Effects of the Spatial Resolution of UAV Images on the Prediction and Transferability of Nitrogen Content Model for Winter Wheat

Abstract

UAV imaging provides an efficient and non-destructive tool for characterizing farm information, but the quality of the UAV model is often affected by the image’s spatial resolution. In this paper, the predictability of models established using UAV multispectral images with different spatial resolutions for nitrogen content of winter wheat was evaluated during the critical growth stages of winter wheat over the period 2021–2022. Feature selection based on UAV image reflectance, vegetation indices, and texture was conducted using the competitive adaptive reweighted sampling, and the random forest machine learning method was used to construct the prediction model with the optimized features. Results showed that model performance increased with decreasing image spatial resolution with a R2, a RMSE, a MAE and a RPD of 0.84, 4.57 g m−2, 2.50 g m−2 and 2.34 (0.01 m spatial resolution image), 0.86, 4.15 g m−2, 2.82 g m−2 and 2.65 (0.02 m), and 0.92, 3.17 g m−2, 2.45 g m−2 and 2.86 (0.05 m), respectively. Further, the transferability of models differed when applied to images with coarser (upscaling) or finer (downscaling) resolutions. For upscaling, the model established with the 0.01 m images had a R2 of 0.84 and 0.89 when applied to images with 0.02 m and 0.05 m resolutions, respectively. For downscaling, the model established with the 0.05 m image features had a R2 of 0.86 and 0.83 when applied to images of 0.01 m and 0.02 m resolutions. Though the image spatial resolution affects image texture features more than the spectral features and the effects of image spatial resolution on model performance and transferability decrease with increasing plant wetness under irrigation treatment, it can be concluded that all the UAV images acquired in this study with different resolutions could achieve good predictions and transferability of the nitrogen content of winter wheat plants.

Predictive modeling of nitrogen content in winter wheat plants based on LASSO feature screening and UAV imagery

“…Through reviewing related literature [7,8,[32][33][34][35][36], 20 commonly used vegetation indices such as NDVI, GNDVI, MSR, etc. were selected for this study.…”

“…Compared with traditional destructive sampling and monitoring methods, remote sensing technology has been recognized as an effective means for rapid, dynamic, and non-destructive monitoring of farmland environment and crop growth conditions [3][4]. Commonly used remote sensing techniques include satellite, UAV, and proximal remote sensing [5][6][7]. Among them, UAV remote sensing has irreplaceable advantages of low cost, being able to obtain high spatial resolution images, being suitable for field-scale observation [7], and is an important technical tool for precision management of smart agriculture in the era of Agriculture 4.0.…”

“…Commonly used remote sensing techniques include satellite, UAV, and proximal remote sensing [5][6][7]. Among them, UAV remote sensing has irreplaceable advantages of low cost, being able to obtain high spatial resolution images, being suitable for field-scale observation [7], and is an important technical tool for precision management of smart agriculture in the era of Agriculture 4.0.…”

“…Currently, researches on crop nitrogen monitoring based on UAV remote sensing mainly utilizes spectral reflectance, vegetation index, and texture features to construct a prediction model for nitrogen prediction or inversion [4,[7][8]. For example, the relationship between vegetation indices such as NDVI and RENDVI and nitrogen nutrition index (NNI) extracted from a UAV-mounted Mini-MCA 6 multispectral camera was used to diagnose the nitrogen status of winter wheat during the key growth stages, with a model R 2 ranging from 0.49 to 0.73 [9]; Li et al [10] estimated winter wheat canopy nitrogen concentration by constructing spectral indices such as MSR based on DJI S1000 UAV equipped with UHD185 hyperspectral imager using a look-up table method, with a model R 2 of 0.83.…”

“…To improve the monitoring accuracy, since researchers revealed the physical mechanism between spectral signals and the leaf chemical components in the 1960s [14][15][16], there have been more and more algorithms used to decipher the relationship between the plant nitrogen content and the spectral information in an attempt to estimate the plant nitrogen content more precisely [7,10,17]. Traditional methods such as linear regression, polynomial regression, and partial least squares regression have been widely used in studies of spectral prediction or inversion of crop nitrogen content due to their simple principles and ease of operation [18].…”