evaluated. (b) Measurement errors due to the offset loadings.

994

TPCA (pixels)

20840 20800 20760 20720 20680 1

2

3

4

5

6

7

8

9 10 11 12 13 14 15 16

Image acquisition event

Figure 8. Effect of air movement on the measured top projected canopy area (TPCA). The error bars show the TPCA deviations of four consecutively acquired images of the same plant at imaging events.

0.35

m

–3

)

0.40

3

Initial moisture: 27% Moisture set point: 38% Error: 0.35% after reaching to set point

0.30

0.25 0

100 200 300 400 500 600 700 800 900 1000 1100 Time (min)

Figure 9. Automatic control of soil moisture, based on feedback from the soil moisture probe. An on–off control algorithm was implemented to achieve minimum overshoot of moisture level in the growth medium.

Side view

9

5

20880

Qv (m

during the tests. The measurement error was found to be ±0.4% for all offset load positions tested with weights from 2000 to 100 g. Percent measurement errors were much larger when the center load was less than 100g. These results suggest that the effect of offset load on the load cell readings would be negligible if the total weight of the material being measured on the weighing platform was greater than 500 g. In this study, the total weight of the pot, soil, plant, and the moisture sensor ranged from 1000 to 1400 g. Figure 8 was plotted to illustrate the magnitude of air movement–induced TPCA change. Each data point in the figure is an averaged TPCA value of four sequentially acquired images of the same plant. The error bar on each data point represents the standard deviation of each point. This graph shows that air movement, chamber vibration, and other random noises affected TPCA measurements. Standard deviations of the four consecutive images were less than ±10 pixels. By averaging four sequentially acquired images to obtain a single TPCA value, we were able to improve the reliability of the determined TPCA values. The irrigation control capability of the system is shown in figure 9. An on–off control algorithm was implemented to achieve minimum overshoot of moisture level in the growth medium. An overshoot of 0.35% was found when a step change from 27% to 38% soil moisture content was implemented. These data showed that accurate control of soil moisture level was possible using an on–off control scheme and soil moisture sensor feedback. Performance of the system and an example of its application are presented in the next section.

POTENTIAL RESEARCH APPLICATIONS OF THE PLANT MONITORING SYSTEM To demonstrate potential applications of the automated plant monitoring system, a brief illustration of its use for plant–water stress detection of New Guinea Impatiens plants is provided below. More detailed information about the experiments is available from Kacira (2000). The plant’s water status was assessed using two non–contact sensing techniques. One of these used an index, called crop water stress index (CWSI), to determine the stress level to which the plants were subjected. CWSI is a function of both plant canopy temperature and the microclimatological data, and has values between zero and one. A value close to zero indicates that the plants experienced no water stress, while larger CWSI values indicate that the plants were under water stress. The second sensing technique used changes of top projected canopy area (TPCA) to assess plant water stress. The TPCA was determined from images captured using the machine vision system. Figure 10 illustrates a typical relationship between CWSI and measured evapotranspiration rates over time of well–watered and water–stressed plants. The data points in the graphs are CWSI and TPCA values collected from each plant at a given time without averaging. The curves in each figure show the data of one plant from each group obtained in one of the experiments. A total of six plants were used in each of the four experiments to determine the feasibility of using CWSI for water stress detection. CWSI of the well–watered plant was found to be close to zero and

TRANSACTIONS OF THE ASAE

0.4 0.2 0.0 –0.2 1

2

3

4

5

6

7

8

9 10 11 12 13 14

–0.4 –0.6 –0.8 –1.0

–1

hr

0.6

1200 1100 1000 900 800 700 600 500 400 300 200 100 0

–2

Crop water stress index

0.8

Measured ET rate (g m

1.0

)

remained close to zero throughout the experimental period, while the CWSI value of the water–stressed plant started to increase on days 4, 8, and 12 when the evapotranspiration rate decreased, causing less cooling effect on the plant. Increasing CWSI values indicated that the plant was experiencing water stress. Following irrigation on days 5, 9, and 12, the CWSI value of the water–stressed plant dropped to lower values. Using the TPCA detection method, a well–watered healthy plant is expected to show an increasing TPCA measurement due to normal growth, as the solid line shows in figure 11. The TPCA curve for a stressed plant shows that as the plant experiences water stress, the turgor pressure in the leaf decreases, resulting in changes in canopy geometry. This was reflected as a decreasing trend on the TPCA curve, indicated by valleys on the dotted line in figure 11. The TPCA value of the stressed plant increased after the plant was watered, indicating that the plant rehydrated and recovered from the stressed condition. As seen in the figure, water stress caused TPCA to decrease temporarily. TPCA recovery and further expansion were observed after the irrigation events.

Days CWSI(Stressed) ETm(Stressed)

CWSI(Well–watered) ETm(Well–watered)

Figure 10. History of crop water stress index (CWSI) and measured ET (ETm) rates of New Guinea Impatiens. CWSI was determined using canopy temperature and other microclimate information. ETm was determined using the micro–lysimeter measured data.

the infrared sensors to see through the canopy to the soil medium, which might be warmer, thereby causing an inaccurate canopy temperature reading and affecting the CWSI values. To assure data integrity, therefore, it is important to make sure the field of view of the infrared temperature sensors is fully occupied by the targeted plant canopy. Otherwise, the canopy temperature will be compromised due to the inclusion of soil background temperature.

CONCLUSIONS An automated plant monitoring system has been developed for plant water status assessment. Based on a distributed signal processing hierarchy, data loggers were used for local sensor excitation and signal processing to reduce the complexity of the computer–based data acquisition system. The system can be used to collect environmental data, measure canopy temperature, and detect morphological changes affected by experimental treatments using non–contact sensors. Contact sensors such as soil moisture probes and micro–lysimeters are available for determining evapotranspiration rate. Sources of sensing and measurement error were analyzed and quantified, and measures were taken to improve the data accuracy. Precise soil moisture level could be maintained using a closed–loop irrigation sub–system. The system is a versatile platform for plant monitoring and closed–loop plant production research with minimum labor requirements. A wide range of data can be collected automatically and continuously, providing temporal data of high resolution for plant characterization in response to research treatments. ACKNOWLEDGEMENTS The authors would like to acknowledge the help of Michael Sciarini, electronic design engineer, and William Kreider and Roger Moss, mechanical design engineers, on the design and construction of the system reported in this article.

24500

T P C A (pixels)

23500

REFERENCES

22500

21500

20500

19500

18500 1

2

3

4

5

6

7

8

9

10

11

12

13

14

Days TPCA(stressed)

TPCA(Well–watered)

Figure 11. History of top projected canopy area (TPCA) change of well– watered and water–stressed plant. TPCA was measured using the machine vision capability of the system.

As seen from figures 10 and 11, CWSI increased at the same time as stress–induced canopy movement was detected. A possible contributing factor is that leaf movement allowed

Vol. 44(4): 989–996

Bugbee, B., M. Droter, O. Monje, and B. Tanner. 1998. Evaluation and modification of commercial infra–red transducer for leaf temperature measurement. Advances in Space Research 22(10): 1425–1434. Choudhury, B. 1983. Simulating the effects of weather variables and soil water potential on a corn canopy temperature. Agric. Meteorology 29: 169–182. Choudhury, B., and S. B. Idso. 1984. Simulating sunflower canopy temperatures to infer root–zone soil water potentials. Agric. and Forest Meteorology 31: 69–78. Clark, R. N., and E. A. Hiler. 1973. Plant measurements as indicators of crop water deficit. Crop Science 13: 466–469. Clawson, K. L., R. D. Jackson, and P. J. Pinter. 1989. Evaluating plant water stress with canopy temperature differences. Agron. J. 81: 858–863. De Lorenzi, F., C. Stanghellini, and A. Pitacco. 1993. Water shortage sensing through infrared canopy temperature: Timely detection is imperative. Acta Horticulturae 335: 373–380.

995

Ehrler, W. L., S. B. Idso, R. D. Jackson, and R. J. Reginato. 1978. Diurnal changes in plant water potential and canopy temperature of wheat as affected by drought. Agron. J. 70: 999–1004. Gates, D. M. 1964. Leaf temperature and transpiration. Agron. J. 56: 273–277. Idso, S. B., R. D. Jackson, P. J. Pinter, R. J. Reginato, and J. L. Hatfield. 1981. Normalizing the stress degree day parameter for environmental variability. Agric. Meteorology 24: 45–55. Jackson, R. D., S. B. Idso, R. J. Reginato, and P. J. Pinter. 1981. Canopy temperature as a crop water stress indicator. Water Resources Res. 17(4): 1133–1138. Kacira, M. 2000. Non–contact and early detection of plant water stress using infrared thermometry and image processing. Ph.D. dissertation. Columbus, Ohio: Ohio State University.

996

Ling, P. P., G. A. Giacomelli, and T. Russell. 1996. Monitoring of plant development in controlled environment with machine vision. Advances in Space Research 18(4–5): 101–112. ML2 User Manual. 1998. ThetaProbe: Soil Moisture Sensor. Burwell, U.K.: Delta–T Devices, Ltd. Oosterhuis, D. M., S. Walker, and J. Eastham. 1985. Soybean leaflet movements as an indicator of crop water stress. Crop Science 25: 1101–1106. Pang, T. 1992. Dynamic analysis of water and nutrient uptake for new guinea impatiens. Ph.D. dissertation. Columbus, Ohio: Ohio State University. Stanghellini, C., F. De Lorenzi, A. H. Bosma, and C. Werkhoven. 1992. Early detection of water stress in sub–humid climates. Report 92–9. Wageningen, The Netherlands: IMAG–DLO. Seginer, I., R. T. Elster, J. W. Goodrum, M. W. Rieger. 1992. Plant wilt detection by computer–vision tracking of leaf tips. Trans. ASAE 35(5): 1563–1567.

TRANSACTIONS OF THE ASAE