Target and off-target spray depositions determine the spray’s effectiveness and impact on the environment. A decisive stage in the measurement of spray deposition and drift is selecting an appropriate sampling approach under field conditions. There are various approaches available for sampling spray deposition and drift, during the evaluation of ground sprayers used for the UAV sprayer assessment, under field conditions. In this study, two sampling approaches (water-sensitive paper, and glass strip collectors) were compared to analyze spray deposition in target and off-target zones. The results showed a variation in the estimation of the spray deposits among the two applied sampling methods. The results showed that the water-sensitive paper recorded the droplet deposition in the target zone with a range from 0.049 to 4.866 µLcm⁻², whereas the glass strip recorded from 0.11 to 0.793 µLcm⁻². The results also showed the water sensitive paper recorded an 80.3% higher deposition than that of the glass strip at zero position during the driving flight height 2 m and flight speed 2 ms⁻¹ (T1 treatment). It can be concluded that variation in recorded depositing is due to the sampling material. It is recommended that the confident deposition results, measurement methods and sampling approaches must be standardized for UAV sprayers according to the field conditions and controlled within artificial assessments.

Manual mango variety classification is time-consuming, error-prone, and contributes significantly to post-harvest losses in developing economies. This study aims to develop a computationally efficient and highly accurate artificial intelligence framework for automated mango variety classification suitable for real-time applications. Eight deep transfer learning models were evaluated as feature extractors and combined with ten classical machine-learning classifiers. Model performance was assessed using accuracy, log loss, memory usage, training time, and inference latency. The hybrid models EfficientNetB0–Linear Discriminant Analysis (LDA) and ResNet50–Logistic Regression achieved 100% test accuracy while reducing inference time by up to 330 times compared to full Convolutional Neural Network (CNN) models. These findings demonstrate that hybrid deep-learning and machine-learning architectures can deliver state-of-the-art accuracy with substantially lower computational cost. Future research will focus on large-scale real-world validation and embedded hardware deployment for industrial fruit sorting systems.

The world food security needs the adoption of effective and sustainable agricultural practices, which is affected by factors like population growth, labor shortage, and rising production costs. Especially in vegetable farming that is an important source of nutrients is impacted. Plug seedling systems, where transplants are cultivated in regulated greenhouse conditions, have much to recommend them, such as lower establishment costs, more efficient resource use (water, nutrients, pesticides) and better uniformity of crops. However, cultivating and transplanting of these seedlings are laborious activities, which require precision. This review explores the critical role of agriculture automation technologies in solving such challenges. In particular, we review the results of the current automated solutions used along the plug seedling production chain, such as automatic seedling quality evaluation, machine vision-controlled robotics to monitor and grade quality, robotic, and automated transplanting systems. Analysis of the findings demonstrate that these technologies can be helpful in increasing productivity, maintaining high quality and greatly decreasing reliance on manual labor. Lastly, we recognize existing constraints and future trends in order to offer future research and development directions in the field of smart greenhouse automation.

Post-harvest losses of mango (Mangifera indica L.) in developing economies are estimated at 5% to 30%, largely due to manual management practices that depend on subjective visual assessments. This paper proposes a lightweight deep learning architecture, termed SE-DBIRNet, for real-time classification of mangoes into five shelf-life stages: unripe, semi-ripe, fully ripe, overripe, and perished. The model incorporates three key design strategies: (i) depthwise separable convolutions, achieving an 88.5% reduction in parameters when integrated into the ResNet50 backbone; (ii) a double-branch inverted residual (DBIR) module designed to enhance feature diversity and richness; and (iii) a squeeze-and-excitation (SE) attention mechanism for adaptive channel-wise recalibration. Using a public benchmark dataset of 4428 RGB images (Mendeley Data) under 10-fold cross-validation, SE-DBIRNet achieved 98.24% accuracy. Among lightweight CNN architectures (EfficientNetB0, MobileNetV2, ResNet50), SE-DBIRNet outperformed the best lightweight baseline (EfficientNetB0: 96.57%) by 1.67 percentage points. While dedicated attention-based DenseNet variants (e.g., DSA-DenseNet: 99.20%) achieved higher accuracy, SE-DBIRNet offers a superior trade-off among accuracy, inference speed (56.9 ± 1.8 FPS), and memory efficiency (8871 ± 45 MB CPU memory). EigenCAM activation visualizations revealed that the model focuses on biologically relevant and stage-discriminative features, including surface color gradients, texture uniformity, lenticel patterns, and decay boundaries. Overall, SE-DBIRNet achieves a Pareto-optimal balance among accuracy, speed, and memory efficiency, making it a strong candidate for real-time, edge-deployable post-harvest mango quality-monitoring systems, particularly when computational resources are limited.

The mechanical weeding system for paddy fields stands out as the leading nonchemical method for weed removal. However, the current weeder configurations face challenges in terms of driving stability and energy efficiency. Traditionally, optimizing the design of weeder mechanisms involves physical experiments, which are both time-consuming and energy-intensive. This study focuses on optimizing the two walk-behind weeder designs, W1 and W2, through kinematic and dynamic analyses using virtual digital twin (DT) models, allowing for rapid iterations of simulated experiments. The digital twin (DT) is a cutting-edge approach that shifts agricultural mechanization from physical assessments to virtual machine model optimization. Initially, the machine model was created in SolidWorks and then imported into MSC Adams View software for kinematic analysis. The numerical simulation of the kinematics of the weeding mechanism was examined using MSC Adams View software, and the model was validated through comparison with the theoretical solution. Subsequently, a digital twin was developed with EDEM software to conduct a three-dimensional (3D) dynamic force analysis of the soil-tool interaction. A comparative analysis of the 3D dynamic forces was carried out to optimize the weeder in terms of torque requirements and energy consumption. According to the findings of the virtual digital twin model application, the W2 weeder configuration performs optimally in paddy fields.