|ZHANG, M. - Jiangsu Academy Agricultural Sciences|
|ZHOU, J. - University Of Missouri|
|Sudduth, Kenneth - Ken|
Submitted to: Biosystems Engineering
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 11/1/2019
Publication Date: 12/5/2019
Citation: Zhang, M., Zhou, J., Sudduth, K.A., Kitchen, N.R. 2019. Estimation of maize yield and effects of variable-rate nitrogen application using UAV-based RGB imagery. Biosystems Engineering. 189:24-35. https://doi.org/10.1016/j.biosystemseng.2019.11.001.
Interpretive Summary: Remote sensing can provide an efficient way to obtain information about differences in how a crop is growing across fields and landscapes. Recent availability of unmanned aerial vehicle (UAV) technology has created interest in developing new remote sensing applications for use in precision agriculture. Within-season measurement of the health of the crop and estimation of crop yield using UAV images would allow farmers to make management and marketing decisions before harvest, potentially improving their profitability. In this project, an index of corn color, as measured by a UAV, was investigated as an estimator of corn yield and the effects of variable nitrogen application in a research field in Central Missouri. Results were promising, especially when images were obtained later in the growing season. This study demonstrated the potential for using UAV-based imaging to estimate within-field yield variation in precision agriculture. This approach may be useful to researchers and to farmers who are interested in obtaining within-season yield estimations at relatively low cost and high resolution.
Technical Abstract: Accurate crop yield estimation is important for agronomic and economic decision-making. This study examined the application of visual image data acquired with a UAV-based remote sensing system for estimating maize (Zea mays L.) yield and the effects of variable-rate nitrogen (N) application. Images of a 27-ha maize field were captured using a digital camera mounted on a UAV flying at ~100 m above ground level at three maize growth stages (around R2, R3 and R6). The collected sequential images were stitched and the Excess Green (ExG) color feature was extracted to develop predication models for maize yield and to examine the effects of variable-rate N application. Various linear regression models between ExG and maize yield were developed for three sample area sizes (21, 106, and 1058 m2). The model performance was evaluated using coefficient of determination (R2), F-test and the mean absolute percentage error (MAPE) between estimated and actual yield. All linear regression models between ExG and yield were significant (p=0.05). The MAPE ranged from 6.2 to 15.1% at the three sample sizes, although R2 values were all < 0.5. From ANOVA, both crop growth stage and area scale affected model accuracy. Prediction error was lower at later growth stages, as the crop approached maturity, and at the largest sample level. The analysis of variable-rate N application showed that the crop grew better and more uniformly. Overall, the ExG color feature from a low-cost camera was found effective for predicting maize yield and evaluating the effect of variable-rate N application on plant growth.