With the rapid development of artificial intelligence, UAV remote sensing, and agricultural Internet of Things technologies, crop growth monitoring is evolving from manual inspection and single-source analysis toward intelligent decision-making based on multisource perception. However, existing methods still suffer from limited robustness under environmental variations, insufficient integration between UAV imagery and sparse ground sensor observations, and weak capability for transforming predictions into practical agricultural management recommendations. This study proposes a UAV–ground sensor collaborative lightweight framework for crop growth assessment and agricultural decision support. The proposed framework integrates UAV RGB and multispectral imagery with ground sensor observations through a region-level aerial–ground alignment mechanism and a sensor-guided attention fusion module, enabling environmental conditions to enhance visual feature interpretation. Furthermore, a fact-constrained decision module is developed to generate management recommendations based on crop status, environmental risks, and field information. Experimental results demonstrate that the proposed method achieves superior performance in crop growth classification and yield-trend prediction, reaching Accuracy, Precision, Recall, and F1-score values of $92.47\%$, $91.86\%$, $91.39\%$, and $91.62\%$, respectively, with an RMSE of $0.381$ and an $R^2$ of $0.902$. The lightweight framework requires only $6.18$M parameters and $0.91$G FLOPs, achieving $39.56$ ms inference latency and $25.28$ FPS on edge devices. The proposed framework also improves decision reliability, achieving an expert agreement rate of $89.34\%$ and a risk identification accuracy of $90.18\%$. Economic analysis indicates that the proposed framework reduces labor cost, water consumption, and fertilizer input by $49.7\%$, $26.7\%$, and $23.0\%$, respectively, while increasing net benefit by $46.1\%$ compared with conventional field management practices. These results demonstrate that the proposed method provides an accurate, interpretable, and deployable AI-driven solution for intelligent crop management in smallholder and medium-sized farming systems.

Smart greenhouse management requires decisions that balance yields, costs, resource use, emissions, and revenue. Existing threshold-based controls and many learning models use sensor data mainly for environmental feedback or yield prediction, leaving limited support for operational decisions tied to farm profitability and resource efficiency. This study proposes AER-SFRDNet, a multisource sensor fusion and deep reinforcement learning framework for revenue-aware greenhouse regulation. Crop images, environmental sensor sequences, management records, resource consumption data, and economic variables from facility tomato production are integrated into a unified agro-economic state representation. A yield–cost–revenue prediction module estimates the yield, resource consumption, production cost, sales revenue, and net revenue. These outputs guide a reinforcement learning module that optimizes continuous actions including irrigation, ventilation, supplemental lighting, heating, and CO₂ application. A comparison with traditional machine learning, unimodal deep learning, multimodal learning, and reinforcement learning baselines shows lower prediction errors and improved regulation outcomes. AER-SFRDNet achieves a yield MAE of $0.486$, yield RMSE of $0.674$, yield $R^2$ of $0.893$, cost RMSE of $0.613$, and net revenue MAPE of $10.82$. In regulation experiments, the cumulative net revenue and input–output ratio reach $1.276$ and $1.263$, while the normalized energy use, water use, and carbon emissions per unit yield decrease to $0.752$, $0.781$, and $0.741$, respectively. The results suggest that agro-economic state modeling can support greenhouse decisions that consider both revenue and sustainability-related resource efficiency.

Fruit quality is jointly influenced by preharvest growing conditions, harvest status, and postharvest storage environments, whereas existing approaches generally focus on individual stages and therefore remain limited in simultaneously supporting cross-stage quality prediction, economic value assessment, and practical operational decision-making. To address the requirements of smart orchard management and postharvest quality control, Pre2Post-EcoLLM is proposed as a unified framework integrating fruit and canopy RGB images, environmental and soil sensing data, postharvest temperature and relative humidity, carbon dioxide and oxygen (${\mathrm{CO}}_2/{\mathrm{O}}_2$), volatile organic compound (VOC) measurements, management records, and economic information. First, the Growth-to-Harvest Multimodal Representation module constructs a Harvest Quality State through growth-stage-aware cross-modal interaction. Subsequently, the Cross-Stage Quality–Economic Co-Evolution module continuously injects the harvest state into the postharvest temporal evolution process to jointly predict soluble solids content (SSC), firmness, weight loss, decay risk, and remaining shelf life, while further estimating commercial grade, saleable rate, economic loss, and net return. Finally, the Evidence-Grounded Agricultural Large Language Model (LLM) QA module combines structured model evidence, retrieval-augmented generation (RAG)-retrieved agricultural knowledge, and outputs from the Economic Calculator to provide traceable natural-language explanations for storage and selling decisions. Experimental results demonstrate that Pre2Post-EcoLLM achieved an overall normalized mean absolute error (MAE) of $0.247$, a root mean square error (RMSE) of $0.394$, a mean absolute percentage error (MAPE) of $7.29\%$, and an $R^2$ of $0.916$ for harvest and postharvest quality prediction, with disaggregated MAEs of $0.42 °\mathrm{Brix}$ for SSC, $0.38\mathrm{kg}/{\mathrm{cm}}^2$ for firmness, and $0.31\%$ for weight loss rate, while the decay-risk F1-score and area under the receiver operating characteristic curve (AUC) reached $0.907$ and $0.954$, respectively. The remaining shelf-life MAE was reduced to $1.68\mathrm{days}$, while the Economic MAE and Economic $R^2$ reached $0.066(\mathrm{normalized}\mathrm{ratio},\mathrm{equivalent}\mathrm{to}561\mathrm{CNY}/\mathrm{ton})$ and $0.921$, respectively. Storage–sale Decision Accuracy and Decision F1 reached $91.7\%$ and $90.9\%$, respectively. For agricultural question answering, Answer Accuracy and Faithfulness reached $92.4\%$ and $95.1\%$, respectively, while the Hallucination Rate was reduced to $4.9\%$. Ablation experiments further confirmed the effectiveness of growth-stage-aware fusion, cross-stage state injection, temporal quality evolution, economic coupling, and evidence-grounded question answering. These results indicate that the proposed framework can effectively connect preharvest sensing, postharvest quality evolution, and economic decision-making, providing a practical intelligent solution for lifecycle fruit quality assessment and smart storage management.

Water and fertilizer management in facility and precision agriculture increasingly depends on images, soil sensors, weather records, management logs, and economic data rather than on empirical operation alone. A cross-modal decision framework is developed to align crop visual traits with soil water–salt dynamics, meteorological variation, and management operations. Image preprocessing, abnormal sensor screening, missing value interpolation, temporal resampling, multi-window alignment, and temporal augmentation are used to form comparable inputs before multimodal fusion. Using plot-blocked evaluation that keeps complete plot trajectories together, AgriWFD-RLNet obtains 93.46% status-recognition accuracy and an $R^2$ of 0.923 for yield–resource–economic response prediction. In offline policy evaluation against the logged conventional management reference, it estimates 16.92% water saving, 14.37% fertilizer saving, and 15.84% Net Return improvement while maintaining 94.68% yield stability and 96.72% safety compliance. These decision outcomes are model-based estimates from the 2024 Wuyuan dataset rather than outcomes of a prospective field deployment.