Using Hybrid Artificial Intelligence and Evolutionary Optimization Algorithms for Estimating Soybean Yield and Fresh Biomass Using Hyperspectral Vegetation Indices

High-Throughput Phenotyping of Soybean Biomass: Conventional Trait Estimation and Novel Latent Feature Extraction Using UAV Remote Sensing and Deep Learning Models

High-throughput phenotyping serves as a framework to reduce chronological costs and accelerate breeding cycles. In this study, we developed models to estimate the phenotypes of biomass-related traits in soybean (Glycine max) using unmanned aerial vehicle (UAV) remote sensing and deep learning models. In 2018, a field experiment was conducted using 198 soybean germplasm accessions with known whole-genome sequences under 2 irrigation conditions: drought and control. We used a convolutional neural network (CNN) as a model to estimate the phenotypic values of 5 conventional biomass-related traits: dry weight, main stem length, numbers of nodes and branches, and plant height. We utilized manually measured phenotypes of conventional traits along with RGB images and digital surface models from UAV remote sensing to train our CNN models. The accuracy of the developed models was assessed through 10-fold cross-validation, which demonstrated their ability to accurately estimate the phenotypes of all conventional traits simultaneously. Deep learning enabled us to extract features that exhibited strong correlations with the output (i.e., phenotypes of the target traits) and accurately estimate the values of the features from the input data. We considered the extracted low-dimensional features as phenotypes in the latent space and attempted to annotate them based on the phenotypes of conventional traits. Furthermore, we validated whether these low-dimensional latent features were genetically controlled by assessing the accuracy of genomic predictions. The results revealed the potential utility of these low-dimensional latent features in actual breeding scenarios.

Prediction accuracy and repeatability of UAV based biomass estimation in wheat variety trials as affected by variable type, modelling strategy and sampling location

BACKGROUND

This study explores the use of Unmanned Aerial Vehicles (UAVs) for estimating wheat biomass, focusing on the impact of phenotyping and analytical protocols in the context of late-stage variety selection programs. It emphasizes the importance of variable selection, model specificity, and sampling location within the experimental plot in predicting biomass, aiming to refine UAV-based estimation techniques for enhanced selection accuracy and throughput in variety testing programs.

RESULTS

The research uncovered that integrating geometric and spectral traits led to an increase in prediction accuracy, whilst a recursive feature elimination (RFE) based variable selection workflowled to slight reductions in accuracy with the benefit of increased interpretability. Models, tailored to specific experiments were more accurate than those modelling all experiments together, while models trained for broad-growth stages did not significantly increase accuracy. The comparison between a permanent and a precise region of interest (ROI) within the plot showed negligible differences in biomass prediction accuracy, indicating the robustness of the approach across different sampling locations within the plot. Significant differences in the within-season repeatability (w2) of biomass predictions across different experiments highlighted the need for further investigation into the optimal timing of measurement for prediction.

CONCLUSIONS

The study highlights the promising potential of UAV technology in biomass prediction for wheat at a small plot scale. It suggests that the accuracy of biomass predictions can be significantly improved through optimizing analytical and modelling protocols (i.e., variable selection, algorithm selection, stage-specific model development). Future work should focus on exploring the applicability of these findings under a wider variety of conditions and from a more diverse set of genotypes.

Artificial intelligence in plant breeding

Harnessing cutting-edge technologies to enhance crop productivity is a pivotal goal in modern plant breeding. Artificial intelligence (AI) is renowned for its prowess in big data analysis and pattern recognition, and is revolutionizing numerous scientific domains including plant breeding. We explore the wider potential of AI tools in various facets of breeding, including data collection, unlocking genetic diversity within genebanks, and bridging the genotype-phenotype gap to facilitate crop breeding. This will enable the development of crop cultivars tailored to the projected future environments. Moreover, AI tools also hold promise for refining crop traits by improving the precision of gene-editing systems and predicting the potential effects of gene variants on plant phenotypes. Leveraging AI-enabled precision breeding can augment the efficiency of breeding programs and holds promise for optimizing cropping systems at the grassroots level. This entails identifying optimal inter-cropping and crop-rotation models to enhance agricultural sustainability and productivity in the field.

AI can empower agriculture for global food security: challenges and prospects in developing nations

Food and nutrition are a steadfast essential to all living organisms. With specific reference to humans, the sufficient and efficient supply of food is a challenge as the world population continues to grow. Artificial Intelligence (AI) could be identified as a plausible technology in this 5th industrial revolution in bringing us closer to achieving zero hunger by 2030-Goal 2 of the United Nations Sustainable Development Goals (UNSDG). This goal cannot be achieved unless the digital divide among developed and underdeveloped countries is addressed. Nevertheless, developing and underdeveloped regions fall behind in economic resources; however, they harbor untapped potential to effectively address the impending demands posed by the soaring world population. Therefore, this study explores the in-depth potential of AI in the agriculture sector for developing and under-developed countries. Similarly, it aims to emphasize the proven efficiency and spin-off applications of AI in the advancement of agriculture. Currently, AI is being utilized in various spheres of agriculture, including but not limited to crop surveillance, irrigation management, disease identification, fertilization practices, task automation, image manipulation, data processing, yield forecasting, supply chain optimization, implementation of decision support system (DSS), weed control, and the enhancement of resource utilization. Whereas AI supports food safety and security by ensuring higher crop yields that are acquired by harnessing the potential of multi-temporal remote sensing (RS) techniques to accurately discern diverse crop phenotypes, monitor land cover dynamics, assess variations in soil organic matter, predict soil moisture levels, conduct plant biomass modeling, and enable comprehensive crop monitoring. The present study identifies various challenges, including financial, infrastructure, experts, data availability, customization, regulatory framework, cultural norms and attitudes, access to market, and interdisciplinary collaboration, in the adoption of AI for developing nations with their subsequent remedies. The identification of challenges and opportunities in the implementation of AI could ignite further research and actions in these regions; thereby supporting sustainable development.

Machine learning-mediated Passiflora caerulea callogenesis optimization

Callogenesis is one of the most powerful biotechnological approaches for in vitro secondary metabolite production and indirect organogenesis in Passiflora caerulea. Comprehensive knowledge of callogenesis and optimized protocol can be obtained by the application of a combination of machine learning (ML) and optimization algorithms. In the present investigation, the callogenesis responses (i.e., callogenesis rate and callus fresh weight) of P. caerulea were predicted based on different types and concentrations of plant growth regulators (PGRs) (i.e., 2,4-dichlorophenoxyacetic acid (2,4-D), 6-benzylaminopurine (BAP), 1-naphthaleneacetic acid (NAA), and indole-3-Butyric Acid (IBA)) as well as explant types (i.e., leaf, node, and internode) using multilayer perceptron (MLP). Moreover, the developed models were integrated into the genetic algorithm (GA) to optimize the concentration of PGRs and explant types for maximizing callogenesis responses. Furthermore, sensitivity analysis was conducted to assess the importance of each input variable on the callogenesis responses. The results showed that MLP had high predictive accuracy (R2 > 0.81) in both training and testing sets for modeling all studied parameters. Based on the results of the optimization process, the highest callogenesis rate (100%) would be obtained from the leaf explant cultured in the medium supplemented with 0.52 mg/L IBA plus 0.43 mg/L NAA plus 1.4 mg/L 2,4-D plus 0.2 mg/L BAP. The results of the sensitivity analysis showed the explant-dependent impact of the exogenous application of PGRs on callogenesis. Generally, the results showed that a combination of MLP and GA can display a forward-thinking aid to optimize and predict in vitro culture systems and consequentially cope with several challenges faced currently in Passiflora tissue culture.

Predicting and optimizing reactive oxygen species metabolism in Punica granatum L. through machine learning: role of exogenous GABA on antioxidant enzyme activity under drought and salinity stress

BACKGROUND

Drought and salinity stress have been proposed as the main environmental factors threatening food security, as they adversely affect crops' agricultural productivity. As a potential solution, the application of plant growth regulators to enhance drought and salinity tolerance has gained considerable attention. γ-aminobutyric acid (GABA) is a four-carbon non-protein amino acid that accumulates in plants as a response to stressful conditions. This study focused on a comparative assessment of several machine learning (ML) regression models, including radial basis function, generalized regression neural network (GRNN), random forest (RF), and support vector regression (SVR) to develop predictive models for assessing the effect of different concentrations of GABA (0, 10, 20, and 40 mM) on various physio-biochemical traits during periods of drought, salinity, and combined stress conditions. The physio-biochemical traits included antioxidant enzyme activities (superoxide dismutase, SOD; peroxidase, POD; catalase, CAT; and ascorbate peroxidase, APX), protein content, malondialdehyde (MDA) levels, and hydrogen peroxide (H2O2) levels. The non‑dominated sorting genetic algorithm‑II (NSGA‑II) was employed for optimizing the superior prediction model.

RESULTS

The GRNN model outperformed the other ML algorithms and was therefore selected for optimization by NSGA-II. The GRNN-NSGA-II model revealed that treatment with GABA at concentrations of 20.90 mM and 20.54 mM, under combined drought and salinity stress conditions at 20.86 and 20.72 days post-treatment, respectively, could result in the maximum values for protein content (by 0.80 and 0.69), APX activity (by 50.63 and 51.51), SOD activity (by 0.54 and 0.53), POD activity (by 1.53 and 1.72), CAT activity (by 4.42 and 5.66), as well as lower MDA levels (by 0.12 and 0.15) and H2O2 levels (by 0.44 and 0.55), respectively, in the 'Atabaki' and 'Rabab' cultivars.

CONCLUSIONS

This study demonstrates that the GRNN-NSGA-II model, as an advanced ML algorithm with a strong predictive ability for outcomes in combined stressful environmental conditions, provides valuable insights into the significant factors influencing such multifactorial processes.

Enhancing petunia tissue culture efficiency with machine learning: A pathway to improved callogenesis

The important feature of petunia in tissue culture is its unpredictable and genotype-dependent callogenesis, posing challenges for efficient regeneration and biotechnology applications. To address this issue, machine learning (ML) can be considered a powerful tool to analyze callogenesis data, extract key parameters, and predict optimal conditions for petunia callogenesis, facilitating more controlled and productive tissue culture processes. The study aimed to develop a predictive model for callogenesis in petunia using ML algorithms and to optimize the concentrations of phytohormones to enhance callus formation rate (CFR) and callus fresh weight (CFW). The inputs for the model were BAP, KIN, IBA, and NAA, while the outputs were CFR and CFW. Three ML algorithms, namely MLP, RBF, and GRNN, were compared, and the results revealed that GRNN (R2≥83) outperformed MLP and RBF in terms of accuracy. Furthermore, a sensitivity analysis was conducted to determine the relative importance of the four phytohormones. IBA exhibited the highest importance, followed by NAA, BAP, and KIN. Leveraging the superior performance of the GRNN model, a genetic algorithm (GA) was integrated to optimize the concentration of phytohormones for maximizing CFR and CFW. The genetic algorithm identified an optimized combination of phytohormones consisting of 1.31 mg/L BAP, 1.02 mg/L KIN, 1.44 mg/L NAA, and 1.70 mg/L IBA, resulting in 95.83% CFR. To validate the reliability of the predicted results, optimized combinations of phytohormones were tested in a laboratory experiment. The results of the validation experiment indicated no significant difference between the experimental and optimized results obtained through the GA. This study presents a novel approach combining ML, sensitivity analysis, and GA for modeling and predicting callogenesis in petunia. The findings offer valuable insights into the optimization of phytohormone concentrations, facilitating improved callus formation and potential applications in plant tissue culture and genetic engineering.

Prediction and optimization of indirect shoot regeneration of Passiflora caerulea using machine learning and optimization algorithms

BACKGROUND

Optimization of indirect shoot regeneration protocols is one of the key prerequisites for the development of Agrobacterium-mediated genetic transformation and/or genome editing in Passiflora caerulea. Comprehensive knowledge of indirect shoot regeneration and optimized protocol can be obtained by the application of a combination of machine learning (ML) and optimization algorithms.

MATERIALS AND METHODS

In the present investigation, the indirect shoot regeneration responses (i.e., de novo shoot regeneration rate, the number of de novo shoots, and length of de novo shoots) of P. caerulea were predicted based on different types and concentrations of PGRs (i.e., TDZ, BAP, PUT, KIN, and IBA) as well as callus types (i.e., callus derived from different explants including leaf, node, and internode) using generalized regression neural network (GRNN) and random forest (RF). Moreover, the developed models were integrated into the genetic algorithm (GA) to optimize the concentration of PGRs and callus types for maximizing indirect shoot regeneration responses. Moreover, sensitivity analysis was conducted to assess the importance of each input variable on the studied parameters.

RESULTS

The results showed that both algorithms (RF and GRNN) had high predictive accuracy (R² > 0.86) in both training and testing sets for modeling all studied parameters. Based on the results of optimization process, the highest de novo shoot regeneration rate (100%) would be obtained from callus derived from nodal segments cultured in the medium supplemented with 0.77 mg/L BAP plus 2.41 mg/L PUT plus 0.06 mg/L IBA. The results of the sensitivity analysis showed the explant-dependent impact of exogenous application of PGRs on indirect de novo shoot regeneration.

CONCLUSIONS

A combination of ML (GRNN and RF) and GA can display a forward-thinking aid to optimize and predict in vitro culture systems and consequentially cope with several challenges faced currently in Passiflora tissue culture.

Application of SVR-Mediated GWAS for Identification of Durable Genetic Regions Associated with Soybean Seed Quality Traits

Soybean (Glycine max L.) is an important food-grade strategic crop worldwide because of its high seed protein and oil contents. Due to the negative correlation between seed protein and oil percentage, there is a dire need to detect reliable quantitative trait loci (QTL) underlying these traits in order to be used in marker-assisted selection (MAS) programs. Genome-wide association study (GWAS) is one of the most common genetic approaches that is regularly used for detecting QTL associated with quantitative traits. However, the current approaches are mainly focused on estimating the main effects of QTL, and, therefore, a substantial statistical improvement in GWAS is required to detect associated QTL considering their interactions with other QTL as well. This study aimed to compare the support vector regression (SVR) algorithm as a common machine learning method to fixed and random model circulating probability unification (FarmCPU), a common conventional GWAS method in detecting relevant QTL associated with soybean seed quality traits such as protein, oil, and 100-seed weight using 227 soybean genotypes. The results showed a significant negative correlation between soybean seed protein and oil concentrations, with heritability values of 0.69 and 0.67, respectively. In addition, SVR-mediated GWAS was able to identify more relevant QTL underlying the target traits than the FarmCPU method. Our findings demonstrate the potential use of machine learning algorithms in GWAS to detect durable QTL associated with soybean seed quality traits suitable for genomic-based breeding approaches. This study provides new insights into improving the accuracy and efficiency of GWAS and highlights the significance of using advanced computational methods in crop breeding research.

Non-Invasive Assessment, Classification, and Prediction of Biophysical Parameters Using Reflectance Hyperspectroscopy

Hyperspectral technology offers significant potential for non-invasive monitoring and prediction of morphological parameters in plants. In this study, UV-VIS-NIR-SWIR reflectance hyperspectral data were collected from Nicotiana tabacum L. plants using a spectroradiometer. These plants were grown under different light and gibberellic acid (GA₃) concentrations. Through spectroscopy and multivariate analyses, key growth parameters, such as height, leaf area, energy yield, and biomass, were effectively evaluated based on the interaction of light with leaf structures. The shortwave infrared (SWIR) bands, specifically SWIR1 and SWIR2, showed the strongest correlations with these growth parameters. When classifying tobacco plants grown under different GA₃ concentrations in greenhouses, artificial intelligence (AI) and machine learning (ML) algorithms were employed, achieving an average accuracy of over 99.1% using neural network (NN) and gradient boosting (GB) algorithms. Among the 34 tested vegetation indices, the photochemical reflectance index (PRI) demonstrated the strongest correlations with all evaluated plant phenotypes. Partial least squares regression (PLSR) models effectively predicted morphological attributes, with R²_CV values ranging from 0.81 to 0.87 and RPD_P values exceeding 2.09 for all parameters. Based on Pearson's coefficient XYZ interpolations and HVI algorithms, the NIR-SWIR band combination proved the most effective for predicting height and leaf area, while VIS-NIR was optimal for optimal energy yield, and VIS-VIS was best for predicting biomass. To further corroborate these findings, the SWIR bands for certain morphological characteristic wavelengths selected with s-PLS were most significant for SWIR1 and SWIR2, while i-PLS showed a more uniform distribution in VIS-NIR-SWIR bands. Therefore, SWIR hyperspectral bands provide valuable insights into developing alternative bands for remote sensing measurements to estimate plant morphological parameters. These findings underscore the potential of remote sensing technology for rapid, accurate, and non-invasive monitoring within stationary high-throughput phenotyping systems in greenhouses. These insights align with advancements in digital and precision technology, indicating a promising future for research and innovation in this field.

Reflectance Spectroscopy for the Classification and Prediction of Pigments in Agronomic Crops

Reflectance spectroscopy, in combination with machine learning and artificial intelligence algorithms, is an effective method for classifying and predicting pigments and phenotyping in agronomic crops. This study aims to use hyperspectral data to develop a robust and precise method for the simultaneous evaluation of pigments, such as chlorophylls, carotenoids, anthocyanins, and flavonoids, in six agronomic crops: corn, sugarcane, coffee, canola, wheat, and tobacco. Our results demonstrate high classification accuracy and precision, with principal component analyses (PCAs)-linked clustering and a kappa coefficient analysis yielding results ranging from 92 to 100% in the ultraviolet-visible (UV-VIS) to near-infrared (NIR) to shortwave infrared (SWIR) bands. Predictive models based on partial least squares regression (PLSR) achieved R² values ranging from 0.77 to 0.89 and ratio of performance to deviation (RPD) values over 2.1 for each pigment in C₃ and C₄ plants. The integration of pigment phenotyping methods with fifteen vegetation indices further improved accuracy, achieving values ranging from 60 to 100% across different full or range wavelength bands. The most responsive wavelengths were selected based on a cluster heatmap, β-loadings, weighted coefficients, and hyperspectral vegetation index (HVI) algorithms, thereby reinforcing the effectiveness of the generated models. Consequently, hyperspectral reflectance can serve as a rapid, precise, and accurate tool for evaluating agronomic crops, offering a promising alternative for monitoring and classification in integrated farming systems and traditional field production. It provides a non-destructive technique for the simultaneous evaluation of pigments in the most important agronomic plants.

Remotely Sensed Phenotypic Traits for Heritability Estimates and Grain Yield Prediction of Barley Using Multispectral Imaging from UAVs

This study tested the potential of parametric and nonparametric regression modeling utilizing multispectral data from two different unoccupied aerial vehicles (UAVs) as a tool for the prediction of and indirect selection of grain yield (GY) in barley breeding experiments. The coefficient of determination (R²) of the nonparametric models for GY prediction ranged between 0.33 and 0.61 depending on the UAV and flight date, where the highest value was achieved with the DJI Phantom 4 Multispectral (P4M) image from 26 May (milk ripening). The parametric models performed worse than the nonparametric ones for GY prediction. Independent of the retrieval method and UAV, GY retrieval was more accurate in milk ripening than dough ripening. The leaf area index (LAI), fraction of absorbed photosynthetically active radiation (fAPAR), fraction vegetation cover (fCover), and leaf chlorophyll content (LCC) were modeled at milk ripening using nonparametric models with the P4M images. A significant effect of the genotype was found for the estimated biophysical variables, which was referred to as remotely sensed phenotypic traits (RSPTs). Measured GY heritability was lower, with a few exceptions, compared to the RSPTs, indicating that GY was more environmentally influenced than the RSPTs. The moderate to strong genetic correlation of the RSPTs to GY in the present study indicated their potential utility as an indirect selection approach to identify high-yield genotypes of winter barley.

Machine Learning-Assisted Approaches in Modernized Plant Breeding Programs

In the face of a growing global population, plant breeding is being used as a sustainable tool for increasing food security. A wide range of high-throughput omics technologies have been developed and used in plant breeding to accelerate crop improvement and develop new varieties with higher yield performance and greater resilience to climate changes, pests, and diseases. With the use of these new advanced technologies, large amounts of data have been generated on the genetic architecture of plants, which can be exploited for manipulating the key characteristics of plants that are important for crop improvement. Therefore, plant breeders have relied on high-performance computing, bioinformatics tools, and artificial intelligence (AI), such as machine-learning (ML) methods, to efficiently analyze this vast amount of complex data. The use of bigdata coupled with ML in plant breeding has the potential to revolutionize the field and increase food security. In this review, some of the challenges of this method along with some of the opportunities it can create will be discussed. In particular, we provide information about the basis of bigdata, AI, ML, and their related sub-groups. In addition, the bases and functions of some learning algorithms that are commonly used in plant breeding, three common data integration strategies for the better integration of different breeding datasets using appropriate learning algorithms, and future prospects for the application of novel algorithms in plant breeding will be discussed. The use of ML algorithms in plant breeding will equip breeders with efficient and effective tools to accelerate the development of new plant varieties and improve the efficiency of the breeding process, which are important for tackling some of the challenges facing agriculture in the era of climate change.

Optimizing genomic selection in soybean: An important improvement in agricultural genomics

Fast-paced yield improvement in strategic crops such as soybean is pivotal for achieving sustainable global food security. Precise genomic selection (GS), as one of the most effective genomic tools for recognizing superior genotypes, can accelerate the efficiency of breeding programs through shortening the breeding cycle, resulting in significant increases in annual yield improvement. In this study, we investigated the possible use of haplotype-based GS to increase the prediction accuracy of soybean yield and its component traits through augmenting the models by using sophisticated machine learning algorithms and optimized genetic information. The results demonstrated up to a 7% increase in the prediction accuracy when using haplotype-based GS over the full single nucleotide polymorphisms-based GS methods. In addition, we discover an auspicious haplotype block on chromosome 19 with significant impacts on yield and its components, which can be used for screening climate-resilient soybean genotypes with improved yield in large breeding populations.

The field phenotyping platform's next darling: Dicotyledons

The genetic information and functional properties of plants have been further identified with the completion of the whole-genome sequencing of numerous crop species and the rapid development of high-throughput phenotyping technologies, laying a suitable foundation for advanced precision agriculture and enhanced genetic gains. Collecting phenotypic data from dicotyledonous crops in the field has been identified as a key factor in the collection of large-scale phenotypic data of crops. On the one hand, dicotyledonous plants account for 4/5 of all angiosperm species and play a critical role in agriculture. However, their morphology is complex, and an abundance of dicot phenotypic information is available, which is critical for the analysis of high-throughput phenotypic data in the field. As a result, the focus of this paper is on the major advancements in ground-based, air-based, and space-based field phenotyping platforms over the last few decades and the research progress in the high-throughput phenotyping of dicotyledonous field crop plants in terms of morphological indicators, physiological and biochemical indicators, biotic/abiotic stress indicators, and yield indicators. Finally, the future development of dicots in the field is explored from the perspectives of identifying new unified phenotypic criteria, developing a high-performance infrastructure platform, creating a phenotypic big data knowledge map, and merging the data with those of multiomic techniques.

SlypNet: Spikelet-based yield prediction of wheat using advanced plant phenotyping and computer vision techniques

The application of computer vision in agriculture has already contributed immensely to restructuring the existing field practices starting from the sowing to the harvesting. Among the different plant parts, the economic part, the yield, has the highest importance and becomes the ultimate goal for the farming community. It depends on many genetic and environmental factors, so this curiosity about knowing the yield brought several precise pre-harvest prediction methods using different ways. Out of those techniques, non-invasive yield prediction techniques using computer vision have been proved to be the most efficient and trusted platform. This study developed a novel methodology, called SlypNet, using advanced deep learning networks, i.e., Mask R-CNN and U-Net, which can extract various plant morphological features like spike and spikelet from the visual image of the wheat plant and provide a high-throughput yield estimate with great precision. Mask R-CNN outperformed previous networks in spike detection by its precise detection performance with a mean average precision (mAP) of 97.57%, a F1 score of 0.67, and an MCC of 0.91 by overcoming several natural field constraints like overlapping and background interference, variable resolution, and high bushiness of plants. The spikelet detection module's accuracy and consistency were tested with about 99% validation accuracy of the model and the least error, i.e., a mean square error of 1.3 from a set of typical and complex views of wheat spikes. Spikelet yield cumulatively showed the probable production capability of each plant. Our method presents an integrated deep learning platform of spikelet-based yield prediction comprising spike and spikelet detection, leading to higher precision over the existing methods.

Selective transplantation method of leafy vegetable seedlings based on ResNet 18 network

To solve the problem of low survival rate caused by unscreened transplanting of seedlings. This study proposed a selective transplanting method of leafy vegetable seedlings based on the ResNet 18 network. Lettuce seedlings were selected as the research object, and a total of 3,388 images were obtained in the dataset. The images were randomly divided into the training set, validation set, and test set in the ratio of 6:2:2. The ResNet 18 network was used to perform transfer learning after tuning, identifying, and classifying leafy vegetable seedlings, and then establishing a model to screen leafy vegetable seedlings. The results showed that the optimal detection accuracy of the presence and health of seedlings in the training data set was above 100%, and the model loss remained at around 0.005. Nine hundred seedlings were selected for the validation test, and the screening accuracy rate was 97.44%, the precision rate of healthy seedlings was 97.56%, the recall rate was 97.34%, the precision rate of unhealthy seedlings was 92%, and the recall rate was 92.62%, which was better than the screening model based on the physical characteristics of seedlings. If they were identified as unhealthy seedlings, the manipulator would remove them during the transplanting process and perform the seedling replenishment operation to increase the survival rate of the transplanted seedlings. Moreover, the seedling image is extracted by background removal technology, so the model processing time for a single image is only 0.0129 s. This research will provide technical support for the selective transplantation of leafy vegetable seedlings.

Temporal Variation and Component Allocation Characteristics of Geometric and Physical Parameters of Maize Canopy for the Entire Growing Season

The accurate monitoring of crop parameters is important for crop yield prediction and canopy parameter inversion from remote sensing. Process-based and semi-empirical crop models are the main approaches to modeling the temporal changes in crop parameters. However, the former requires too many input parameters and the latter has the problem of poor portability. In this study, new semi-empirical geometric and physical parameters of the maize canopy model (GPMCM) crop model adapted to northeast China were proposed based on a time-series field datasets collected from 11 sites in the Nong’an and Changling Counties of Jilin Province, China, during DOY (day of year) 163 to DOY 278 in 2021. The allocation characteristics of and correlations between each maize canopy parameter were investigated for the whole growing season using the 22 algorithms of crop parameters, and the following conclusions were obtained. (1) The high correlation coefficient (R mean = 0.79) of LAI with other canopy parameters indicated that it was a good indicator for predicting other parameters. (2) Better performance was achieved by the regression method based on the two-stage simulation. The root-mean-squared error (RMSE) of geometric parameters including maize height, stem long radius, and short radius were 12.91 cm, 0.74 mm, and 0.73 mm, respectively, and the RMSE of the physical parameters including the FAGB, AGB, VWC, and RWC of the stems and leaves, ranged from 0.05 kg/m2 to 4.24 kg/m2 (2.0% to 12.9% for mean absolute percentage error (MAPE)). (3) The extension of the field-scale GPMCM to the 500 m MODIS-scale still provided a good accuracy (MAPE: 11% to 18.5%) and confirmed the feasibility of the large-scale application of the GPMCM. The proposed CPMCM can predict the temporal dynamics of maize geometric and physical parameters, and it is helpful to establish the forward and reverse models of remote sensing and improve the inversion accuracy of crop parameters.

Construction of Remote Sensing Model of Fresh Corn Biomass Based on Neural Network

Corn has a high yield and is widely used. Therefore, developing corn production and accurately estimating corn biomass yield are of great significance to improving people's lives, developing rural economy and climate issues. In this paper, a 3-layer BP neural network model is constructed by using the LM algorithm as the training algorithm of the corn biomass BP network model. From the three aspects of elevation, slope, and aspect, combined with the BP neural network model of corn biomass, the spatial distribution of corn biomass in the study area is analyzed. The results showed that the average biomass per unit area of maize increased with the increase in altitude below 1000 m. There are relatively more human activities in low altitude areas, which are more active in forestry production. The best planting altitude of corn is 0 ∼ 1000 m. When the altitude is higher than 1000 m, the corn biomass gradually decreases. In terms of slope, if the slope is lower than 15°, the biomass of maize increases with the increase in slope. If the slope is lower than 15°, the biomass of maize decreases gradually with the increase in slope. The biomass of maize on sunny slope was higher than that on shady slope.

Application of Convolutional Neural Network-Based Detection Methods in Fresh Fruit Production: A Comprehensive Review

As one of the representative algorithms of deep learning, a convolutional neural network (CNN) with the advantage of local perception and parameter sharing has been rapidly developed. CNN-based detection technology has been widely used in computer vision, natural language processing, and other fields. Fresh fruit production is an important socioeconomic activity, where CNN-based deep learning detection technology has been successfully applied to its important links. To the best of our knowledge, this review is the first on the whole production process of fresh fruit. We first introduced the network architecture and implementation principle of CNN and described the training process of a CNN-based deep learning model in detail. A large number of articles were investigated, which have made breakthroughs in response to challenges using CNN-based deep learning detection technology in important links of fresh fruit production including fruit flower detection, fruit detection, fruit harvesting, and fruit grading. Object detection based on CNN deep learning was elaborated from data acquisition to model training, and different detection methods based on CNN deep learning were compared in each link of the fresh fruit production. The investigation results of this review show that improved CNN deep learning models can give full play to detection potential by combining with the characteristics of each link of fruit production. The investigation results also imply that CNN-based detection may penetrate the challenges created by environmental issues, new area exploration, and multiple task execution of fresh fruit production in the future.

Machine-Learning-Based Genome-Wide Association Studies for Uncovering QTL Underlying Soybean Yield and Its Components

A genome-wide association study (GWAS) is currently one of the most recommended approaches for discovering marker-trait associations (MTAs) for complex traits in plant species. Insufficient statistical power is a limiting factor, especially in narrow genetic basis species, that conventional GWAS methods are suffering from. Using sophisticated mathematical methods such as machine learning (ML) algorithms may address this issue and advance the implication of this valuable genetic method in applied plant-breeding programs. In this study, we evaluated the potential use of two ML algorithms, support-vector machine (SVR) and random forest (RF), in a GWAS and compared them with two conventional methods of mixed linear models (MLM) and fixed and random model circulating probability unification (FarmCPU), for identifying MTAs for soybean-yield components. In this study, important soybean-yield component traits, including the number of reproductive nodes (RNP), non-reproductive nodes (NRNP), total nodes (NP), and total pods (PP) per plant along with yield and maturity, were assessed using a panel of 227 soybean genotypes evaluated at two locations over two years (four environments). Using the SVR-mediated GWAS method, we were able to discover MTAs colocalized with previously reported quantitative trait loci (QTL) with potential causal effects on the target traits, supported by the functional annotation of candidate gene analyses. This study demonstrated the potential benefit of using sophisticated mathematical approaches, such as SVR, in a GWAS to complement conventional GWAS methods for identifying MTAs that can improve the efficiency of genomic-based soybean-breeding programs.

Phenotypic Traits Estimation and Preliminary Yield Assessment in Different Phenophases of Wheat Breeding Experiment Based on UAV Multispectral Images

The utility of unmanned aerial vehicles (UAV) imagery in retrieving phenotypic data to support plant breeding research has been a topic of increasing interest in recent years. The advantages of image-based phenotyping are related to the high spatial and temporal resolution of the retrieved data and the non-destructive and rapid method of data acquisition. This study trains parametric and nonparametric regression models to retrieve leaf area index (LAI), fraction of absorbed photosynthetically active radiation (fAPAR), fractional vegetation cover (fCover), leaf chlorophyll content (LCC), canopy chlorophyll content (CCC), and grain yield (GY) of winter durum wheat breeding experiment from four-bands UAV images. A ground dataset, collected during two field campaigns and complemented with data from a previous study, is used for model development. The dataset is split at random into two parts, one for training and one for testing the models. The tested parametric models use the vegetation index formula and parametric functions. The tested nonparametric models are partial least square regression (PLSR), random forest regression (RFR), support vector regression (SVR), kernel ridge regression (KRR), and Gaussian processes regression (GPR). The retrieved biophysical variables along with traditional phenotypic traits (plant height, yield, and tillering) are analysed for detection of genetic diversity, proximity, and similarity in the studied genotypes. Analysis of variance (ANOVA), Duncan’s multiple range test, correlation analysis, and principal component analysis (PCA) are performed with the phenotypic traits. The parametric and nonparametric models show close results for GY retrieval, with parametric models indicating slightly higher accuracy (R2 = 0.49; MRSE = 0.58 kg/plot; rRMSE = 6.1%). However, the nonparametric model, GPR, computes per pixel uncertainty estimation, making it more appealing for operational use. Furthermore, our results demonstrate that grain filling was better than flowering phenological stage to predict GY. The nonparametric models show better results for biophysical variables retrieval, with GPR presenting the highest prediction performance. Nonetheless, robust models are found only for LAI (R2 = 0.48; MRSE = 0.64; rRMSE = 13.5%) and LCC (R2 = 0.49; MRSE = 31.57 mg m−2; rRMSE = 6.4%) and therefore these are the only remotely sensed phenotypic traits included in the statistical analysis for preliminary assessment of wheat productivity. The results from ANOVA and PCA illustrate that the retrieved remotely sensed phenotypic traits are a valuable addition to the traditional phenotypic traits for plant breeding studies. We believe that these preliminary results could speed up crop improvement programs; however, stronger interdisciplinary research is still needed, as well as uncertainty estimation of the remotely sensed phenotypic traits.

Background-Aware Domain Adaptation for Plant Counting

Deep learning-based object counting models have recently been considered preferable choices for plant counting. However, the performance of these data-driven methods would probably deteriorate when a discrepancy exists between the training and testing data. Such a discrepancy is also known as the domain gap. One way to mitigate the performance drop is to use unlabeled data sampled from the testing environment to correct the model behavior. This problem setting is also called unsupervised domain adaptation (UDA). Despite UDA has been a long-standing topic in machine learning society, UDA methods are less studied for plant counting. In this paper, we first evaluate some frequently-used UDA methods on the plant counting task, including feature-level and image-level methods. By analyzing the failure patterns of these methods, we propose a novel background-aware domain adaptation (BADA) module to address the drawbacks. We show that BADA can easily fit into object counting models to improve the cross-domain plant counting performance, especially on background areas. Benefiting from learning where to count, background counting errors are reduced. We also show that BADA can work with adversarial training strategies to further enhance the robustness of counting models against the domain gap. We evaluated our method on 7 different domain adaptation settings, including different camera views, cultivars, locations, and image acquisition devices. Results demonstrate that our method achieved the lowest Mean Absolute Error on 6 out of the 7 settings. The usefulness of BADA is also supported by controlled ablation studies and visualizations.

Study on the Classification Method of Rice Leaf Blast Levels Based on Fusion Features and Adaptive-Weight Immune Particle Swarm Optimization Extreme Learning Machine Algorithm

Leaf blast is a disease of rice leaves caused by the Pyricularia oryzae. It is considered a significant disease is affecting rice yield and quality and causing economic losses to food worldwide. Early detection of rice leaf blast is essential for early intervention and limiting the spread of the disease. To quickly and non-destructively classify rice leaf blast levels for accurate leaf blast detection and timely control. This study used hyperspectral imaging technology to obtain hyperspectral image data of rice leaves. The descending dimension methods got rice leaf disease characteristics of different disease classes, and the disease characteristics obtained by screening were used as model inputs to construct a model for early detection of leaf blast disease. First, three methods, ElasticNet, principal component analysis loadings (PCA loadings), and successive projections algorithm (SPA), were used to select the wavelengths of spectral features associated with leaf blast, respectively. Next, the texture features of the images were extracted using a gray level co-occurrence matrix (GLCM), and the texture features with high correlation were screened by the Pearson correlation analysis. Finally, an adaptive-weight immune particle swarm optimization extreme learning machine (AIPSO-ELM) based disease level classification method is proposed to further improve the model classification accuracy. It was also compared and analyzed with a support vector machine (SVM) and extreme learning machine (ELM). The results show that the disease level classification model constructed using a combination of spectral characteristic wavelengths and texture features is significantly better than a single disease feature in terms of classification accuracy. Among them, the model built with ElasticNet + TFs has the highest classification accuracy, with OA and Kappa greater than 90 and 87%, respectively. Meanwhile, the AIPSO-ELM proposed in this study has higher classification accuracy for leaf blast level classification than SVM and ELM classification models. In particular, the AIPSO-ELM model constructed with ElasticNet+TFs as features obtained the best classification performance, with OA and Kappa of 97.62 and 96.82%, respectively. In summary, the combination of spectral characteristic wavelength and texture features can significantly improve disease classification accuracy. At the same time, the AIPSO-ELM classification model proposed in this study has sure accuracy and stability, which can provide a reference for rice leaf blast disease detection.

Genome-Wide Association Study Statistical Models: A Review

Statistical models are at the core of the genome-wide association study (GWAS). In this chapter, we provide an overview of single- and multilocus statistical models, Bayesian, and machine learning approaches for association studies in plants. These models are discussed based on their basic methodology, cofactors adjustment accounted for, statistical power and computational efficiency. New statistical models and machine learning algorithms are both showing improved performance in detecting missed signals, rare mutations and prioritizing causal genetic variants; nevertheless, further optimization and validation studies are required to maximize the power of GWAS.

Genome-Wide Association Studies of Soybean Yield-Related Hyperspectral Reflectance Bands Using Machine Learning-Mediated Data Integration Methods

In conjunction with big data analysis methods, plant omics technologies have provided scientists with cost-effective and promising tools for discovering genetic architectures of complex agronomic traits using large breeding populations. In recent years, there has been significant progress in plant phenomics and genomics approaches for generating reliable large datasets. However, selecting an appropriate data integration and analysis method to improve the efficiency of phenome-phenome and phenome-genome association studies is still a bottleneck. This study proposes a hyperspectral wide association study (HypWAS) approach as a phenome-phenome association analysis through a hierarchical data integration strategy to estimate the prediction power of hyperspectral reflectance bands in predicting soybean seed yield. Using HypWAS, five important hyperspectral reflectance bands in visible, red-edge, and near-infrared regions were identified significantly associated with seed yield. The phenome-genome association analysis of each tested hyperspectral reflectance band was performed using two conventional genome-wide association studies (GWAS) methods and a machine learning mediated GWAS based on the support vector regression (SVR) method. Using SVR-mediated GWAS, more relevant QTL with the physiological background of the tested hyperspectral reflectance bands were detected, supported by the functional annotation of candidate gene analyses. The results of this study have indicated the advantages of using hierarchical data integration strategy and advanced mathematical methods coupled with phenome-phenome and phenome-genome association analyses for a better understanding of the biology and genetic backgrounds of hyperspectral reflectance bands affecting soybean yield formation. The identified yield-related hyperspectral reflectance bands using HypWAS can be used as indirect selection criteria for selecting superior genotypes with improved yield genetic gains in large breeding populations.

Assessing the Impacts of Species Composition on the Accuracy of Mapping Chlorophyll Content in Heterogeneous Ecosystems

Chlorophyll is an essential vegetation pigment influencing plant photosynthesis rate and growth conditions. Remote sensing images have been widely used for mapping vegetation chlorophyll content in different ecosystems (e.g., farmlands, forests, grasslands, and wetlands) for evaluating vegetation growth status and productivity of these ecosystems. Compared to farmlands and forests that are more homogeneous in terms of species composition, grasslands and wetlands are more heterogeneous with highly mixed species (e.g., various grass, forb, and shrub species). Different species contribute differently to the ecosystem services, thus, monitoring species-specific chlorophyll content is critical for better understanding their growth status, evaluating ecosystem functions, and supporting ecosystem management (e.g., control invasive species). However, previous studies in mapping chlorophyll content in heterogeneous ecosystems have rarely estimated species-specific chlorophyll content, which was partially due to the limited spatial resolution of remote sensing images commonly used in the past few decades for recognizing different species. In addition, many previous studies have used one universal model built with data of all species for mapping chlorophyll of the entire study area, which did not fully consider the impacts of species composition on the accuracy of chlorophyll estimation (i.e., establishing species-specific chlorophyll estimation models may generate higher accuracy). In this study, helicopter-acquired high-spatial resolution hyperspectral images were acquired for species classification and species-specific chlorophyll content estimation. Four estimation models, including a universal linear regression (LR) model (i.e., built with data of all species), species-specific LR models (i.e., built with data of each species, respectively), a universal random forest regression (RFR) model, and species-specific RFR models, were compared to determine their performance in mapping chlorophyll and to evaluate the impacts of species composition. The results show that species-specific models performed better than the universal models, especially for species with fewer samples in the dataset. The best performed species-specific models were then used to generate species-specific chlorophyll content maps using the species classification results. Impacts of species composition on the retrieval of chlorophyll content were further assessed to support future chlorophyll mapping in heterogeneous ecosystems and ecosystem management.

Machine Learning-Mediated Development and Optimization of Disinfection Protocol and Scarification Method for Improved In Vitro Germination of Cannabis Seeds

In vitro seed germination is a useful tool for developing a variety of biotechnologies, but cannabis has presented some challenges in uniformity and germination time, presumably due to the disinfection procedure. Disinfection and subsequent growth are influenced by many factors, such as media pH, temperature, as well as the types and levels of contaminants and disinfectants, which contribute independently and dynamically to system complexity and nonlinearity. Hence, artificial intelligence models are well suited to model and optimize this dynamic system. The current study was aimed to evaluate the effect of different types and concentrations of disinfectants (sodium hypochlorite, hydrogen peroxide) and immersion times on contamination frequency using the generalized regression neural network (GRNN), a powerful artificial neural network (ANN). The GRNN model had high prediction performance (R² > 0.91) in both training and testing. Moreover, a genetic algorithm (GA) was subjected to the GRNN to find the optimal type and level of disinfectants and immersion time to determine the best methods for contamination reduction. According to the optimization process, 4.6% sodium hypochlorite along with 0.008% hydrogen peroxide for 16.81 min would result in the best outcomes. The results of a validation experiment demonstrated that this protocol resulted in 0% contamination as predicted, but germination rates were low and sporadic. However, using this sterilization protocol in combination with the scarification of in vitro cannabis seed (seed tip removal) resulted in 0% contamination and 100% seed germination within one week.