In response to the challenges of detecting rice pests and diseases at different scales and the difficulties associated with deploying and running models on embedded devices with limited computational resources, this study proposes a multi-scale rice pest and disease recognition model (RGC-YOLO). Based on the YOLOv8n network, which includes an SPPF layer, the model introduces a structural reparameterization module (RepGhost) to achieve implicit feature reuse through reparameterization. GhostConv layers replace some standard convolutions, reducing the model’s computational cost and improving inference speed. A Hybrid Attention Module (CBAM) is incorporated into the backbone network to enhance the model’s ability to extract important features. The RGC-YOLO model is evaluated for accuracy and inference time on a multi-scale rice pest and disease dataset, including bacterial blight, rice blast, brown spot, and rice planthopper. Experimental results show that RGC-YOLO achieves a precision (P) of 86.2%, a recall (R) of 90.8%, and a mean average precision at Intersection over Union 0.5(mAP50) of 93.2%. In terms of model size, the parameters are reduced by 33.2%, and GFLOPs decrease by 29.27% compared to the base YOLOv8n model. Finally, the RGC-YOLO model is deployed on an embedded Jetson Nano device, where the inference time per image is reduced by 21.3% compared to the base YOLOv8n model, reaching 170 milliseconds. This study develops a multi-scale rice pest and disease recognition model, which is successfully deployed on embedded field devices, achieving high-accuracy real-time monitoring and providing valuable reference for intelligent equipment in unmanned farms.

Due to nutrient stress, which is an important constraint to the development of the global agricultural sector, it is now vital to timely evaluate plant health. Remote sensing technology, especially hyperspectral imaging technology, has evolved from spectral response modes to pattern recognition and vegetation monitoring. This study established a hyperspectral library of 14 NPK (nitrogen, phosphorus, potassium) nutrient stress conditions in rice. The terrestrial hyperspectral camera (SPECIM-IQ) collected 420 rice stress images and extracted as well as analyzed representative spectral reflectance curves under 14 stress modes. The canopy spectral profile characteristics, vegetation index, and principal component analysis demonstrated the differences in rice under different nutrient stresses. A transformer-based deep learning network SHCFTT (SuperPCA-HybridSN-CBAM-Feature tokenization transformer) was established for identifying nutrient stress patterns from hyperspectral images while being compared with classic support vector machines, 1D-CNN (1D-Convolutional Neural Network), and 3D-CNN. The total accuracy of the SHCFTT model under different modeling strategies and different years ranged from 93.92% to 100%, indicating the positive effect of the proposed method on improving the accuracy of identifying nutrient stress in rice.

Weeds contribute to biodiversity and a wide range of ecosystem functions. It is crucial to map different weed species and analyze their physiological activities. Remote sensing techniques for plant identification, especially hyperspectral imaging, are being developed using spectral response patterns to vegetation for detection and species identification. A library of hyperspectral images of 40 urban weed species in northeast China was established in this study. A terrestrial hyperspectral camera was used to acquire 435 hyperspectral images. The hyperspectral information for each weed species was extracted and analyzed. The spectral characteristics and vegetation indices of different weeds revealed the differences between weed species in the cities of northeast China and indirectly characterized the growth and physiological activity levels of different species, but could not effectively distinguish different species. Five methods—first derivative spectrum (FDS), second derivative spectrum (SDS), standard normal variate (SNV), moving averages (MA), and Savitzky-Golay (SG) smoothing—were used to pretreat the spectral curves to maximize the retention of spectral characteristics while removing the influence of noise. We investigated the application of a convolutional neural network (CNN) with terrestrial hyperspectral remote sensing to identify urban weeds in northeast China. A CNN classification model was established to distinguish weeds from the hyperspectral images and demonstrated a test accuracy of 95.32% to 98.15%. The accuracy of the original spectrum was 97.45%; SNV had the best accuracy (98.15%) and SG was the least accurate (95.32%). This provides a baseline for understanding the hyperspectral characteristics of urban weed species and monitoring their growth. It also contributes to the development of a hyperspectral imaging database with global applicability.

Accurate estimation of wheat leaf nitrogen concentration (LNC) is critical for characterizing ecosystem and plant physiological processes; it can further guide fertilization and other field management operations, and promote the sustainable development of agriculture. In this study, a wheat LNC test method based on multi-source spectral data and a convolutional neural network is proposed. First, interpolation reconstruction was performed on the wheat spectra data collected by different spectral instruments to ensure that the number of spectral channels and spectral range were consistent, and multi-source spectral data were constructed using interpolated, reconstructed imaging spectral data and non-imaging spectral data. Afterwards, the convolutional neural network DshNet and machine learning methods (PLSR, SVR, and RFR) were compared under various scenarios (non-imaging spectral data, imaging spectral data, and multi-source spectral data). Finally, the competitive adaptive reweighted sampling (CARS) and successive projections algorithm (SPA) were used to optimize the LNC detection model. The results show that the model based on DshNet has the highest test accuracy. The CARS method is more suitable for DshNet model optimization than SPA. In the modeling scenario with non-imaging spectral, imaging spectral, and multi-source spectral, the optimized R² is 0.86, 0.82, and 0.82, and the RMSE is 0.29, 0.31, and 0.31, respectively. The LNC visualization results show that DshNet modeling using multi-source spectral data is conducive to the visualization expansion of non-imaging spectral data. Therefore, the method presented in this paper provides new considerations for spectral data from different sources and is helpful for related research on the chemometric task of multi-source spectral data.

Aiming at the issues of complex manufacturing processes and unstable bonding after individual manufacturing in current soft manipulator forming methods, this study investigated the mechanical characteristics of a pneumatically driven soft-picking manipulator formed by the stereolithography (SLA) process and evaluated its application in grasping fruits and vegetables. The soft-picking manipulator mainly consists of three soft actuators designed in an integrated folded structure to simplify the manufacturing process compared to a conventional one. A finite element model (FEM) of the actuator was created to analyze the bending deformation capability under different pressures, and the simulated results match well with the experimental ones. Under the 60 kPa pressure, the maximum grasping force for fingertip- and envelope-grasping is 3.94 N and 8.87 N, respectively. The grasping tests of several fruits and vegetables of different weights and sizes by the soft manipulator were examined, and the results showed that the manipulator has strong adaptability. For spherical and elongated fruits and vegetables, the completion time for grasping is approximately 8.59 ± 1.26 s and 10.99 ± 1.79 s, respectively, and for irregularly shaped ones, the pressure is increased accordingly to the increased grasping stability. This study may provide a basis for the development of a soft manipulator for sorting and picking fruits and vegetables.