THE INFLUENCE OF COTTON VARIETY IN THE ... - Semantic Scholar
THE INFLUENCE OF COTTON VARIETY IN THE CALIBRATION FACTOR OF A COTTON YIELD MONITOR A. T. Markinos1, T. A. Gemtos1, D. Pateras , L. Toulios2, G. Zerva2, M. Papaeconomou3 2
1
Laboratory of Farm Mechanization, School of Agricultural Sciences, Faculty of Crop Production and Rural Environment, University of Thessaly, Fytoko Street, N. Ionia, Magnisias 384 46, Tel: +302421093228, email: [email protected] 2 NAGREF/ISM, Theofrastou 1, Larissa 413 35, email: [email protected] 3 Papaeconomou Agrochemicals S.A., email: [email protected]
ABSTRACT A three years study on precision farming in cotton in central Greece showed serious infield variability of yield and soil physical and chemical properties. Yield mapping constitutes the starting and ending point of the whole process chain in a precision farming system. Yield mapping was performed for three consecutive years in the same cotton fields of central Greece in a wide range of cotton varieties and field conditions with a yield monitor installed on a cotton picker. The yield sensors estimate the flow volume of cotton conveyed from the picking units in the basket through air ducts. The main control unit performs the transformation of the estimated cotton volume in cotton weight using a factor named Calibration Coefficient or Factor. Every time the picking conditions change, like going to a different field or change cotton variety, there is the need to pick a whole load basket and make an actual weighing to inform the main unit to adapt the calibration factor to the correct value. After three cotton yield mapping seasons (overall 200ha) it was observed that there is a straight relation of the value of calibration factor with the cotton variety. The present study shows these results that group the values of calibration factor with corresponding cotton varieties. The results would help in the calibration of the monitors. Keywords: cotton, yield mapping, varieties, yield monitor, calibration Introduction Agricultural production sector faces a twofold problem in the beginning of new century. At one side, the globalization framework increases the competition, decreases product substitutes that cause decreased product prices and profits. On the other side, the food safety, traceability and environmental surcharge drive to a transformation in production practices. The new applications of information technology and electronics seem to have a serious potential to overcome all these issues. Precision Farming constitutes of the application of new technology that deals with soil and crop variability to drive in management practices optimizing yield and profit and decreasing environmental impact as well [1]. Many precision farming applications have been evaluated in USA, Australia and North Europe last decade giving positive results, especially in large farms with variable soil conditions [2]. South Europe farms characterized of the small area fields and the low adaptation to the new technology along with the aging of the farmer [3].
One of the characteristics of Greek agriculture is the small farms (average area 5.6 ha) in more than one plots. The main cultivated crop is cotton that is entirely irrigated and mechanized using cotton pickers for harvesting. A project to apply the PF techniques to the small cotton growing farms was undertaken for the last three years to investigate the feasibility of using Precision Farming techniques in small farms and the justification of the involved investment [4, 5]. The main target was to find better crop management methods to assist farmers. A precision agriculture system can be divided in three different phases (fig. 1) [6]. The first one relates to the acquisition of crop data, mainly old yield maps and soil analysis data. Second phase concerns the processing and calculations on the data that have been gathered. The way of combining data in every system depends on the crop type and the implemented algorithm. The existence of an advanced upgradeable library or database provides the system with the necessary intelligence to work out many different conditions and solutions. Third phase relates to the implementation and adaptation of cultivation practices according to system results. It needs the modification of the equipment to accommodate variable rate application of agricultural inputs in the field area. This specific operation targets to counterbalance the variations of the physical and chemical soil properties of different parts of the field. All these tasks establish a cyclic process with a cause-result procedure. Yield mapping constitutes the starting and ending point of the whole process chain in a precision farming system. The basic procedure of data acquisition system is the yield mapping during the crop harvest. In crops like grains, the operation of these systems has already a life of more than a decade. All this time, many commercial systems have been developed from different manufacturers [7, 8]. Current systems have sensors based on different working operations like mass flow, volumetric flow, impact and others [9]. Sensors recording the cotton flow have recently developed. The development was late due to the difficulty to measure the erratic shape and texture of cotton during the conveyance with air stream through air ducts to the storage basket [10]. The majority of cotton sensors in the market, at the moment, are based on measuring the volume of cotton flow through air ducts from the picking point to the basket of the machine [11, 12]. Vellidis et al. [13] have compared different cotton yield monitoring systems that were used in the USA. They found that all the systems compared could produce adequate yield maps, provided they are properly calibrated, operated and maintained. All systems converted the volumetric flow of cotton to mass flow using a calibration factor. A GPS receiver with the corresponding antenna provides the system with the actual position of the machine in the field area. Additional sensors inform the system about the vertical position of picking units and count the actual speed of cotton picker. All these parts of equipment connected to the main control unit, installed in the cabin of the machine. Real time yield determination needs the measure of flow rate of seed cotton as it is conveyed from the picking unit to the machine basket through air ducts. That means that an optically based systems measures cotton flow. An early work described such an optical system that performed well in preliminary laboratory tests [14]. But, it had problems with dynamically changing ambient light when applied to a cotton picker machine. Another system consisted of a light source paired with a light sensing bar has been tested since 1989 [15]. Although the results of this research were positive, a second optical device was tested in a related study [15]. A version of this system was mounted on a harvester and tested. Test results showed promising results but some problems were reported with the dust and cotton strings formed in any surface undulation. Later, several studies were performed that been tested two commercially available optical sensors [16, 17]. The main problems reported were the dust and debris built-up on sensor eyes. Many other researchers have suggested ways to overcome these problems [18].
Every commercial optical yield sensor installed on an air duct consisted of a pair of an infrared light emitter array and a detector array mounted exactly opposite each other, figure 1 [19].
Figure 1. The emitter light array mounted opposite the detector array on air duct Each eye of the light detector array has a limited field of view and is paired with the corresponding emitter-eye at the opposite side of the air duct, from which receives the light energy. Every eye of the detector is an integrated circuit that has a frequency-based digital output. The output frequency is linearly related to the intensity of the received light to the active area of the eye. The more light striking the active area of detector element, the greater frequency of the device output pulses occurs. Each of the detector outputs is coupled to digital hardware counter that is interfaced with the main control unit or data logger. The counters sum all the pulses produced by the detectors over an integration interval. At the end of the integration interval the data acquisition system reads and clears all the corresponding counters. The resulting sums that have been measured in each integration interval represent the total amount of the light energy received by each detector in this time cycle. Conclusively, each measurement of total light energy is inversely related to the amount of cotton that passed between the emitter and detector during the integration interval. After a transformation process the attenuation of the light beam for each pair of emitter-detector sensor, calculated using proprietary algorithms unique to each yield monitor [15, 19]. Assuming that we use n yield sensor pairs mounted on n air ducts of the cotton picker, we can accumulate the sensing unit outputs Aj from each duct and multiply by a scalar factor CF to obtain an estimation of the cotton mass flow passed through air ducts during the integration interval: n
M = CF ∑ Aj
(1)
j =1
where: • M is the cotton mass flow passed through the air ducts during the integration interval, • CF is the scalar factor or calibration factor, • n is the number of sensing units of yield monitor, • Aj is the sensing unit output from the jth sensing unit. During cotton picking operation the main control unit of yield monitor stores the readings of the cotton mass parcels Μi generated by the yield sensors readings in the corresponding i integration interval. Simultaneously, the main control unit estimates the weight of cotton seed picked in the
machine basket by accumulating the Μi mass parcels over the whole number m of integration interval have been passed. That is: m
m
n
i =1
i =1 j =1
M est .load = ∑ M i = CF ∑∑ Aij
(2)
The total accumulation of the n sensing unit outputs over the m integration intervals passed gives the summed amount A of the total attenuated light energy or the linearly related cotton flow volume:
M est .load = CF ⋅ A
(3)
The Calibration Coefficient or Calibration Factor CF is a native number that used by the main control unit to calculate the estimated weight of cotton have been picked. According to the literature this number depends mainly from the picking conditions and the cotton variety [15, 19]. The present study is based on the three years yield mapping data and tries to group the values of calibration factor with corresponding cultivated cotton varieties. Materials and Methods During the 2001, 2002 and 2003 harvesting periods, yield mapping was performed in order to provide maps for farmer-owned and managed cotton fields in central Greece. All the fields located at Myrina, in Karditsa prefecture where cotton is the major cultivated crop in the small fields of the area. A commercial yield monitor system from Australian Farmscan was used on a two row John DeereTM model 9910 cotton picker. The yield mapping system consists of the main control unit (canlink 3000); the two pairs of infrared yield sensors, the speed and units position sensors and the GPS antenna with the corresponding receiver. Figure 2 represents the picker diagram and the install points of the system components.
Figure 2: Cotton picker diagram showing yield monitor components installed The central unit is the basis of the whole system (fig. 3a). During the operation, it gathers and stores the sensor data every 2s and simultaneously shows them on the incorporated LCD display in a form of current yield, average yield, harvested area and the estimated weight of cotton in the basket. Auxiliary signals from a height hold and a speed sensor are also sensed (fig. 2). Logged data are a combination of geographical coordinates (latitude and longitude) of the spot of cotton picker, with the speed and yield at the same point. All data stored in a memory card
(SRAM) with 2-MB capacity. Through the main control unit the driver can set a new value for the calibration factor CF (eq. 3).
Figure 3: a) the main control unit in the cabin, b) the yield sensors installed After harvesting completion the yield data were transferred to a personal computer where using a GIS software yield maps are generated. Several methods (inverse distance weight, kriging, etc.) can be used for spatially data interpolation in order to produce solid yield maps with discrete zones of equal yield (fig. 4a, b).
Figure 4: a) 1st picking yield map for 2001, b) 1st picking yield map for 2002 (4.3 ha) The system operation was successful and there were no problems. One point has to be mentioned, is the attention in the sensors cleaning from dust. When the system operated without sensor cleaning for long period of time, there were eye blockages of the sensors and the system stopped measuring. For smooth operation of the system, the yield sensors should be cleaned at every picking of a full basket. The first year (2001), yield mapping was performed successfully in two fields, 4.3 ha and 1.7 ha respectively [6]. During harvesting season of the second year (2002), yield maps were made in 15 fields and about 80 ha including the fields of the first year. The yield mapping of the third year (2003) was performed in about 80 ha and in majority at the same fields with the second year. All the three years of evaluation observed that the actual value of the calibration factor depends of the picking conditions, especially moisture, and the cultivated cotton variety. If we keep the picking conditions in a stable range that means picking performed only at day without moisture and only at defoliated open crops, a possible relation between the value of calibration factor and cotton variety could be investigated. Every time picking was performed the main control unit has a unique value for the calibration factor using it to estimate the weight of the picked cotton. Suppose that a full load basket have been picked and the main control unit uses the value of CF to estimate the weight of the picked
load that is Μest (eq. 3). Performing an actual weighing to the load of seed cotton picked found the actual weight Mactual, which differs from the Mest. That means that the system needs to adapt the CF to a new value CF´ that is:
M act .load = CF′ ⋅ A
(4)
Equations (3) and (4) give:
CF′ =
M act ⋅ CF M est
(5)
During yield mapping and at every different field the system had by default a known calibration factor value. For each load harvested, the field details (owner, farmer, location, id and area), cotton variety, estimated load weight and actual load weight have been logged in a database. A two wheel platform trailed by a tractor was used to emptying every picked basket. The platform filled with two consecutive depletions of the machine basket (one platform load equal to two basket loads) and transported to the nearest scale for actual weighing. The database was updated by the actual weight of every platform load. During the first evaluation of the yield monitor (year 2001) were observed that under normal conditions like day picking, defoliated crop, common varieties, a calibration factor value around 2850 gave the most reliable results. For this reason the default calibration factor value of 2858 was used for yield estimation (eq. 5). For every platform load having weighted in the database the new value of the calibration factor was calculated using equation 5. A filtering of the results was performed removing the loads that had been picked with abnormal picking conditions. Only loads that had been picked between 11:00 in the morning to 17:00 in the afternoon in a defoliated crop were accepted. Then, results were grouped according the type of cotton variety giving nine different classes. The nine different varieties that used are: Bolina, Carmen, Celia, Ermis, Fotini, Lider, Millenium, Sandra, Christina. Results and Discussion Figure 5 shows the calibration factors in relation to the cotton variety have been calculated following the above procedure. Picked platform loads from year 2002 and 2003 were used. The following Figure 6 represents the relation of the estimated load weights using the default calibration factor of 2858 to the actual load weights according the nine different cotton varieties.
3200.00
R R R R
3100.00 R R R R R
3000.00
Calibration Factor
R R
2900.00
2800.00
R
R R R R
R R R R R R R R R R R R R R R R R R R R R R R R R R R R R R R R
R R R R R
R R R
R R R R R
2700.00
R
2600.00 R
2500.00
R
XRISTINA
SANDRA
MILLENIUM
LIDER
FOTINI
ERMIS
CELIA
CARMEN
BOLINA
R R R
Variety
Figure 5: Calculated calibration factors for every load and cotton variety y = 0.9771x R2 = 1
Estimated Load Weight(Kg)
2500
2000
CELIA CARMEN
1500
XRISTINA LIDER ERMIS SANDRA
1000
BOLINA FOTINI MILLENIUM 500 500
1000
1500
2000
2500
Actual Load Weight (Kg)
Figure 6: Comparison of the estimated weights to actual load weights by variety (CF=2858).
Next Figure 7 shows the mean value of the calculated calibration factor for every cotton variety using equation 5. The standard deviation range is shown in the same diagram. 3250.00
W
3000.00
W
3061.50
2990.00
W
2927.50
2872.84 W
2799.40
W
2750.00
W
2775.57
2760.00
2632.33
W
MILLENIUM
FOTINI
BOLINA
2488.75
SANDRA
XRISTINA
CARMEN
CELIA
W
ERMIS
2500.00
LIDER
Calibration Factor Value
W
Variety
Figure 7: The mean and standard deviation for the calibration factor according the corresponding cotton variety Conclusions These first results show that there is a strong correlation of the value of calibration factor to the cotton variety in stable picking conditions. Furthermore, some conclusions could be written: - Every variety has a corresponding range of calibration factor values, - It is possible to improve the estimation of the actual yield during the picking putting a prescribed calibration factor according the cultivated variety of every field at the beginning of harvest, - It is possible to characterize and classify a harvesting load in specific categories from the calibration factor, - It is possible to reduce the estimation error knowing the variety type, when it is impossible to make an actual weighing based calibration of the system.
- The value of the calibration factor gives a first indication about the density of every variety and it is of value to the producer, since he is paid for the weight of seed cotton, More research and observations are needed covering many different varieties and commercial yield monitors to establish a clear deductive methodology. More effort is necessary to eliminate and model the external factors that influence the picking conditions. References 1.
Auernhammer, H., (2001), Precision farming – the environmental challenge, Computers and Electronics in Agriculture, 30, pp. 31-43.
2.
Stafford J.V.. (2000), Implementing Precision Agriculture in the 21st Century, J. Agr. Engng Res.(76) 267-275
3.
Gemtos, T., Fountas, S., Blackmore, B.S., Greipentrog, H.W. (2002), Precision farming experience in Europe and the Greek potential, In: Proceedings of the 1st HAICTA Conference, Athens, 6-7 June 2002: 45-55.
4.
Gemtos, T.A., Markinos, Ath., Toulios, L., Pateras, D., Zerva, G. (2003), A Precision Farming Application in the Small Cotton Farms of Greece, Presented in ITAFE 2003 conference, Izmir, Turkey, 7-9 October 2003, pp.
5.
Markinos, Ath, Toulios, L., Pateras, D., Zerva, G., Gemtos, T. A., (2003), Poster in the 4th European Conference on Precision Agriculture, Berlin 16-18 June 2003, pp. 489-490.
6.
Markinos, A., Gemtos, T., Toulios, L., Pateras, D., Zerva, G. (2002), Yield Mapping of Cotton crop in Greece, In: Proceedings of the 1st HAICTA Conf., Athens, 6-7 June 2002: 56-62.
7.
Reyns, P., Missotten, B., Ramon, H., De Baerdemaeker, J., (2002), A Review of Combine Sensors for Precision Farming, Precision Agriculture, 3, pp. 169-182.
8.
Grisso, R.D., Jasa, P.J., Schroeder, M.A., Wilcox, J.C. (2002), Yield monitor accuracy: Successful farming magazine case study, Applied Engineering in Agriculture 18(2), pp. 147-151.
9.
Arslan, S., Colvin, T.S., (2002), Grain Yield Mapping: Yield Sensing, Yield Reconstruction, and Errors, Precision Agriculture, 3, pp. 135-154.
10. Searcy, S.W., Motz, D.S., Inayatullah, A., Goering, K.J., (1997), Cotton yield mapping. In: Dugger, P., Richter, D.A. (Eds.), Proceedings Beltwide Conference, 6-10 January 1997, New Orleans, LA. National Cotton Council of America, Memphis, TN, pp. 601-603. 11. Perry, C.D., Durrence, J.S., Thomas, D.L., Vellidis, G., Sobolik, C.J., Dzubak, A. (1998), Evaluation of commercial cotton yield monitors in Georgia field conditions, In. Proc. Fourth International Conference on Site-Specific Management for Agricultural Systems. Madison Wis., pp. 1227-1240. 12. Wilkerson, J.B., Moody, F.H., Hart, W.E. (2002), Implementation and field evaluation of a cotton yield monitor. Applied Engineering in Agriculture 18(2), pp. 153-159. 13. Vellidis, G.C., Perry, D., Rains, G.C., Thomas, D.L., Wells, N., Kvien, C.K. (2003), Simultaneous assessment of cotton yield monitors. Applied Engineering in Agriculture, Vol 19(3), pp. 259-272.
14. Wilkerson, J.B., Kirby, J.S., Hart, W.E., Womac, A.R., (1994), Real-time cotton flow sensor. ASAE Paper No. 941054. American Society of Agricultural Engineers, St. Joseph, MI. 15. Thomasson, J.A., Pennington, D.A., Pringle, H.C., Columbus, E.P., Thomson, S.J., Byler, R.K. (1999), Cotton mass flow measurements: Experiments with two optical devices. Applied Engineering in Agriculture 15(1), pp. 11-17. 16. Wallace, T.P., (1999), Small plot evaluation of an electro-optical cotton yield monitor. Computers and Electronics in Agriculture, 23, pp. 1-8. 17. Durrence, J.S., Perry, C.D., Vellidis, G., Thomas, D.L., Kvien, C.K. (1998), Evaluation of commercially-available cotton yield monitors in Georgia field conditions. ASAE Paper No. 98-3106, St. Joseph, Mich.: ASAE. 18. Wolak, F.J., Khalilian, A., Dodd, R.B., Han, Y.J., Keshkin, M., Lippert, R.M., Hair, W. (1999), Cotton yield monitor evaluation, South Carolina: Year 2. In Proc. Beltwide Cotton Conf., pp. 361-364. Orlando, Fla. 3-7 January. Memphis, Tenn.: National Cotton Council. 19. Wilkerson, J.B., Moody, F.H., Hart, W.E., Funk, P.A. (2001), Design and evaluation of a cotton flow rate sensor. Transactions of the ASAE 44(6), pp. 1415-1420.
1
Laboratory of Farm Mechanization, School of Agricultural Sciences, Faculty of Crop Production and Rural Environment, University of Thessaly, Fytoko Street, N. Ionia, Magnisias 384 46, Tel: +302421093228, email: [email protected] 2 NAGREF/ISM, Theofrastou 1, Larissa 413 35, email: [email protected] 3 Papaeconomou Agrochemicals S.A., email: [email protected]
ABSTRACT A three years study on precision farming in cotton in central Greece showed serious infield variability of yield and soil physical and chemical properties. Yield mapping constitutes the starting and ending point of the whole process chain in a precision farming system. Yield mapping was performed for three consecutive years in the same cotton fields of central Greece in a wide range of cotton varieties and field conditions with a yield monitor installed on a cotton picker. The yield sensors estimate the flow volume of cotton conveyed from the picking units in the basket through air ducts. The main control unit performs the transformation of the estimated cotton volume in cotton weight using a factor named Calibration Coefficient or Factor. Every time the picking conditions change, like going to a different field or change cotton variety, there is the need to pick a whole load basket and make an actual weighing to inform the main unit to adapt the calibration factor to the correct value. After three cotton yield mapping seasons (overall 200ha) it was observed that there is a straight relation of the value of calibration factor with the cotton variety. The present study shows these results that group the values of calibration factor with corresponding cotton varieties. The results would help in the calibration of the monitors. Keywords: cotton, yield mapping, varieties, yield monitor, calibration Introduction Agricultural production sector faces a twofold problem in the beginning of new century. At one side, the globalization framework increases the competition, decreases product substitutes that cause decreased product prices and profits. On the other side, the food safety, traceability and environmental surcharge drive to a transformation in production practices. The new applications of information technology and electronics seem to have a serious potential to overcome all these issues. Precision Farming constitutes of the application of new technology that deals with soil and crop variability to drive in management practices optimizing yield and profit and decreasing environmental impact as well [1]. Many precision farming applications have been evaluated in USA, Australia and North Europe last decade giving positive results, especially in large farms with variable soil conditions [2]. South Europe farms characterized of the small area fields and the low adaptation to the new technology along with the aging of the farmer [3].
One of the characteristics of Greek agriculture is the small farms (average area 5.6 ha) in more than one plots. The main cultivated crop is cotton that is entirely irrigated and mechanized using cotton pickers for harvesting. A project to apply the PF techniques to the small cotton growing farms was undertaken for the last three years to investigate the feasibility of using Precision Farming techniques in small farms and the justification of the involved investment [4, 5]. The main target was to find better crop management methods to assist farmers. A precision agriculture system can be divided in three different phases (fig. 1) [6]. The first one relates to the acquisition of crop data, mainly old yield maps and soil analysis data. Second phase concerns the processing and calculations on the data that have been gathered. The way of combining data in every system depends on the crop type and the implemented algorithm. The existence of an advanced upgradeable library or database provides the system with the necessary intelligence to work out many different conditions and solutions. Third phase relates to the implementation and adaptation of cultivation practices according to system results. It needs the modification of the equipment to accommodate variable rate application of agricultural inputs in the field area. This specific operation targets to counterbalance the variations of the physical and chemical soil properties of different parts of the field. All these tasks establish a cyclic process with a cause-result procedure. Yield mapping constitutes the starting and ending point of the whole process chain in a precision farming system. The basic procedure of data acquisition system is the yield mapping during the crop harvest. In crops like grains, the operation of these systems has already a life of more than a decade. All this time, many commercial systems have been developed from different manufacturers [7, 8]. Current systems have sensors based on different working operations like mass flow, volumetric flow, impact and others [9]. Sensors recording the cotton flow have recently developed. The development was late due to the difficulty to measure the erratic shape and texture of cotton during the conveyance with air stream through air ducts to the storage basket [10]. The majority of cotton sensors in the market, at the moment, are based on measuring the volume of cotton flow through air ducts from the picking point to the basket of the machine [11, 12]. Vellidis et al. [13] have compared different cotton yield monitoring systems that were used in the USA. They found that all the systems compared could produce adequate yield maps, provided they are properly calibrated, operated and maintained. All systems converted the volumetric flow of cotton to mass flow using a calibration factor. A GPS receiver with the corresponding antenna provides the system with the actual position of the machine in the field area. Additional sensors inform the system about the vertical position of picking units and count the actual speed of cotton picker. All these parts of equipment connected to the main control unit, installed in the cabin of the machine. Real time yield determination needs the measure of flow rate of seed cotton as it is conveyed from the picking unit to the machine basket through air ducts. That means that an optically based systems measures cotton flow. An early work described such an optical system that performed well in preliminary laboratory tests [14]. But, it had problems with dynamically changing ambient light when applied to a cotton picker machine. Another system consisted of a light source paired with a light sensing bar has been tested since 1989 [15]. Although the results of this research were positive, a second optical device was tested in a related study [15]. A version of this system was mounted on a harvester and tested. Test results showed promising results but some problems were reported with the dust and cotton strings formed in any surface undulation. Later, several studies were performed that been tested two commercially available optical sensors [16, 17]. The main problems reported were the dust and debris built-up on sensor eyes. Many other researchers have suggested ways to overcome these problems [18].
Every commercial optical yield sensor installed on an air duct consisted of a pair of an infrared light emitter array and a detector array mounted exactly opposite each other, figure 1 [19].
Figure 1. The emitter light array mounted opposite the detector array on air duct Each eye of the light detector array has a limited field of view and is paired with the corresponding emitter-eye at the opposite side of the air duct, from which receives the light energy. Every eye of the detector is an integrated circuit that has a frequency-based digital output. The output frequency is linearly related to the intensity of the received light to the active area of the eye. The more light striking the active area of detector element, the greater frequency of the device output pulses occurs. Each of the detector outputs is coupled to digital hardware counter that is interfaced with the main control unit or data logger. The counters sum all the pulses produced by the detectors over an integration interval. At the end of the integration interval the data acquisition system reads and clears all the corresponding counters. The resulting sums that have been measured in each integration interval represent the total amount of the light energy received by each detector in this time cycle. Conclusively, each measurement of total light energy is inversely related to the amount of cotton that passed between the emitter and detector during the integration interval. After a transformation process the attenuation of the light beam for each pair of emitter-detector sensor, calculated using proprietary algorithms unique to each yield monitor [15, 19]. Assuming that we use n yield sensor pairs mounted on n air ducts of the cotton picker, we can accumulate the sensing unit outputs Aj from each duct and multiply by a scalar factor CF to obtain an estimation of the cotton mass flow passed through air ducts during the integration interval: n
M = CF ∑ Aj
(1)
j =1
where: • M is the cotton mass flow passed through the air ducts during the integration interval, • CF is the scalar factor or calibration factor, • n is the number of sensing units of yield monitor, • Aj is the sensing unit output from the jth sensing unit. During cotton picking operation the main control unit of yield monitor stores the readings of the cotton mass parcels Μi generated by the yield sensors readings in the corresponding i integration interval. Simultaneously, the main control unit estimates the weight of cotton seed picked in the
machine basket by accumulating the Μi mass parcels over the whole number m of integration interval have been passed. That is: m
m
n
i =1
i =1 j =1
M est .load = ∑ M i = CF ∑∑ Aij
(2)
The total accumulation of the n sensing unit outputs over the m integration intervals passed gives the summed amount A of the total attenuated light energy or the linearly related cotton flow volume:
M est .load = CF ⋅ A
(3)
The Calibration Coefficient or Calibration Factor CF is a native number that used by the main control unit to calculate the estimated weight of cotton have been picked. According to the literature this number depends mainly from the picking conditions and the cotton variety [15, 19]. The present study is based on the three years yield mapping data and tries to group the values of calibration factor with corresponding cultivated cotton varieties. Materials and Methods During the 2001, 2002 and 2003 harvesting periods, yield mapping was performed in order to provide maps for farmer-owned and managed cotton fields in central Greece. All the fields located at Myrina, in Karditsa prefecture where cotton is the major cultivated crop in the small fields of the area. A commercial yield monitor system from Australian Farmscan was used on a two row John DeereTM model 9910 cotton picker. The yield mapping system consists of the main control unit (canlink 3000); the two pairs of infrared yield sensors, the speed and units position sensors and the GPS antenna with the corresponding receiver. Figure 2 represents the picker diagram and the install points of the system components.
Figure 2: Cotton picker diagram showing yield monitor components installed The central unit is the basis of the whole system (fig. 3a). During the operation, it gathers and stores the sensor data every 2s and simultaneously shows them on the incorporated LCD display in a form of current yield, average yield, harvested area and the estimated weight of cotton in the basket. Auxiliary signals from a height hold and a speed sensor are also sensed (fig. 2). Logged data are a combination of geographical coordinates (latitude and longitude) of the spot of cotton picker, with the speed and yield at the same point. All data stored in a memory card
(SRAM) with 2-MB capacity. Through the main control unit the driver can set a new value for the calibration factor CF (eq. 3).
Figure 3: a) the main control unit in the cabin, b) the yield sensors installed After harvesting completion the yield data were transferred to a personal computer where using a GIS software yield maps are generated. Several methods (inverse distance weight, kriging, etc.) can be used for spatially data interpolation in order to produce solid yield maps with discrete zones of equal yield (fig. 4a, b).
Figure 4: a) 1st picking yield map for 2001, b) 1st picking yield map for 2002 (4.3 ha) The system operation was successful and there were no problems. One point has to be mentioned, is the attention in the sensors cleaning from dust. When the system operated without sensor cleaning for long period of time, there were eye blockages of the sensors and the system stopped measuring. For smooth operation of the system, the yield sensors should be cleaned at every picking of a full basket. The first year (2001), yield mapping was performed successfully in two fields, 4.3 ha and 1.7 ha respectively [6]. During harvesting season of the second year (2002), yield maps were made in 15 fields and about 80 ha including the fields of the first year. The yield mapping of the third year (2003) was performed in about 80 ha and in majority at the same fields with the second year. All the three years of evaluation observed that the actual value of the calibration factor depends of the picking conditions, especially moisture, and the cultivated cotton variety. If we keep the picking conditions in a stable range that means picking performed only at day without moisture and only at defoliated open crops, a possible relation between the value of calibration factor and cotton variety could be investigated. Every time picking was performed the main control unit has a unique value for the calibration factor using it to estimate the weight of the picked cotton. Suppose that a full load basket have been picked and the main control unit uses the value of CF to estimate the weight of the picked
load that is Μest (eq. 3). Performing an actual weighing to the load of seed cotton picked found the actual weight Mactual, which differs from the Mest. That means that the system needs to adapt the CF to a new value CF´ that is:
M act .load = CF′ ⋅ A
(4)
Equations (3) and (4) give:
CF′ =
M act ⋅ CF M est
(5)
During yield mapping and at every different field the system had by default a known calibration factor value. For each load harvested, the field details (owner, farmer, location, id and area), cotton variety, estimated load weight and actual load weight have been logged in a database. A two wheel platform trailed by a tractor was used to emptying every picked basket. The platform filled with two consecutive depletions of the machine basket (one platform load equal to two basket loads) and transported to the nearest scale for actual weighing. The database was updated by the actual weight of every platform load. During the first evaluation of the yield monitor (year 2001) were observed that under normal conditions like day picking, defoliated crop, common varieties, a calibration factor value around 2850 gave the most reliable results. For this reason the default calibration factor value of 2858 was used for yield estimation (eq. 5). For every platform load having weighted in the database the new value of the calibration factor was calculated using equation 5. A filtering of the results was performed removing the loads that had been picked with abnormal picking conditions. Only loads that had been picked between 11:00 in the morning to 17:00 in the afternoon in a defoliated crop were accepted. Then, results were grouped according the type of cotton variety giving nine different classes. The nine different varieties that used are: Bolina, Carmen, Celia, Ermis, Fotini, Lider, Millenium, Sandra, Christina. Results and Discussion Figure 5 shows the calibration factors in relation to the cotton variety have been calculated following the above procedure. Picked platform loads from year 2002 and 2003 were used. The following Figure 6 represents the relation of the estimated load weights using the default calibration factor of 2858 to the actual load weights according the nine different cotton varieties.
3200.00
R R R R
3100.00 R R R R R
3000.00
Calibration Factor
R R
2900.00
2800.00
R
R R R R
R R R R R R R R R R R R R R R R R R R R R R R R R R R R R R R R
R R R R R
R R R
R R R R R
2700.00
R
2600.00 R
2500.00
R
XRISTINA
SANDRA
MILLENIUM
LIDER
FOTINI
ERMIS
CELIA
CARMEN
BOLINA
R R R
Variety
Figure 5: Calculated calibration factors for every load and cotton variety y = 0.9771x R2 = 1
Estimated Load Weight(Kg)
2500
2000
CELIA CARMEN
1500
XRISTINA LIDER ERMIS SANDRA
1000
BOLINA FOTINI MILLENIUM 500 500
1000
1500
2000
2500
Actual Load Weight (Kg)
Figure 6: Comparison of the estimated weights to actual load weights by variety (CF=2858).
Next Figure 7 shows the mean value of the calculated calibration factor for every cotton variety using equation 5. The standard deviation range is shown in the same diagram. 3250.00
W
3000.00
W
3061.50
2990.00
W
2927.50
2872.84 W
2799.40
W
2750.00
W
2775.57
2760.00
2632.33
W
MILLENIUM
FOTINI
BOLINA
2488.75
SANDRA
XRISTINA
CARMEN
CELIA
W
ERMIS
2500.00
LIDER
Calibration Factor Value
W
Variety
Figure 7: The mean and standard deviation for the calibration factor according the corresponding cotton variety Conclusions These first results show that there is a strong correlation of the value of calibration factor to the cotton variety in stable picking conditions. Furthermore, some conclusions could be written: - Every variety has a corresponding range of calibration factor values, - It is possible to improve the estimation of the actual yield during the picking putting a prescribed calibration factor according the cultivated variety of every field at the beginning of harvest, - It is possible to characterize and classify a harvesting load in specific categories from the calibration factor, - It is possible to reduce the estimation error knowing the variety type, when it is impossible to make an actual weighing based calibration of the system.
- The value of the calibration factor gives a first indication about the density of every variety and it is of value to the producer, since he is paid for the weight of seed cotton, More research and observations are needed covering many different varieties and commercial yield monitors to establish a clear deductive methodology. More effort is necessary to eliminate and model the external factors that influence the picking conditions. References 1.
Auernhammer, H., (2001), Precision farming – the environmental challenge, Computers and Electronics in Agriculture, 30, pp. 31-43.
2.
Stafford J.V.. (2000), Implementing Precision Agriculture in the 21st Century, J. Agr. Engng Res.(76) 267-275
3.
Gemtos, T., Fountas, S., Blackmore, B.S., Greipentrog, H.W. (2002), Precision farming experience in Europe and the Greek potential, In: Proceedings of the 1st HAICTA Conference, Athens, 6-7 June 2002: 45-55.
4.
Gemtos, T.A., Markinos, Ath., Toulios, L., Pateras, D., Zerva, G. (2003), A Precision Farming Application in the Small Cotton Farms of Greece, Presented in ITAFE 2003 conference, Izmir, Turkey, 7-9 October 2003, pp.
5.
Markinos, Ath, Toulios, L., Pateras, D., Zerva, G., Gemtos, T. A., (2003), Poster in the 4th European Conference on Precision Agriculture, Berlin 16-18 June 2003, pp. 489-490.
6.
Markinos, A., Gemtos, T., Toulios, L., Pateras, D., Zerva, G. (2002), Yield Mapping of Cotton crop in Greece, In: Proceedings of the 1st HAICTA Conf., Athens, 6-7 June 2002: 56-62.
7.
Reyns, P., Missotten, B., Ramon, H., De Baerdemaeker, J., (2002), A Review of Combine Sensors for Precision Farming, Precision Agriculture, 3, pp. 169-182.
8.
Grisso, R.D., Jasa, P.J., Schroeder, M.A., Wilcox, J.C. (2002), Yield monitor accuracy: Successful farming magazine case study, Applied Engineering in Agriculture 18(2), pp. 147-151.
9.
Arslan, S., Colvin, T.S., (2002), Grain Yield Mapping: Yield Sensing, Yield Reconstruction, and Errors, Precision Agriculture, 3, pp. 135-154.
10. Searcy, S.W., Motz, D.S., Inayatullah, A., Goering, K.J., (1997), Cotton yield mapping. In: Dugger, P., Richter, D.A. (Eds.), Proceedings Beltwide Conference, 6-10 January 1997, New Orleans, LA. National Cotton Council of America, Memphis, TN, pp. 601-603. 11. Perry, C.D., Durrence, J.S., Thomas, D.L., Vellidis, G., Sobolik, C.J., Dzubak, A. (1998), Evaluation of commercial cotton yield monitors in Georgia field conditions, In. Proc. Fourth International Conference on Site-Specific Management for Agricultural Systems. Madison Wis., pp. 1227-1240. 12. Wilkerson, J.B., Moody, F.H., Hart, W.E. (2002), Implementation and field evaluation of a cotton yield monitor. Applied Engineering in Agriculture 18(2), pp. 153-159. 13. Vellidis, G.C., Perry, D., Rains, G.C., Thomas, D.L., Wells, N., Kvien, C.K. (2003), Simultaneous assessment of cotton yield monitors. Applied Engineering in Agriculture, Vol 19(3), pp. 259-272.
14. Wilkerson, J.B., Kirby, J.S., Hart, W.E., Womac, A.R., (1994), Real-time cotton flow sensor. ASAE Paper No. 941054. American Society of Agricultural Engineers, St. Joseph, MI. 15. Thomasson, J.A., Pennington, D.A., Pringle, H.C., Columbus, E.P., Thomson, S.J., Byler, R.K. (1999), Cotton mass flow measurements: Experiments with two optical devices. Applied Engineering in Agriculture 15(1), pp. 11-17. 16. Wallace, T.P., (1999), Small plot evaluation of an electro-optical cotton yield monitor. Computers and Electronics in Agriculture, 23, pp. 1-8. 17. Durrence, J.S., Perry, C.D., Vellidis, G., Thomas, D.L., Kvien, C.K. (1998), Evaluation of commercially-available cotton yield monitors in Georgia field conditions. ASAE Paper No. 98-3106, St. Joseph, Mich.: ASAE. 18. Wolak, F.J., Khalilian, A., Dodd, R.B., Han, Y.J., Keshkin, M., Lippert, R.M., Hair, W. (1999), Cotton yield monitor evaluation, South Carolina: Year 2. In Proc. Beltwide Cotton Conf., pp. 361-364. Orlando, Fla. 3-7 January. Memphis, Tenn.: National Cotton Council. 19. Wilkerson, J.B., Moody, F.H., Hart, W.E., Funk, P.A. (2001), Design and evaluation of a cotton flow rate sensor. Transactions of the ASAE 44(6), pp. 1415-1420.
