Machine learning approaches for crop yield prediction and nitrogen status estimation in precision agriculture: A review

The on-line measurement of soil properties is essential for site-specific application of different inputs into agricultural soils. Using a previously developed soil sensor for on-line measurement of soil properties, primary results about carbon (C), moisture content (MC), pH and phosphorous (P) were reported. The on-line sensor consisted of a soil penetration unit (subsoiler), to which the optical probe to acquire soil spectra from the bottom of the trench opened by the subsoiler chisel is attached. A mobile, fibre-type, visible (VIS) and near infrared (NIR) spectrophotometer (Zeiss Corona 45 visnir fibre, Germany), with a measurement range of 306.5–1710.9 nm was used to measure soil spectra in reflectance mode. General calibration models were established under non-mobile laboratory conditions, on the basis of two sample sets collected from large geographical areas covering Belgium and northern France. These models were developed using partial least squares (PLS) regression coupled with the full cross-validation technique. On the basis of the values of coefficient of multiple determination ( R 2 ) and the ratio of standard deviation of calibration set (S.D.) to root mean square error of cross-validation (RMSECV), the prediction of MC was evaluated as good, whereas the predictions of C, pH and P were evaluated as only possible to provide quantitative approximations. When these models were validated using on-line measured spectra, they provided moderately similar maps to those measured with the reference methods. The best similarity was obtained for MC maps. Mean error values of 5.97, 0.37, 27.48 and 5.10% were found between the on-line and reference measurements of total carbon (C-tot), MC, pH and available phosphorous (P-avl), respectively, suggesting the potential use of the VIS–NIR sensing system for on-line measurement of soil properties.

On-line measurement of some selected soil properties using a VIS–NIR sensor

Current and future applications of statistical machine learning algorithms for agricultural machine vision systems

The evaluation of unmanned aerial system-based photogrammetry and terrestrial laser scanning to generate DEMs of agricultural watersheds

A B S T R A C T The yield of rainfed crops is commonly limited by the availability of soil water during the summer growing season. Channels produced by cover crop roots in fall/winter when soils are relatively moist may facilitate the penetration of compacted soils by subsequent crop roots in summer when soils are relatively dry and hard. Our objective was to determine the effects of fall cover crops on maize (Zea mays) growth and soil water status under three levels (high, medium, and no) of imposed traffic compaction. The study was conducted on coastal plain soils (fine-loamy Typic/Aquic hapludults and siliceous, Psammentic hapludults) in the mid-Atlantic region of the United States from 2006 to 2008. Cover crop treatments were FR (forage radish: Raphanus sativus var. longipinnatus, cv. ‘Daikon’), rapeseed (Brassica napus, cv. ‘Essex’), rye (cereal rye: Secale cereale L., cv. ‘Wheeler’) and NCC (no cover crop). Maize under high compaction achieved more deep-roots following FR and rapeseed than following rye or NCC. However, maize had greater yield following all cover crops than NCC control regardless of compaction levels and soil texture. Compaction reduced maize yield only under the high compaction in the lightly textured soils. During 24 June–24 July 2008, soils at 15 and 50 cm depths were drier under no compaction than high compaction and drier following FR than other cover crop treatments. Our results suggest that FR benefited maize root penetration in compacted soils while rye provided the best availability of surface soil water; rapeseed tended to provide both benefits. However, as rapeseed is relatively difficult to kill in spring, a mixture of FR and rye cover crops might be most practical and beneficial for rainfed summer crops under no-till systems in regions with cool to temperate, humid climates.

Root growth and yield of maize as affected by soil compaction and cover crops

Management zones based on correlation between soil compaction, yield and crop data

A field was investigated with precision farming techniques to delineate zones with different yield potential due to previous soil erosion. Winter wheat (Triticum aestivum L.) grain yield, straw yield, biomass and harvest index were measured with a combine harvester. Fuzzy clustering of grain and straw yield provided good delineation of zones in the field with different yield potential. Entropy and fuzziness calculations for different number of classes resulted in a division of the field into five clearly differentiated yield potential zones. Straw yield, biomass and harvest index were significantly different among zones. Grain yield was significantly different for all zones, except for two. Elevation and slope of the field were measured from a global positioning system (GPS) unit on the combine. Both were related to yield variability in the field. Average elevation, slope and soil type were calculated per cluster class. High grain yield, straw yield and biomass could be related to flat, high places in the field with little erosion. Good grain yield, low straw yield and high harvest index were found on relatively steep slopes subjected to erosion. High straw yield and low grain yield were found at low places in the field on relatively steep slopes. Lowest grain yield, straw yield and biomass were located on steepest slopes with high erosion and in depressions where accumulation of eroded soil took place and slumping and crusting of the soil were present. This information suggests that variable management on a site-specific basis would optimize yield and inputs.

Yield variability related to landscape properties of a loamy soil in central Belgium

An on-the-go monitoring algorithm for separation processes in combine harvesters.

An analytical grain flow model for a combine harvester, part I: Design of the model

The threshing section affects grain separation performance in combine harvesters Proper configuration is critical
to ensure an optimal economic working of combine harvesters Predicting grain separation online is essential for an
automatic machine tuning system 
An indirect process–monitoring device was designed based upon feedrate measurements Stationary tests provided
information about the behavior of grain separation during the threshing process as a function of different feedrates and crop
properties Six different theoretical threshing models were analyzed For three models, a comparative study has been carried
out to choose an optimal predictor of the separation process The most accurate prediction model is based on Rusanov’s
exponential model 
The confidence interval necessary to analyze the residual value of the predictor was calculated for each predicted
separation curve

K. Maertens

Papers

Management zones based on correlation between soil compaction, yield and crop data

Yield variability related to landscape properties of a loamy soil in central Belgium

An on-the-go monitoring algorithm for separation processes in combine harvesters.

An analytical grain flow model for a combine harvester, part I: Design of the model

Flow rate based prediction of threshing process in combine harvesters