The most essential component of precision farming is the yield monitor, a sensor or group of sensors installed on harvesting equipment that dynamically measure spatial yield variability. Yield maps, which are produced using data fi-om yield monitors, are extremely useful in providing the farmer a color-coded visual image clearly showing the variability of yield across a field. University of Georgia scientists recently completed development work on PYMS, the Peanut Yield Monitoring System. PYMS uses load cells for instantaneous load measurements of harvested peanuts and has proven to be accurate to between 2% and 3% on a trailer-load basis and to approximately 1% on afield basis when using data collected during combine operation. PYMS data are accurate to around 1% on a basket-load basis when using data collected under static conditions. The instantaneous accuracy of PYMS was calculated to be 700 kg/ha. Basing management decisions on the yield of individual pixels of PYMS yield maps is not realistic. The strength of PYMS is in differentiating yield trends and evaluating management practices. The system was extensively and successfully field-tested over a 3-year period and evaluated by 11 users during 1999, all of whom were able to use the resulting yield maps to evaluate current management practices or to develop future management plans. The University of Georgia has submitted a patent application for PYMS, and the technology has been licensed.

Precision farming describes the process of measuring and mapping land crop characteristics and then using these measurements to develop precise and intelligent application strategies that improve overall farm production. Yield monitoring is the phase of precision farming in which the crop yield variation within a field is measured and mapped. Yield maps from previous seasons can be used to determine the needed inputs in the field, whereas post harvest yield maps can be used to evaluate the implemented methods and make adjustments for the next season. Grain yield monitors are available, however there are few if any monitors for other crops. This paper examines the use of strain gauge load cells in a yield monitor for peanut combines and the development of methods for minimizing measurement noises thereby increasing reliability

During the 1997 harvest season, a peanut combine equipped with a load cell yield monitoring system was used to harvest 37.6 hal (119 ac) of irrigated land in Georgia Yield data from two fields are presented to demonstrate the yield monitor's ability to characterize spatial yield variation as well as quantify cumulative yield. Data recorded from individual wagon loads averaged less than 5% error, and whole field errors were <1%. Analog and digital signal conditioning methods used in the combine instrumentation are discussed. Practical techniques for correcting erroneous data are also described. Interpolation methods used for map creation are identified and justified according to the spatial characteristics of the collected data. Improvements in the yield monitoring system over previous systems are discussed along with the system's limitations and suitability for commercial use.

During the development of a peanut yield monitoring system, experiments were conducted on a two-rowpeanut combine to determine the duration of time lag between pickup and yield measurement, and to characterize theconvolution of peanut flow within the combine. The research indicates that the two-row peanut combine used in theexperiment subjects harvested product to significant convolution. A simple time lag correction will not recover the sitespecific (short term accuracy) of yield measurements. The distance and time period required to achieve a yield estimateerror less than 20% (95% confidence) is greater than 19.7 m (17 s) for simple time lag correction while it is 5.8 m (5 s)for deconvoluted data. The net result is that smaller regions of yield variability may be recognized with greater confidenceusing the deconvolution method than with the simple time delay method.

A prototype peanut yield monitoring system based on load cell transducers was evaluated for use in precision farming applications. Noise characteristics under simulated field conditions were examined, and the effect of mixing within the peanut combine during harvest was also investigated. Evaluation results showed that the system has potential for providing limited quality site-specific yield measurements for yield mapping applications.

Two commercial cotton yield monitoring systems (YMS) were marketed in the Fall of 1997. The systems were purchased and installed on three cotton pickers in south Georgia. Installation of both YMS involved much time, labor, and specialized tools. Field-scale accuracies of the systems were investigated by using the systems to map yields on three fields (113, 19, and 7 ha) representing extremes in cotton production levels in Georgia. Results indicated one system's accuracies were within 3% of actual yield while the other system exhibited a considerably larger error. Instantaneous accuracies of the two systems were researched by harvesting small plots of varying cotton yield levels. The errors for both systems over 33 m (100 ft) plots were large but were not representative of actual system performance. The growers involved in the investigations were impressed that commercial yield monitors were available, but the YMS overall performance was not sufficiently rigorous or accurate to encourage their purchase. Improved versions of both systems are needed before widespread use will begin in Georgia.

A yield monitoring system, which uses load cells mounted below the basket, was developed for application to existing and new peanut combines. A tank weighing assembly (free movement in both lateral directions) mount was used with double ended shear beam load cells in a 90 degrees configuration to create the most flexibility, with least opportunity for binding at the mounts. Slopes less than 2 degrees did not appreciably affect the load cell response. Moving average and regression methods were used to translate the continuous loading of the basket to incremental loading rates. The system was capable of reproducing input loading rates to within a 1.5 to 7.3% error over an assumed ground area of 20 m(2) to 10 m(2) (220 ft(2) to 110 ft(2)), respectively (under static operating conditions). The larger the area (number of values) included in the average, the smaller the error, but the less sensitivity to changing yield conditions. Additional integration with position signals, improved filtering, and some method of adjusting for pod moisture and trash are necessary to fully develop the system for commercial application.