Sclerotinia sclerotiorum, one of the major diseases infecting oilseed rape leaves, has seriously affected crop yield and quality. In this study, an indoor unmanned aerial vehicle (UAV) low-altitude remote sensing simulation platform was built for disease detection. Thermal, multispectral and RGB images were acquired before and after being artificially inoculated with Sclerotinia sclerotiorum on oilseed rape leaves. New image registration and fusion methods based on scale-invariant feature transform (SIFT) were presented to construct a fused database using multi-model images. The changes of temperature distribution in different sections of infected areas were analyzed by processing thermal images, the maximum temperature difference (MTD) on a single leaf reached 1.7 degrees Celsius 24 h after infection. Four machine learning models were established using thermal images and fused images respectively, including support vector machine (SVM), random forest (RF), K-nearest neighbor (KNN) and naïve Bayes (NB). The results demonstrated that the classification accuracy was improved by 11.3% after image fusion, and the SVM model obtained a classification accuracy of 90.0% on the task of classifying disease severity. The overall results indicated the UAV low-altitude remote sensing simulation platform equipped with multi-sensors could be used to early detect Sclerotinia sclerotiorum on oilseed rape leaves.

Unmanned Aerial Vehicle (UAV) spray has been used for efficient and adaptive pesticide applications with its low costs. However, droplet drift is the main problem for UAV spray and will induce pesticide waste and safety concerns. Droplet size and deposition distribution are both highly related to droplet drift and spray effect, which are determined by the nozzle. Therefore, it is necessary to propose an evaluating method for a specific UAV spray nozzles. In this paper, four machine learning methods (REGRESS, least squares support vector machines (LS-SVM), extreme learning machine, and radial basis function neural network (RBFNN)) were applied for quantitatively evaluating one type of UAV spray nozzle (TEEJET XR110015VS), and the case of twin nozzles was investigated. The results showed REGRESS and LS-SVM are good candidates for droplet size evaluation with the coefficient of determination in the calibration set above 0.9 and root means square errors of the prediction set around 2 µm. RBFNN achieved the best performance for the evaluation of deposition distribution and showed its potential for determining the droplet size of overlapping area. Overall, this study proved the accuracy and efficiency of using the machine learning method for UAV spray nozzle evaluation. Additionally, the study demonstrated the feasibility of using machine learning model to predict the droplet size in the overlapping area of twin nozzles.

In the demand for renewable energy sources, algae are considered to have a good potential for biodiesel production. Fast detection of heavy metals in energetic algae is of vital importance for algal biomass and biodiesel production environment monitoring. This study aims to determine cadmium (Cd) and zinc (Zn) contents in algal pellets with laser-induced breakdown spectroscopy (LIBS) technology. Partial least squares (PLS) regression and extreme learning machine (ELM) were constructed after spectral preprocessing by standard normal variate transformation (SNV), multiplicative scatter correction (MSC) and Savitzky–Golay smoothing (SG). For univariate analysis, ELM models based on Cd II 226.45 nm and Zn II 206.19 nm of SNV preprocessed spectra achieved the best results for Cd and Zn content prediction respectively, with R_p, RMSEP and RPD values of 0.9866, 50.71 mg kg⁻¹ and 6.95 for the former, and 0.9873, 30.08 mg kg⁻¹ and 6.44 for the latter. Multivariate analysis of Cd based on the PLS model with global spectra achieved the best performance with R_c and R_p values of 0.9965 and 0.9972, RMSECV and RMSEP values of 25.57 mg kg⁻¹ and 23.63 mg kg⁻¹, and an RPD value as high as 13.27, showing the excellent robustness and effectiveness of the model for Cd detection. For Zn analysis, the best performance was achieved by the ELM model based on feature variables selected by regression coefficients, with R_c and R_p values of 0.9836 and 0.9920, RMSEC and RMSEP values of 34.25 mg kg⁻¹ and 24.64 mg kg⁻¹, and an RPD value of 7.87. The results indicated that the LIBS technique combined with appropriate preprocessing algorithms and multivariate chemometric methods could be a rapid and accurate way for quantitative analysis of Cd and Zn contents in algae and the aquatic environment.