Machine Learning-Based Approaches for Predicting SPAD Values of Maize Using Multi-Spectral Images

A Comprehensive Review of Crop Chlorophyll Mapping Using Remote Sensing Approaches: Achievements, Limitations, and Future Perspectives

Chlorophyll absorbs light energy and converts it into chemical energy, making it a crucial biochemical parameter for monitoring vegetation health, detecting environmental stress, and predicting physiological states. Accurate and rapid estimation of canopy chlorophyll content is crucial for assessing vegetation dynamics, ecological changes, and growth patterns. Remote sensing technology has become an indispensable tool for monitoring vegetation chlorophyll content since 2015, with more than 50 research papers published annually, contributing to a substantial body of case studies. This review discusses remote sensing technologies currently used for estimating vegetation chlorophyll content, focusing on four key aspects: the acquisition of reference datasets, the identification of optimal spectral variables, the selection of estimation models, and the analysis of application scenarios. The results indicate that spectral bands in the visible and red-edge regions (e.g., 530 nm, 670 nm, and 705 nm) provide high prediction accuracy. Machine learning methods, such as random forest and support vector regression, exhibit excellent performance, with determination coefficients (R2) typically exceeding 0.9, although overfitting remains an issue. Although radiative transfer models are slightly less accurate (R2 = 0.6-0.8), they provide greater interpretability. Hybrid models integrating machine learning and radiative transfer show strong potential to balance accuracy and generalizability. Future research should improve model generalizability for different vegetation types and environmental conditions and integrate multi-source remote sensing data to improve spatial and temporal resolution. Combining physical models with data processing methods, such as artificial intelligence, can improve scalability, cost-effectiveness, and real-time monitoring capabilities.

Nondestructive estimation of leaf chlorophyll content in banana based on unmanned aerial vehicle hyperspectral images using image feature combination methods

Introduction

Nondestructive quantification of leaf chlorophyll content (LCC) of banana and its spatial distribution across growth stages from remotely sensed data provide an effective avenue to diagnose nutritional deficiency and guide management practices. Unmanned aerial vehicle (UAV) hyperspectral imagery can document abundant texture features (TFs) and spectral information in a field experiment due to the high spatial and spectral resolutions. However, the benefits of using the fine spatial resolution accessible from UAV data for estimating LCC for banana have not been adequately quantified.

Methods

In this study, two types of image features including vegetation indices (VIs) and TFs extracted from the first-three-principal-component-analyzed images (TFs-PC1, TFs-PC2, and TFs-PC3) were employed. We proposed two methods of image feature combination for banana LCC inversion, which are a two-pair feature combination and a multivariable feature combination based on four machine learning algorithms (MLRAs).

Results

The results indicated that compared to conventionally used VIs alone, the banana LCC estimations with both proposed VI and TF combination methods were all significantly improved. Comprehensive analyses of the linear relationships between all constructed two-pair feature combinations and LCC indicated that the ratio of mean to modified red-edge sample ratio index (MEA/MSRre) stood out (R 2 = 0.745, RMSE = 2.17). For multivariable feature combinations, four MLRAs using original or two selected VIs and TFs-PC1 combination groups resulted in better LCC estimation than the other input variables. We concluded that the nonlinear Gaussian process regression model with the VIs and TFs-PC1 combination selected by maximal information coefficient as input achieved the highest accuracy in LCC prediction for banana, with the highest R 2 of 0.776 and lowest RMSE of 2.04. This study highlights the potential of the proposed image feature combination method for deriving high-resolution maps of banana LCC fundamental for precise nutritional diagnosing and operational agriculture management.

Identification of maize kernel varieties based on interpretable ensemble algorithms

Introduction

Maize kernel variety identification is crucial for reducing storage losses and ensuring food security. Traditional single models show limitations in processing large-scale multimodal data.

Methods

This study constructed an interpretable ensemble learning model for maize seed variety identification through improved differential evolutionary algorithm and multimodal data fusion. Morphological and hyperspectral data of maize samples were extracted and preprocessed, and three methods were used to screen features, respectively. The base learner of the Stacking integration model was selected using diversity and performance indices, with parameters optimized through a differential evolution algorithm incorporating multiple mutation strategies and dynamic adjustment of mutation factors and recombination rates. Shapley Additive exPlanation was applied for interpretable ensemble learning.

Results

The HDE-Stacking identification model achieved 97.78% accuracy. The spectral bands at 784 nm, 910 nm, 732 nm, 962 nm, and 666 nm showed positive impacts on identification results.

Discussion

This research provides a scientific basis for efficient identification of different corn kernel varieties, enhancing accuracy and traceability in germplasm resource management. The findings have significant practical value in agricultural production, improving quality management efficiency and contributing to food security assurance.

Hyperspectral indices data fusion-based machine learning enhanced by MRMR algorithm for estimating maize chlorophyll content

Accurate estimation of chlorophyll is essential for monitoring maize health and growth, for which hyperspectral imaging provides rich data. In this context, this paper presents an innovative method to estimate maize chlorophyll by combining hyperspectral indices and advanced machine learning models. The methodology of this study focuses on the development of machine learning models using proprietary hyperspectral indices to estimate corn chlorophyll content. Six advanced machine learning models were used, including robust linear stepwise regression, support vector machines (SVM), fine Gaussian SVM, Matern 5/2 Gaussian stepwise regression, and three-layer neural network. The MRMR algorithm was integrated into the process to improve feature selection by identifying the most informative spectral bands, thereby reducing data redundancy and improving model performance. The results showed significant differences in the performance of the six machine learning models applied to chlorophyll estimation. Among the models, the Matern 5/2 Gaussian process regression model showed the highest prediction accuracy. The model achieved R2 = 0.71 for the training set, RMSE = 338.46 µg/g and MAE = 264.30 µg/g. In the case of the validation set, the Matern 5/2 Gaussian process regression model further improved its performance, reaching R2 =0.79, RMSE=296.37 µg/g, MAE=237.12 µg/g. These metrics show that Matern's 5/2 Gaussian process regression model combined with the MRMR algorithm to select optimal traits is highly effective in predicting corn chlorophyll content. This research has important implications for precision agriculture, particularly for real-time monitoring and management of crop health. Accurate estimation of chlorophyll allows farmers to take timely and targeted action.

Random forest machine learning for maize yield and agronomic efficiency prediction in Ghana

Maize (Zea mays) is an important staple crop for food security in Sub-Saharan Africa. However, there is need to increase production to feed a growing population. In Ghana, this is mainly done by increasing acreage with adverse environmental consequences, rather than yield increment per unit area. Accurate prediction of maize yields and nutrient use efficiency in production is critical to making informed decisions toward economic and ecological sustainability. We trained the random forest machine learning algorithm to predict maize yield and agronomic efficiency in Ghana using soil, climate, environment, and management factors, including fertilizer application. We calibrated and evaluated the performance of the random forest machine learning algorithm using a 5 × 10-fold nested cross-validation approach. Data from 482 maize field trials consisting of 3136 georeferenced treatment plots conducted in Ghana from 1991 to 2020 were used to train the algorithm, identify important predictor variables, and quantify the uncertainties associated with the random forest predictions. The mean error, root mean squared error, model efficiency coefficient and 90 % prediction interval coverage probability were calculated. The results obtained on test data demonstrate good prediction performance for yield (MEC = 0.81) and moderate performance for agronomic efficiency (MEC = 0.63, 0.55 and 0.54 for AE-N, AE-P and AE-K, respectively). We found that climatic variables were less important predictors than soil variables for yield prediction, but temperature was of key importance to yield prediction and rainfall to agronomic efficiency. The developed random forest models provided a better understanding of the drivers of maize yield and agronomic efficiency in a tropical climate and an insight towards improving fertilizer recommendations for sustainable maize production and food security in Sub-Saharan Africa.

Machine learning approaches for estimation of the fraction of absorbed photosynthetically active radiation and net photosynthesis rate of maize using multi-spectral sensor

The fraction of absorbed photosynthetically active radiation (FAPAR) and the photosynthesis rate (Pn) of maize canopies were identified as essential photosynthetic parameters for accurately estimating vegetation growth and productivity using multispectral vegetation indices (VIs). Despite their importance, few studies have compared the effectiveness of multispectral imagery and various machine learning techniques in estimating these photosynthetic traits under high vegetation coverage. In this study, seventeen multispectral VIs and four machine learning (ML) algorithms were utilized to determine the most suitable model for estimating maize FAPAR and Pn during the kharif and rabi seasons at Tamil Nadu Agricultural University, Coimbatore, India. Results demonstrate that indices such as OSAVI, SAVI, EVI-2, and MSAVI-2 during the kharif and MNDVIRE and MSRRE during the rabi season outperformed others in estimating FAPAR and Pn values. Among the four ML methods of random forest (RF), extreme gradient boosting (XGBoost), support vector regression (SVR), and multiple linear regression (MLR) considered, RF consistently showed the most effective fitting effect and XGBoost demonstrated the least fitting accuracy for FAPAR and Pn estimation. However, SVR with R2 = 0.873 and RMSE = 0.045 during the kharif and MLR with R2 = 0.838 and RMSE = 0.053 during the rabi season demonstrated higher fitting accuracy, particularly notable for FAPAR prediction. Similarly, in the prediction of Pn, MLR showed higher fitting accuracy with R2 = 0.741 and RMSE = 2.531 during the kharif and R2 = 0.955 and RMSE = 1.070 during the rabi season. This study demonstrated the potential of combining UAV-derived VIs with ML to develop accurate FAPAR and Pn prediction models, overcoming VI saturation in dense vegetation. It underscores the importance of optimizing these models to improve the accuracy of maize vegetation assessments during various growing seasons.

Enhancing Winter Wheat Soil-Plant Analysis Development Value Prediction through Evaluating Unmanned Aerial Vehicle Flight Altitudes, Predictor Variable Combinations, and Machine Learning Algorithms

Monitoring winter wheat Soil-Plant Analysis Development (SPAD) values using Unmanned Aerial Vehicles (UAVs) is an effective and non-destructive method. However, predicting SPAD values during the booting stage is less accurate than other growth stages. Existing research on UAV-based SPAD value prediction has mainly focused on low-altitude flights of 10-30 m, neglecting the potential benefits of higher-altitude flights. The study evaluates predictions of winter wheat SPAD values during the booting stage using Vegetation Indices (VIs) from UAV images at five different altitudes (i.e., 20, 40, 60, 80, 100, and 120 m, respectively, using a DJI P4-Multispectral UAV as an example, with a resolution from 1.06 to 6.35 cm/pixel). Additionally, we compare the predictive performance using various predictor variables (VIs, Texture Indices (TIs), Discrete Wavelet Transform (DWT)) individually and in combination. Four machine learning algorithms (Ridge, Random Forest, Support Vector Regression, and Back Propagation Neural Network) are employed. The results demonstrate a comparable prediction performance between using UAV images at 120 m (with a resolution of 6.35 cm/pixel) and using the images at 20 m (with a resolution of 1.06 cm/pixel). This finding significantly improves the efficiency of UAV monitoring since flying UAVs at higher altitudes results in greater coverage, thus reducing the time needed for scouting when using the same heading overlap and side overlap rates. The overall trend in prediction accuracy is as follows: VIs + TIs + DWT > VIs + TIs > VIs + DWT > TIs + DWT > TIs > VIs > DWT. The VIs + TIs + DWT set obtains frequency information (DWT), compensating for the limitations of the VIs + TIs set. This study enhances the effectiveness of using UAVs in agricultural research and practices.

Combining the fractional order derivative and machine learning for leaf water content estimation of spring wheat using hyper-spectral indices

Leaf water content (LWC) is a vital indicator of crop growth and development. While visible and near-infrared (VIS-NIR) spectroscopy makes it possible to estimate crop leaf moisture, spectral preprocessing and multiband spectral indices have important significance in the quantitative analysis of LWC. In this work, the fractional order derivative (FOD) was used for leaf spectral processing, and multiband spectral indices were constructed based on the band-optimization algorithm. Eventually, an integrated index, namely, the multiband spectral index (MBSI) and moisture index (MI), is proposed to estimate the LWC in spring wheat around Fu-Kang City, Xinjiang, China. The MBSIs for LWC were calculated from two types of spectral data: raw reflectance (RR) and the spectrum based on FOD. The LWC was estimated by combining machine learning (K-nearest neighbor, KNN; support vector machine, SVM; and artificial neural network, ANN). The results showed that the fractional derivative pretreatment of spectral data enhances the implied information of the spectrum (the maximum correlation coefficient appeared using a 0.8-order differential) and increases the number of sensitive bands, especially in the near-infrared bands (700-1100 nm). The correlations between LWC and the two-band index (RVI1156, 1628 nm), three-band indices (3BI-3(766, 478, 1042 nm), 3BI-4(1129, 1175, 471 nm), 3BI-5(814, 929, 525 nm), 3BI-6(1156, 1214, 802 nm), 3BI-7(929, 851, 446 nm)) based on FOD were higher than that of moisture indices and single-band spectrum, with r of - 0.71**, 0.74**, 0.73**, - 0.72**, 0.75** and - 0.76** for the correlation. The prediction accuracy of the two-band spectral indices (DVI(698, 1274 nm) DVI(698, 1274 nm) DVI(698, 1274 nm)) was higher than that of the moisture spectral index, with R2 of 0.81 and R2 of 0.79 for the calibration and validation, respectively. Due to a large amount of spectral indices, the correlation coefficient method was used to select the characteristic spectral index from full three-band indices. Among twenty seven models, the FWBI-3BI- 0.8 order model performed the best predictive ability (with an R2 of 0.86, RMSE of 2.11%, and RPD of 2.65). These findings confirm that combining spectral index optimization with machine learning is a highly effective method for inverting the leaf water content in spring wheat.

A method for durian precise fertilization based on improved radial basis neural network algorithm

Introduction

Durian is one of the tropical fruits that requires soil nutrients in its cultivation. It is important to understand the relationship between the content of critical nutrients, such as nitrogen (N), phosphorus (P), and potassium (K) in the soil and durian yield. How to optimize the fertilization plan is also important to the durian planting.

Methods

Thus, this study proposes an Improved Radial Basis Neural Network Algorithm (IM-RBNNA) in the durian precision fertilization. It uses the gray wolf algorithm to optimize the weights and thresholds of the RBNNA algorithm, which can improve the prediction accuracy of the RBNNA algorithm for the soil nutrient content and its relationship with the durian yield. It also collects the soil nutrients and historical yield data to build the IM-RBNNA model and compare with other similar algorithms.

Results

The results show that the IM-RBNNA algorithm is better than the other three algorithms in the average relative error, average absolute error, and coefficient of determination between the predicted and true values of soil N, K, and P fertilizer contents. It also predicts the relationship between soil nutrients and yield, which is closer to the true value.

Discussion

It shows that the IM-RBNNA algorithm can accurately predict the durian soil nutrient content and yield, which is benefited for farmers to make agronomic plans and management strategies. It uses soil nutrient resources efficiently, which reduces the environmental negative impacts. It also ensures that the durian tree can obtain the appropriate amount of nutrients, maximize its growth potential, reduce production costs, and increase yields.

Machine Learning Application in Horticulture and Prospects for Predicting Fresh Produce Losses and Waste: A Review

The current review examines the state of knowledge and research on machine learning (ML) applications in horticultural production and the potential for predicting fresh produce losses and waste. Recently, ML has been increasingly applied in horticulture for efficient and accurate operations. Given the health benefits of fresh produce and the need for food and nutrition security, efficient horticultural production and postharvest management are important. This review aims to assess the application of ML in preharvest and postharvest horticulture and the potential of ML in reducing postharvest losses and waste by predicting their magnitude, which is crucial for management practices and policymaking in loss and waste reduction. The review starts by assessing the application of ML in preharvest horticulture. It then presents the application of ML in postharvest handling and processing, and lastly, the prospects for its application in postharvest loss and waste quantification. The findings revealed that several ML algorithms perform satisfactorily in classification and prediction tasks. Based on that, there is a need to further investigate the suitability of more models or a combination of models with a higher potential for classification and prediction. Overall, the review suggested possible future directions for research related to the application of ML in postharvest losses and waste quantification.

Quantitative and Qualitative Analysis of Agricultural Fields Based on Aerial Multispectral Images Using Neural Networks

This article presents an integrated system that uses the capabilities of unmanned aerial vehicles (UAVs) to perform a comprehensive crop analysis, combining qualitative and quantitative evaluations for efficient agricultural management. A convolutional neural network-based model, Detectron2, serves as the foundation for detecting and segmenting objects of interest in acquired aerial images. This model was trained on a dataset prepared using the COCO format, which features a variety of annotated objects. The system architecture comprises a frontend and a backend component. The frontend facilitates user interaction and annotation of objects on multispectral images. The backend involves image loading, project management, polygon handling, and multispectral image processing. For qualitative analysis, users can delineate regions of interest using polygons, which are then subjected to analysis using the Normalized Difference Vegetation Index (NDVI) or Optimized Soil Adjusted Vegetation Index (OSAVI). For quantitative analysis, the system deploys a pre-trained model capable of object detection, allowing for the counting and localization of specific objects, with a focus on young lettuce crops. The prediction quality of the model has been calculated using the AP (Average Precision) metric. The trained neural network exhibited robust performance in detecting objects, even within small images.

Improving grain yield prediction through fusion of multi-temporal spectral features and agronomic trait parameters derived from UAV imagery

Rapid and accurate prediction of crop yield is particularly important for ensuring national and regional food security and guiding the formulation of agricultural and rural development plans. Due to unmanned aerial vehicles' ultra-high spatial resolution, low cost, and flexibility, they are widely used in field-scale crop yield prediction. Most current studies used the spectral features of crops, especially vegetation or color indices, to predict crop yield. Agronomic trait parameters have gradually attracted the attention of researchers for use in the yield prediction in recent years. In this study, the advantages of multispectral and RGB images were comprehensively used and combined with crop spectral features and agronomic trait parameters (i.e., canopy height, coverage, and volume) to predict the crop yield, and the effects of agronomic trait parameters on yield prediction were investigated. The results showed that compared with the yield prediction using spectral features, the addition of agronomic trait parameters effectively improved the yield prediction accuracy. The best feature combination was the canopy height (CH), fractional vegetation cover (FVC), normalized difference red-edge index (NDVI_RE), and enhanced vegetation index (EVI). The yield prediction error was 8.34%, with an R2 of 0.95. The prediction accuracies were notably greater in the stages of jointing, booting, heading, and early grain-filling compared to later stages of growth, with the heading stage displaying the highest accuracy in yield prediction. The prediction results based on the features of multiple growth stages were better than those based on a single stage. The yield prediction across different cultivars was weaker than that of the same cultivar. Nevertheless, the combination of agronomic trait parameters and spectral indices improved the prediction among cultivars to some extent.

Paddy Rice mapping in fragmented lands by improved phenology curve and correlation measurements on Sentinel-2 imagery in Google earth engine

Accurate and timely rice crop mapping is important to address the challenges of food security, water management, disease transmission, and land use change. However, accurate rice crop mapping is difficult due to the presence of mixed pixels in small and fragmented rice fields as well as cloud cover. In this paper, a phenology-based method using Sentinel-2 time series images is presented to solve these problems. First, the improved rice phenology curve is extracted based on Normalized Difference Vegetation Index and Land Surface Water Index time series data of rice fields. Then, correlation was taken between rice phenology curve and time series data of each pixel. The correlation result of each pixel shows the similarity of its time series behavior with the proposed rice phenology curve. In the next step, the maximum correlation value and its occurrence time are used as the feature vectors of each pixel to classification. Since correlation measurement provides data with better separability than its input data, training the classifier can be done with fewer samples and the classification is more accurate. The implementation of the proposed correlation-based algorithm can be done in a parallel computing. All the processes were performed on the Google Earth Engine cloud platform on the time series images of the Sentinel 2. The implementations show the high accuracy of this method.

Predicting Maize Theoretical Methane Yield in Combination with Ground and UAV Remote Data Using Machine Learning

The accurate, timely, and non-destructive estimation of maize total-above ground biomass (TAB) and theoretical biochemical methane potential (TBMP) under different phenological stages is a substantial part of agricultural remote sensing. The assimilation of UAV and machine learning (ML) data may be successfully applied in predicting maize TAB and TBMP; however, in the Nordic-Baltic region, these technologies are not fully exploited. Therefore, in this study, during the maize growing period, we tracked unmanned aerial vehicle (UAV) based multispectral bands (blue, red, green, red edge, and infrared) at the main phenological stages. In the next step, we calculated UAV-based vegetation indices, which were combined with field measurements and different ML models, including generalized linear, random forest, as well as support vector machines. The results showed that the best ML predictions were obtained during the maize blister (R2)-Dough (R4) growth period when the prediction models managed to explain 88-95% of TAB and 88-97% TBMP variation. However, for the practical usage of farmers, the earliest suitable timing for adequate TAB and TBMP prediction in the Nordic-Baltic area is stage V7-V10. We conclude that UAV techniques in combination with ML models were successfully applied for maize TAB and TBMP estimation, but similar research should be continued for further improvements.

A new deep learning-based technique for rice pest detection using remote sensing

Background

Agriculture plays a vital role in the country's economy and human society. Rice production is mainly focused on financial improvements as it is demanding worldwide. Protecting the rice field from pests during seedling and after production is becoming a challenging research problem. Identifying the pest at the right time is crucial so that the measures to prevent rice crops from pests can be taken by considering its stage. In this article, a new deep learning-based pest detection model is proposed. The proposed system can detect two types of rice pests (stem borer and Hispa) using an unmanned aerial vehicle (UAV).

Methodology

The image is captured in real time by a camera mounted on the UAV and then processed by filtering, labeling, and segmentation-based technique of color thresholding to convert the image into greyscale for extracting the region of interest. This article provides a rice pests dataset and a comparative analysis of existing pre-trained models. The proposed approach YO-CNN recommended in this study considers the results of the previous model because a smaller network was regarded to be better than a bigger one. Using additional layers has the advantage of preventing memorization, and it provides more precise results than existing techniques.

Results

The main contribution of the research is implementing a new modified deep learning model named Yolo-convolution neural network (YO-CNN) to obtain a precise output of up to 0.980 accuracies. It can be used to reduce rice wastage during production by monitoring the pests regularly. This technique can be used further for target spraying that saves applicators (fertilizer water and pesticide) and reduces the adverse effect of improper use of applicators on the environment and human beings.

Machine Learning in the Classification of Soybean Genotypes for Primary Macronutrients’ Content Using UAV–Multispectral Sensor

Using spectral data to quantify nitrogen (N), phosphorus (P), and potassium (K) contents in soybean plants can help breeding programs develop fertilizer-efficient genotypes. Employing machine learning (ML) techniques to classify these genotypes according to their nutritional content makes the analyses performed in the programs even faster and more reliable. Thus, the objective of this study was to find the best ML algorithm(s) and input configurations in the classification of soybean genotypes for higher N, P, and K leaf contents. A total of 103 F2 soybean populations were evaluated in a randomized block design with two repetitions. At 60 days after emergence (DAE), spectral images were collected using a Sensefly eBee RTK fixed-wing remotely piloted aircraft (RPA) with autonomous take-off, flight plan, and landing control. The eBee was equipped with the Parrot Sequoia multispectral sensor. Reflectance values were obtained in the following spectral bands (SBs): red (660 nm), green (550 nm), NIR (735 nm), and red-edge (790 nm), which were used to calculate the vegetation index (VIs): normalized difference vegetation index (NDVI), normalized difference red edge (NDRE), green normalized difference vegetation index (GNDVI), soil-adjusted vegetation index (SAVI), modified soil-adjusted vegetation index (MSAVI), modified chlorophyll absorption in reflectance index (MCARI), enhanced vegetation index (EVI), and simplified canopy chlorophyll content index (SCCCI). At the same time of the flight, leaves were collected in each experimental unit to obtain the leaf contents of N, P, and K. The data were submitted to a Pearson correlation analysis. Subsequently, a principal component analysis was performed together with the k-means algorithm to define two clusters: one whose genotypes have high leaf contents and another whose genotypes have low leaf contents. Boxplots were generated for each cluster according to the content of each nutrient within the groups formed, seeking to identify which set of genotypes has higher nutrient contents. Afterward, the data were submitted to machine learning analysis using the following algorithms: decision tree algorithms J48 and REPTree, random forest (RF), artificial neural network (ANN), support vector machine (SVM), and logistic regression (LR, used as control). The clusters were used as output variables of the classification models used. The spectral data were used as input variables for the models, and three different configurations were tested: using SB only, using VIs only, and using SBs+VIs. The J48 and SVM algorithms had the best performance in classifying soybean genotypes. The best input configuration for the algorithms was using the spectral bands as input.

Visible and near-infrared spectroscopy and deep learning application for the qualitative and quantitative investigation of nitrogen status in cotton leaves

Leaf nitrogen concentration (LNC) is a critical indicator of crop nutrient status. In this study, the feasibility of using visible and near-infrared spectroscopy combined with deep learning to estimate LNC in cotton leaves was explored. The samples were collected from cotton's whole growth cycle, and the spectra were from different measurement environments. The random frog (RF), weighted partial least squares regression (WPLS), and saliency map were used for characteristic wavelength selection. Qualitative models (partial least squares discriminant analysis (PLS-DA), support vector machine for classification (SVC), convolutional neural network classification (CNNC) and quantitative models (partial least squares regression (PLSR), support vector machine for regression (SVR), convolutional neural network regression (CNNR)) were established based on the full spectra and characteristic wavelengths. Satisfactory results were obtained by models based on CNN. The classification accuracy of leaves in three different LNC ranges was up to 83.34%, and the root mean square error of prediction (RMSEP) of quantitative prediction models of cotton leaves was as low as 3.36. In addition, the identification of cotton leaves based on the predicted LNC also achieved good results. These results indicated that the nitrogen content of cotton leaves could be effectively detected by deep learning and visible and near-infrared spectroscopy, which has great potential for real-world application.

UAV-Based Forest Health Monitoring: A Systematic Review

In recent years, technological advances have led to the increasing use of unmanned aerial vehicles (UAVs) for forestry applications. One emerging field for drone application is forest health monitoring (FHM). Common approaches for FHM involve small-scale resource-extensive fieldwork combined with traditional remote sensing platforms. However, the highly dynamic nature of forests requires timely and repetitive data acquisition, often at very high spatial resolution, where conventional remote sensing techniques reach the limits of feasibility. UAVs have shown that they can meet the demands of flexible operation and high spatial resolution. This is also reflected in a rapidly growing number of publications using drones to study forest health. Only a few reviews exist which do not cover the whole research history of UAV-based FHM. Since a comprehensive review is becoming critical to identify research gaps, trends, and drawbacks, we offer a systematic analysis of 99 papers covering the last ten years of research related to UAV-based monitoring of forests threatened by biotic and abiotic stressors. Advances in drone technology are being rapidly adopted and put into practice, further improving the economical use of UAVs. Despite the many advantages of UAVs, such as their flexibility, relatively low costs, and the possibility to fly below cloud cover, we also identified some shortcomings: (1) multitemporal and long-term monitoring of forests is clearly underrepresented; (2) the rare use of hyperspectral and LiDAR sensors must drastically increase; (3) complementary data from other RS sources are not sufficiently being exploited; (4) a lack of standardized workflows poses a problem to ensure data uniformity; (5) complex machine learning algorithms and workflows obscure interpretability and hinders widespread adoption; (6) the data pipeline from acquisition to final analysis often relies on commercial software at the expense of open-source tools.