Modeling of Diurnal Changing Patterns in Airborne Crop Remote Sensing Images

Abstract

Airborne remote sensing technologies have been widely applied in field crop phenotyping. However, the quality of current remote sensing data suffers from significant diurnal variances. The severity of the diurnal issue has been reported in various plant phenotyping studies over the last four decades, but there are limited studies on the modeling of the diurnal changing patterns that allow people to precisely predict the level of diurnal impacts. In order to comprehensively investigate the diurnal variability, it is necessary to collect time series field images with very high sampling frequencies, which has been difficult. In 2019, Purdue agricultural (Ag) engineers deployed their first field visible to near infrared (VNIR) hyperspectral gantry platform, which is capable of repetitively imaging the same field plots every 2.5 min. A total of 8631 hyperspectral images of the same field were collected for two genotypes of corn plants from the vegetative stage V4 to the reproductive stage R1 in the 2019 growing season. The analysis of these images showed that although the diurnal variability is very significant for almost all the image-derived phenotyping features, the diurnal changes follow stable patterns. This makes it possible to predict the imaging drifts by modeling the changing patterns. This paper reports detailed diurnal changing patterns for several selected plant phenotyping features such as Normalized Difference Vegetation Index (NDVI), Relative Water Content (RWC), and single spectrum bands. For example, NDVI showed a repeatable V-shaped diurnal pattern, which linearly drops by 0.012 per hour before the highest sun angle and increases thereafter by 0.010 per hour. The different diurnal changing patterns in different nitrogen stress treatments, genotypes and leaf stages were also compared and discussed. With the modeling results of this work, Ag remote sensing users will be able to more precisely estimate the deviation/change of crop feature predictions caused by the specific imaging time of the day. This will help people to more confidently decide on the acceptable imaging time window during a day. It can also be used to calibrate/compensate the remote sensing result against the time effect.

Elimination of Leaf Angle Impacts on Plant Reflectance Spectra Using Fusion of Hyperspectral Images and 3D Point Clouds

“…We have built calibration models for the other sources of variability. Two examples can be found in the published papers [ 33 , 34 ]. In this paper, we are focused on only calibrating the slope factor, and hopefully by combining this with the other previous models, we will keep improving the remote sensing quality.…”

Comparing Machine Learning Algorithms for Estimating the Maize Crop Water Stress Index (CWSI) Using UAV-Acquired Remotely Sensed Data in Smallholder Croplands

“…[48] established that the RF algorithm was best for estimating maize's specific leaf area, equivalent water thickness, and fuel moisture content to rRMSEs of 3.48%, 3.13%, and 1%, respectively. RF has been applied to predict the CWSI for crops other than maize [49][50][51]. For instance, Ref.…”

“…[50] observed that RF outperformed SVM in estimating chlorophyll content, with average RMSEs of 2.90 and 3.11, respectively. However, SVM has also shown promising results in predicting relative water content, achieving an R 2 of 0.72 and an RMSE of 6.22% [49]. Nonetheless, the literature also shows that, despite the optimum performance of these machine learning algorithms, no algorithm has been exhaustively tested to enable the accurate and effective identification and mapping of plant characteristics in a variety of environments [52,53].…”

LeafSpec-Dicot: An Accurate and Portable Hyperspectral Imaging Device for Dicot Leaves

“…Depending on the time of the day and the location of the field, the sunlight intensity and angle can change drastically [ 27 ]. The normalized difference vegetation index (NDVI) of the plant follows a V-shaped pattern where the minimum is at solar noon time when the sun is at the highest location in the sky [ 28 ]. Additionally, a field hyperspectral camera can only achieve a 10 mm spatial resolution after orthorectification [ 29 ].…”