Search

The concept of precision agriculture (PA) is becoming an interesting idea for managing natural resources and realizing modern sustainable agricultural development. The general objective of this study is to better understanding of the current use of PA practices in Florida, USA. Specific objectives were to identify PA trends, information source preferences and the constraints to wider use of PA technology. Online and in-person survey results were collected and analyzed to provide an overview of the adoption for specific PA technologies in Florida and compared grower-reported challenges of adoption and attitudes to those found in 2005. In 2018, the plant tissue sampling was the highest PA technic that grower used (81.6%) while in 2005, the highest rate was for using GPS receivers and Soil variability mapping (16.1%). In the meantime, growers had plan to use remote sensing as well as sensor based variable rate applicator (77.3 and 77.7%) in 2005 whilst the more demanding technic for grower in 2018 was yield mapping (27.27%). And the main reason for not adapting PA technology was vary based on the technics individually for both years. However, the most common response for not using PA in general referred to the producer satisfaction with their current production practices, for all of the investigated technologies in 2005 and 2018. The second most common reason for not adapting PA technology was lack of information in 2005 while it was lack of capital in 2018. Respondents were also questioned on their “adoption attitude”. In 2018, 37.29 % described themselves as the first one who like to try new technology and method while in 2005, 16.1% showed their interest to work with new method and technology. In both years, the best description of starting new technology was “normally wait to see other's success with new technologies and production methods” which were categorized at “coat-tailors”.

Yield mapping is the first step for site-specific crop management. Many yield mapping systems have been developed for grain crops. However, it remains a difficult task for tree crops. In this study an autonomous yield mapping system for citrus crops was developed. The system was designed to detect fruit and create yield maps at early stages so that farmers could manage the grove site specifically based on the maps. It consisted of two major sub-systems, an autonomous navigation system and an imaging system. Robot Operating System (ROS) was used for developing the autonomous navigation system on top of an unmanned ground vehicle. An inertial measurement unit (IMU), wheel encoders and a GPS were integrated using an extended Kalman filter to provide reliable navigation solutions. In the imaging system, a high-resolution visible camera was carried by the vehicle for image acquisition. All the video frames were associated with latitude and longitude coordinates automatically. Detection of fruit from the video frames utilized a VGG16 model, which was trained with Faster-RCNN. Fruit detection was evaluated and an accuracy of 77% was achieved. The Lucas-Kanade optical flow method was used for tracking each detected fruit and counting the total number of fruit. The complete system was tested in a citrus grove in Florida.

Yield forecasting is important for farm management. In this study, red, green, and blue (RGB), near-infrared (NIR), and depth sensors were implemented in an outdoor machine vision system to determine the number of immature citrus in tree canopies in a citrus grove. The main objective was to compare the performances of three image data types for citrus yield forecasting. The performance comparison was conducted with two machine vision algorithm steps: 1) circular object detection for potential fruit areas and 2) classification of citrus fruit from the background. For circular object detection, circular Hough transform was used in the RGB and NIR images. For the depth images, CHOI‘s Circle Estimation (‘CHOICE‘) algorithm was developed using depth divergence and vorticity to find circular objects in the depth images. The classification process was conducted using AlexNet, a deep learning algorithm for all three image types. The implementation of a convolutional neural network allowed the machine vision algorithms to remain bias-free process during feature generation and selection. NIR images performed best with 96% true positive rate for both the circular object detection and classification. A machine vision system using this image type will produce a more objective yield prediction with a higher accuracy than other types.

Forecasting the number of immature green citrus and its size prior to harvesting period would help the growers plan application of nutrients during the fruit maturing stages and estimate their yield and profit. Yield mapping by machine-aided harvesting requires automatic detection and the counting of fruit in a tree canopy. However, occlusion, varying illumination, and color similarity with the background make green citrus fruit identification a very challenging task. The overall goals of this research were to develop a real-time yield mapping system that could detect immature green citrus using color images with multiple features and to estimate yield in the complex outdoor environment. This study concentrated on estimating number of citrus fruit in images of tree canopy acquired under natural outdoor conditions. First, a linear color space, OHTA, was used to segment green citrus from the complex background. In order to overcome recognition challenges in uncontrolled environments caused by uneven illumination conditions, similar background features, and partial occlusion, multiple features such as color, shape and texture features of the different objects (immature green citrus, leaves, soil, sky, twigs and branches) were used for feature extraction to find differences between green citrus and background. Texture classification, based on two major types of texture features, gray gradient co-occurrence matrix and Tamura texture features, combined with a support vector machine (SVM) were used to remove non-citrus regions. After removing false positives, a final decision was made using shape features to detect individual citrus fruit. The proposed method was tested in a validation set of images obtained from the outdoor environment, detecting and counting over 82.4% of citrus fruit. The algorithm mentioned in this paper was promising in green citrus detection although improvement would be needed to decrease missed fruit and false positives.

Although the coffee harvest has been done mechanically since 1979, with theintroduction of the Jacto K-3 coffee harvester (Fig.1), the mapping of the spatialvariability of the coffee crops has not yet become a commercially available practice.An automatic yield measuring system was successfully developed and installed onsaid coffee harvester. The production data, at every moment related to the data of aDGPS receiver, were automatically processed by means of a software speciallydeveloped for this acquisition and storage. Yields mapping in liters per hectare weregenerated for two trial fields in two farms. The necessity of a transformation factoramong the different coffee maturation stages has shown itself very important and isanalyzed. Soil fertility mapping and aerial infrared photographs complement theinformation for future analysis. The accuracy, difficulties and future possibilities arepresented.

This research developed systems to measure mass-flow of hay and forage using a self-propelled windrower, self-propelled forage harvester and large square baler in order to generate yield maps of hay and forage production. The windrower was used to collect mass-flow and yield data in alfalfa using system to measure: (1) impact force at the swath forming shield; (2) crop volumetric flow past the swath forming shield; (3) conditioning roll speed; (4) platform pitch; and (5) pressure of platform drive motor. Some sensors systems were promising but even the most promising will require further evaluation because the average absolute mass-flow prediction error and 90th percentile error were 13.4 and 26.4%, respectively, above the desired maximum of 5 and 10%. The forage harvester was used to collect mass-flow and yield data in wilted alfalfa and wholeplant corn silage using systems to measure: (1) feedroll displacement; (2) crop impact force in the spout; (3) blower speed; and (4) depth of material in the spout. The accuracy of these sensors was much better in whole-plant corn silage than in wilted alfalfa, primarily because the former fed much more uniformly. The average absolute and 90th percentile errors were 12.3 and 24.2%, respectively, for wilted alfalfa, and 4.4 and 7.4%, respectively, for whole-plant corn silage. The large square baler was used to collect mass-flow and yield data in dry alfalfa using sensor systems to measure: (1) bale velocity; and (2) dynamic bale weight on the chute. The bale velocity and weight measurement systems were excellent at predicting mass-flow with typical absolute and 90th percentile errors of 1.4 and 3.2%, respectively. Multi-parameter regression models were developed to predict mass-flow for each of the three machines. These models were then used to create yield maps using the output from selected sensors. The resolution of the yield maps was a function of the width of the harvested strip. The resolution of the yield maps was the best with the windrower and forage harvester in whole- plant corn silage because these machines had the narrowest harvesting widths. The map resolution was less with the forage harvester harvesting wilted alfalfa and with the large square baler because multiple windrows were often merged together to meet the capacity needs of the machines. The developed yield maps were quite capable of showing spatial differences in yield that corresponded very well to differences in elevation, soil type, traffic patterns and fertility. The hay and forage mass-flow sensor systems and yield mapping systems should become a valuable precision farming tool.

Several studies have been discussing problems related to the quality of yield data information that comes from yield monitors. Errors re usually caused by GPS, yield sensor, calibration, moisture sensor, errors related to the combine dynamics and errors related to the measured area. This paper deals with raw data from different yield monitors. Each file was studied, identifying the possible errors. Once they were identified, the errors were classified and characterized. Yield monitors available in the market generate data that may be submitted to a filtering process to improve the quality of information. Fill time error was found in all beginnings an ends of rows. Its characteristic is a slow increase on yield measured as the combine is being filled up. Lag time error is a systematic displacement of the yield points lead by wrong time delay in the yield monitor and was found in all data. Wrong interpretation of the cutting width when using partial header results in error on measured area and incorrect yield. The header cut off error happens when the combine is not harvesting but area is being added because the system was not shut off. Failures on collecting data was found in several situations, when the monitor should be recording. Bad yield sensor calibration is a systematic error evidenced by a difference between monitor and the total weighed yield. Some errors occurred on specific monitors while others are present in all files. Filters are being proposed with specific routines for each equipment to eliminate errors.

The objective of this work was to present the Analysis of Correspondence as a multivariate statistical method that can help in the yield mapping comparison. There is a concern about its indiscriminate use as tool of predictive information, without taking into consideration the spatial and temporal variability of the yield. The idea that the use of yield mapping is strictly related to the variability of the components was verified. Four yield maps were used corresponding to collected yield samples for rice in 1994, corn in 1995, cotton in 1996 and corn in 1998, in a grid with 110 points, spaced of 10 x 10 m, at the experimental station of the Agronomic Institute, located at Votuporanga, SP, Brazil. The original data were transformed using its order and analyzed using Simple and Multiple Correspondence Analysis. A great dissimilarity between the variables or classes of variables was found as measured by the Euclidean Distance in the plan.

Miscanthus is emerging as a promising feedstock for domestic biofuel production. However, inefficiencies of machinery that are used for harvesting bioenergy crops such as miscanthus currently prohibit commercial production. The performance of a mower-conditioner used to harvest miscanthus was evaluated and modifications to the disk head were made to allow the machine to operate more efficiently. Previous work on cutting parameters for harvesting miscanthus concluded that the energy requirement could be reduced by changing blade oblique angles and varying cutting speeds. Results from studies involving a single disk cutter system were evaluated for use in determining the cutting parameters that should be implemented on the mower-conditioner. As a result, blades with oblique angles of 20° and 30° were manufactured and fitted to the mower-conditioner. The 30° oblique angle resulted in the greatest energy reduction with an energy consumption of 13.5 MJ Mg -1 as compared to 18.5 MJ Mg -1 for the conventional straight 0° blades. Further studies could reveal the effects of the replacement cost of the blades as compared to the savings achieved by using them.

Crop yield data is a key component of precision agriculture, critical for both development and evaluation of precision management strategies. Ideally, software that generates grain yield maps from raw yield monitor data should automatically correct common errors associated with machine and operating characteristics. Perhaps the most basic correction required is to properly compensate for the time lag (or position lag) between the cutting of the crop from the field and the grain flow measurement by the flow sensor in the combine. Past research has suggested several approaches to automatically determine delay time but for various reasons these have not been implemented in mapping software. In this paper we present a new, computationally efficient method that can accurately determine delay time for individual fields using the image processing method of phase correlation. This method was evaluated using a number of yield maps with varying degrees of harvest complexity and results were compared to a geostatistical method. The new method performed satisfactorily on larger datasets with a significant amount of spatial variability. Additionally, this method was more computationally efficient than previous methods. Results of this study will increase the feasibility of including automatic delay time compensation in yield mapping software.

Yield mapping as a prerequisite of “precision farming evolution” needs more both field and laboratory investigations for wide varied slope effect based crops combining as an important influenced factor. Naturally, optimised precision and accuracy of continuous crop yield metering on combine harvesters requires more detailed studies on points such the field-slope variations which affect not only on soil characteristics and in-field spatial crop yield but it is a serious problem to measure the reliable continuous spatial variable yield variations during combine harvesting. Author through his practical studies conducted at Nafferton research Farm of Newcastle University in UK found out very varied topographic aspects. Previous author’s experiences in both workshop and fields showed that the slope affects accurate yield measurement through combining crops (Sanaei, A. 2008) and then the research was extended to design and construct a test rig using clean grain elevator’s parts of Class Combine for more controlled fruitful trials by simulating varied slopes(pitch and roll) in workshop site. Today, there are only a few harvester combines with installed full package of yield meter kits equipped with a slope (Pitch and Roll) sensor as well. These studies showed that the accuracy of Ceres2 also may be improved by addition of an optional hillside slope sensor. It measures the angle of side-slopes and the instrument will correct the yield measurement for the effect of the slope sensor location. This paper examined and analysed data of multidimensional detailed wide range of slope effects between 0-15 degrees on measuring yield through accessing to algorithmic models by installing previous version of Ceres2 yield meter on constructed test rig which showed comprehensive results including significant differences on yield meter accuracy.

The performance of the new generation Ag Leader cotton yield monitoring system, based on the Insight display, was evaluated for yield accuracy and ease of use during the 2007 harvest season. A four-sensor Ag Leader Insight system was installed on a grower-owned John Deere 9996 cotton picker in south Georgia and was used to harvest his 25 fields. Basket load, module, and field total weights were compared to yield monitor weights to evaluate accuracy. The grower's experience with the system was used to evaluate ease of use. After proper initial weight calibration, the system under-predicted actual weight with a percent error of approximately 12% when comparing basket loads, modules or field totals. Nevertheless, the predicted weight values were fairly consistent and the resulting yield data produced yield maps that represented field variations quite well. Once re-calibrated later in the season, the system began over-predicting cotton weight and the percent error reduced to approximately 7%. Ease of use was very good with the grower able to navigate menus and settings easily.

Yield mapping technology is considerably well established for grain combines, but not as readily available other harvesting machinery; therefore, researchers have been developing systems to map crops other than grain. Moreover, cost effective alternatives of mapping systems are desired for lower cost harvesting machines such as the pull type dry edible bean harvester. Previous studies have presented low cost alternative for measuring torque on a drive chain driven shaft. This concept of drive torque measurement was implemented on a clean grain bucket elevator of a dry bean harvester, and evaluated as mass flow sensor. Tests were conducted in the Yield Monitor Test Facility (YMTF) of University of Kentucky following the recommendation of ASABE Standard S578. Device was tested within the flow rate range of 0 to 3.4 kg·s-1. At two standard deviation level, the deviation at 3.3 kg·s-1 was ± 4.2% and the deviation at 1.6 kg·s-1 was ± 4%. The average accumulated mass errors were less than 3.1% and the maximum accumulated error was 4.9% at a flow rate of 1.8 kg·s-1.

The objective of this research was to develop a non-invasive yield mapping system for sugar beets by applying ultra wideband (UWB) radar technology. Similar to radars used in geological engineering, these devices acquire data via electromagnetic waves. The basic idea was to use such data to identify sugar beets in agricultural soils. The initial research task was to distinguish sugar beets from the surrounding soil.

A complete instrumentation system on-board a New Holland TC 56 rice combine harvester has been successfully integrated for mapping crop yield, grain losses, grain properties and combine field performances at sub meter accuracy. Earlier, laboratory calibration tests on all the sensors showed excellent measurement linearity (correlation coefficients closed to 1.0000). Field calibration work has been successfully carried out to the combine actual speed, grain flow, and grain moisture and temperature sensors of the system for the purpose of improving their measurement accuracies. A generalised time delay equation for the grain flow, grain moisture and temperature was formulated based on the combine average actual speed for the instrumented combine to take care of the time delay in the passage of grain to these sensors. Further work is need to explore the possibility of improving the measurement sensitivity of the present grain loss sensors or identification for other sensor types that could to measure accurately the actual magnitude of grain losses from the instrumented combine. Maps of crop yield, grain properties and combine field performances could be easily developed using ArcGis 8.3 software from the collected field data by the instrumented combine. Multiple regression analysis were conducted to develop the mathematical relationship of instantaneous harvested crop with soil, grain, and combine attributes and combine fuel consumption with grain and combine attributes. Correlation analysis was also conducted between potential crop yield from the Crop Cut Test (CCT) and harvested crop yield recorded from the instrumented combine.

Non-contacting methods for material properties determination and crop yield determination are the best solution especially for rough conditions in agriculture, however, it requires reliable and durable sensors as a source of primary data. The capacitance sensor based technique can be used for forage material mass determination in order to utilize obtained data for yield maps creation.

A machine vision system consisting of two CCD color cameras, a DGPS receiver, an encoder and ultrasonic sensors was developed to obtain citrus yield information of a citrus grove. A total of 512 images from 32 trees were taken during static operation from a citrus grove located in the Winter Garden, Florida. A fruit counting algorithm was developed in this research to estimate citrus yield. Noise removal and illumination adjustment were conducted. The color segmentation and marker-controlled Watershed algorithm were developed to segment and split touching fruits. The algorithm was tested on 176 images and the number of fruits counted by the algorithm was compared with the number of fruits counted by human observation. The coefficient of determination, R2, was 0.83. Moreover, the number of fruits counted by the algorithm and actual harvested number of boxes was compared. The coefficient of determination was 0.64. An average fruit size was also determined from each image using ultrasonic measurements.

A remotely-piloted Gas Xcell mini-helicopter with GPS, wireless Ethernet, and a machinevision system was flown over orange trees to take overhead images. The individual tree images were processed to estimate the number of orange pixels in the images. The fruit yield of individualtrees at harvest a month later had an r-squared value of 0.373 with the number of orange pixels onthe irrigated tree images. While it was demonstrated that a remotely-piloted mini-helicopter could beused to acquire useful overhead images of citrus trees, substantial further research is needed toimprove the accuracy of estimates of individual tree yields or health.

A John Deere 9860 STS combine equipped with an 11 m (36 ft) wide draper header and GreenStar™ yield mapping system was used to evaluate the impact of automatic feedrate and steering control systems on combine harvesting efficiency while operating in soft white wheat. Five operators were used to evaluate the feedrate control system and the combination of feedrate and steering control systems. A benchmark for each operator was established while he had complete control of the combine. Data were gathered with the Greenstar™ yield mapping system and DGPS receiver during harvest. Harvest efficiency was evaluated based on improvement in material capacity and ground speed and the variability of those two items. Though there were some differences among operators, the automatic feedrate control system resulted in an average material capacity increase of 1.7 Mg/h (61 bu/h {5.3%}). Using the automatic steering control in conjunction with the automatic feedrate control resulted in a slight (0.2 Mg/h {9 bu/h}) average decrease in material capacity relative to using the automatic feedrate only. The slight average decrease may be a result of low operating speeds at which the operator may do a better job of using the full header width. The automatic feedrate control system resulted in more ground speed and feedrate variation, as measured by the standard deviation and interquartile range, compared to the standard operator.

Pasture management is one of the most important factors in developing a successful dairy production system in New Zealand. Key performance indicators include the quantity of pasture grown and how well it is utilized. For a pasture management system to be successful, it requires accurate pasture measurement and the ability to use the collected information to make decisions at all management levels. Industry commentators often talk about a ten per cent improvement in production being possible from improved utilization of pasture. On a national scale a 10% improvement would result in an increased return of NZ$560M. (US$400M) to dairy farmers. Improvement in pasture production and utilization has been hampered by the lack of tools to accurately measure pasture production in a way that is time efficient and cost effective. This needs to be managed so that demand can be met and excess conserved. This paper describes work which provides an ATV mounted yield mapping capability for dairy farmers. The most common form of pasture measurement previously used has been the rising plate meter (RPM). During the testing phase it was found there was a strong linear relationship between the two measurement instruments (R2 = 0.955) in terms of predicting pasture dry matter (kg DM ha-1), when exactly the same points were used. When full paddock measurements were undertaken by taking measurements at a ten metre swath width using the new high speed sensor there was found to be a significant difference between the predicted dry matter readings and those gained by a single transect of the plate meter. Indicating that the transect was not necessarily a representative sample.

Yield mapping techniques are beginning to cross over into orchard crop production. Perennial orchard/tree crops are high dollar commodities that can greatly benefit from precision agriculture applications. Tree crops introduce new challenges with creating yield maps as well as any management map because accurate positional data cannot be obtained by the use of Global Positioning Systems (GPS) when a clear view of the sky does not exist under the tree canopy. External to the canopy a GPS can yield accurate positional data whereas underneath the canopy it cannot. This research focuses on the development of a system that can effectively detect every tree inside a given field/orchard inspired by the interpoint method used in forestry. An encoder based system and a DGPS based system collected data concurrently in an open canopy orchard and a dense canopy orchard. The accuracy of each system was analyzed using a “stress value”. The encoder based system had an average stress value of 2.3 in the open canopy and 0.9 in the dense canopy orchards. The DGPS based system had an average stress value of 138.3 in the open canopy and 128.8 in the dense canopy orchards. The encoder based system in association with the snap method successfully detected all trees in both orchard configurations whereas the DGPS system did not.

Up to date, the use of precision agriculture concepts on sugar cane is limited due to the difficulty in detecting the spatial yield variability and the effects of site specific input applications. A yield monitor, produced by Austoft® (CNH) and installed on a sugar cane harvester was evaluated during the 2002 season under field conditions. The main objectives of this work were to characterize the performance of the yield monitor, evaluate its accuracy under different setups for yield calculation and evaluate the use of a non differential correction positioning system on producing yield maps. Field testing limitations did not allow for more conclusive information regarding the accuracy of the load sensor. Even so it was possible to take some good indications, especially related to yield mapping with maps showing consistent information on yield variability in the field. The test with the use of non DGPS positioning system indicated that it is possible to use it on positioning the yield points. The difference between maps produced with GPS and DGPS did not justify the use of DGPS when considering the cost of the differential correction under local conditions.

Currently the sugar cane fields are burned as a pre-harvest practice. If such practice is abandoned, the trash could be collected and used for vapor production for electric energy generation in the sugar mills and distilleries. The objective of this study was to map the remaining trash of the mechanically harvested raw sugar cane (not burned) as a result from bailing the material with two different balers (in cylindrical and prismatic bales), using Precision Agriculture concepts. This activity allowed to measure the sugar cane trash mass bailed, the moisture content and the energetic equivalent, in petroleum barrels. The studied area belongs to Costa Pinto Sugar Mill, in Piracicaba, Sao Paulo, Brazil (22o40’30”S, 47o36’38”W). Determinations were carried out with different bailing process that were made from three different raking process. The average results for mass and volume were 336.4kg with an 1.54m3 for prismatic bales and 452.9kg and 2.72m3 for cylindrical bales. Using the obtained maps it was observed that the sugar cane bailed trash yield on the area varied from 6.0 to 15.0Mg.ha-1, the moisture content varied from 14.68 up to 30.50% and the energetic equivalent was estimated from 6.33 up to 17.80 petroleum barrels per hectare, in non homogeneous distribution.

This research was conducted to develop an image processing algorithm to identify and count the number of citrus fruits in an image. Once this algorithm is completed it will be incorporated into a machine vision system consisting of a GPS receiver and distance measuring devices in a pick-up truck to estimate yield of a citrus grove on-the-go. A total of 90 images were acquired in an experimental citrus grove. Images of the citrus grove were analyzed and histogram & pixel distribution of various classes (citrus, leaf, and background) were developed. The threshold of segmentation of the images to recognize citrus fruits was estimated from the pixel distribution of hue and saturation color plane. A computer vision algorithm was developed to enhance and extract information from the images. Preprocessing steps for removing noise and identifying properly the number of citrus fruits were carried out using a combination of erosion and dilation. Finally the number of fruits was counted using blob analysis. The total time for processing an image was 283 ms excluding image acquisition time. The algorithm was tested on 59 validation images and the R2 value between the number of fruits counted by the machine vision algorithm and the average number of fruits by manual counting was 0.76.

Research was conducted to develop a device/procedure to measure mass flow into a round baler. Development of this technology is of interest to industry because it would provide the means of yield mapping of a forage field. By using this method, there is an opportunity to increase the capacity (ton/h) of the baler by increasing the average flow of material into the bale chamber. This objective is long-term as compared to the short-term objective of providing the means for yield mapping. The results demonstrated that a dynamic measure of windrow volume is possible. This measurement may be a useful surrogate for mass flow.

GPS data obtained during yield mapping operations for 23 fields were used to glean machinery management information. Yield monitor data were exported into the Ag Leader advanced format and GPS time, logging interval, mass flow, and distance were used to help assess combine performance. Field efficiency and capacity were determined. Time required to harvest, turn, and unload the crop were estimated. The relationship between efficiency and crop yield or unloading time was weak. Turning and efficiency were moderately correlated. Optimizing field traffic patterns to minimize turns appeared more opportunity for increasing field efficiency than unloading on the go.

As hyperspectral imagery is becoming more available, it is necessary to evaluate its potential for crop monitoring and precision agriculture applications. In this study airborne hyperspectral imagery was examined for mapping grain sorghum yield variability as compared with yield monitor data. Hyperspectral images were acquired using a CCD camera-based hyperspectral imaging system from two grain sorghum fields during the 2000 growing season, and yield data were also collected from the fields using a yield monitor. The hyperspectral images contained 128 bands covering a spectral range from 457.2 to 921.7 nm with a band width of 3.63 nm. The images had a swath width of 640 pixels and each pixel had a gray level between 0 and 4095. The raw hyperspectral images were corrected for the geometric distortion caused by the motion of the aircraft flying the imaging system. The corrected images were rectified to the UTM coordinate system with 1 m resolution and the raw digital numbers were converted to reflectance. The calibrated image data were then aggregated into images with a cell size of 9 m, close to the combine.s effective cutting width. Correlation analysis showed that grain yield was significantly related to the image data for all the bands except for a few in the transitional range from the red to the near-infrared region. Principal components analysis indicated that the first few principal components of the hyperspectral images accounted for 99% of variability in the image data. Regression analysis based on the principal components effectively quantified the amount of yield variability explained by image data. Stepwise regression analysis performed directly on the yield and hyperspectral data identified the optimum bands and band combinations for mapping yield variability, though the optimum bands differed between the two fields.

The starting point for precision farming is the yield map. A yield sensing system is specific for the equipment and harvest methodology employed in a crop. In this work, a yield sensing system for broccoli was developed for the broccoli harvesting method used when a harvesting aid conveyor and a trailer with containers following the conveyor are used. The harvesting aid conveyor was modified by adding two load cells that carry an idler roller which is under the belt of the conveyor, a speed sensor, a travel distance meter and a GPS with DGPS capabilities. The load cells signals, a sensor speed signal representing the belt speed and time were integrated to get weight of broccoli harvested in a defined period of time. The harvest equipment traveled distance multiplied by the width harvested allowed the system to get harvested area for the same period of time. Then weight and area were used to calculate yield data which were combined with GPS position for making the yield maps. The yield map done by the system was compared to a yield map done with weight data obtained in a separate bascule for each sampled area. For an area harvested of 0.39 ha the weight of broccoli obtained was 1,771 kg and the total weight obtained by the system was 1,778 kg so the error was only 0.4%. A visual comparison of maps shows a good correspondence between them.

An automatic weighing system was developed and interfaced to a GPS for use during mechanical coffee harvesting. The canvas hopper and load cells system successfully measured crop yield. The obtained results showed a great spatial variability in coffee yield, ranging from 2.1 to 16.0 Mg/ha in a 1.1 hectare plot in a farm in Sao Paulo state.

For the study, two “low-cost” handheld GPS receivers were compared to a DGPS receiver. A combine equipped with a yield monitoring system was configured so that yield data could be collected simultaneously using a “low-cost” handheld GPS receiver and a DGPS receiver. Results in 2000 showed average relative positioning differences of 1.78 meters and that similar yield maps could be created. Due to the limited hectares harvested in 2000 further analysis was desired. During the 2001 fall harvest, 212.2 hectares of corn and 283.7 hectares of soybeans were harvested using the above mentioned yield monitor configuration. In 2001, average relative positioning differences of 2.97 meters was obtained between GPS receivers. Yield maps were produced from the same field using standard yield mapping techniques. Comparisons were made between the maps of how similar they were at showing variability and yield in Mg/ha. Visually one could distinguish yield data collected between the two “low-cost” GPS receivers. Using regression analysis for map comparison, R-squared values ranged from 0.3 to 0.9 with a mean of 0.73.

This research was conducted to develop a silage yield mapping system using DGPS, load cells, Bluetooth modules for wireless data transmission, and a moisture sensor. A silage yield mapping system was successfully developed and tested in a silage corn field. The system yielded an error in the range of 0.37 ~ 1.96% of total harvested crop, compared with measurements by a platform scale. A Bluetooth module (wireless transmission) was successfully implemented to transfer moisture sensor information to a host computer. A silage corn yield map was created for a site-specific crop management.

A study was set forth to determine if lower cost GPS receivers are suitable for precision farming practices such as yield mapping. A combine equipped with a yield monitoring system was configured so that yield data could be collected simultaneously using a “lower cost” GPS receiver and a DGPS receiver. Yield maps were then produced from the same field using standard yield mapping techniques. Comparisons were then made between the maps. One comparison made was analyzing the positioning differences between the two receivers. Results showed average relative positioning differences of 1.78 meters. The other comparison made was how similar the yield maps were at showing variability and yield. Visually one could not tell any difference between the maps. Using the yield maps to create management zones, similarity between the maps ranged from 66.4% to 79.7% of being the same.

It was investigated whether the use of high-frequency electromagnetic fields allows a satisfactory estimation of the dry matter (DM) content of chopped maize (Zea mays). For this purpose, a power of 50 W was applied to maize with a known DM content in a radio frequency (RF) unit. Subsequently, it was determined by regression analysis to which extent the DM content can be explained by the measurands of the RF unit. Under laboratory conditions, the DM content of maize of different origins can be estimated significantly with the aid of this RF unit. A satisfactory estimation without knowing the sample mass and its density does not seem realizable with the method described. In contrast to this, a good fit (R 2 adj.) of the regression equation could be attained if the mass of water was used as a dependent variable. Under these conditions it also appears possible to estimate the moisture content of unknown batches.

An optically-based cotton flow rate measurement system was developed for site-specific yield assessment applications. The final design was field tested in Tennessee during three harvest seasons. Performance was evaluated by harvesting loads of cotton, integrating flow rate measurements over time to obtain predicted total mass for each load, and comparing the predicted values with actual mass measurements acquired with scales. Integrated flow rate measurements consistently predicted actual load masses with a mean error magnitude of 4.0 percent for all of the loads harvested during the tests. Performance degradation due to debris accumulation on sensor faces was not observed, even though the sensors were never cleaned during any of the harvest seasons. A subset of test data was analyzed to determine how changing field conditions affected system performance. Average cotton flow rate, cotton moisture content, and the amount of time elapsed since last system calibration had no significant effect on load prediction errors, however cotton variety did significantly affect system performance. Results suggest that the system calibration should be checked when the variety being harvested changes, as an additional calibration may be necessary for optimum results.

It was investigated whether the use of high-frequency electromagnetic fields allows a satisfactory estimation of the dry matter (DM) content of grass. For this purpose, a power of 50 or 250 W was applied to grass with a known DM content in a radio frequency (RF) unit. Subsequently, it was determined by regression analysis to which extent the DM content can be explained by the measurands of the RF unit. Under laboratory conditions, the DM content of grass of different origins and harvesting times can be estimated with the aid of this RF unit. Increasing the sample density improved the accuracy of the estimation. Whereas increasing the applied power did not cause further improvement. In the most favorable case, R2 adj. amounted to 0.89. A satisfactory estimation of the DM content [%] without knowing the sample mass and its density does not seem possible with the method described. In contrast to this, the water mass [g] could be estimated quite precisely (R2 adj. = 0.94).