Digital eyes in the field: Smart pest surveillance for farmers: Review

Introduction

Pest and disease outbreaks represent one of the most critical constraints to achieving global food security and sustainable agricultural productivity. The Food and Agriculture Organization (FAO, 2023) [8] estimates that approximately 20-40% of global agricultural yields are lost annually due to pests and diseases, amounting to hundreds of billions of dollars in economic losses. Such extensive damage not only undermines production efficiency but also threatens the livelihoods of smallholder farmers, particularly in developing economies where agriculture forms the backbone of national income and employment. In India, where agriculture contributes nearly 18% to the national GDP and sustains more than half of the population, pest and pathogen-induced crop losses remain a major economic burden. Studies by Dhaliwal, Jindal, and Dhawan (2015) [7] revealed that the country suffers 25-30% yield losses annually due to insect pests, weeds, and diseases, equating to billions of rupees in economic losses across staple crops. For instance, the Fall Armyworm (Spodoptera frugiperda) invasion, first reported in India during 2018, resulted in a 45% decline in maize yields in affected regions (Prasanna et al., 2021) [37] . Similarly, the cotton bollworm complex (Helicoverpa armigera) continues to cause annual losses exceeding ₹1,500 crore, despite decades of integrated pest management efforts (Kranthi & Stone, 2020) [17] . Such outbreaks have far-reaching implications-not only reducing productivity but also influencing input costs, farmer profitability, export potential, and national food reserves. Traditional pest surveillance techniques, including manual field scouting, pheromone trapping, and visual inspection, remain the primary methods adopted across Indian agriculture (Reynolds, Chapman, & Harrington, 2017) [41] . However, these approaches are labor-intensive, timeconsuming, and spatially constrained, often producing inconsistent or delayed results. The inherent delay between pest emergence and detection allows populations to exceed the Economic Threshold Level (ETL), triggering indiscriminate pesticide use. This overreliance on chemical control contributes to pest resistance, resurgence, biodiversity loss, environmental contamination, and food safety concerns (Sharma, Singh, & Kumar, 2022) [19,45] . Additionally, the increasing influence of climate change, characterized by rising temperatures, altered rainfall patterns, and shifting pest habitats, further complicates surveillance and management strategies, necessitating more adaptive and data-driven systems. In recent years, advances in digital agriculture and precision pest management have revolutionized surveillance methodologies. Technologies such as Artificial Intelligence (AI), Internet of Things (IoT), Unmanned Aerial Vehicles (UAVs), machine vision, and remote sensing have enabled real-time, automated, and predictive pest monitoring at various spatial and temporal scales. These systems facilitate early warning, dynamic threshold modeling, and climate-based pest forecasting, thereby improving the timeliness and accuracy of decision-making (FAO, 2023; Reynolds et al., 2017) [8,41] . Integration of such technologies into Decision Support Systems (DSS) and Integrated Pest Management (IPM) frameworks allows for optimized pesticide use, cost reduction, and environmental sustainability (Sharma et al., 2022) [45] . Globally, and particularly in India, the transition toward AI-powered pest surveillance systems signifies a paradigm shift from reactive to proactive pest management. By combining field-level sensor data, satellite imagery, and machine learning algorithms, modern surveillance networks can detect, quantify, and forecast pest dynamics with unprecedented accuracy. These innovations hold immense promise in mitigating crop losses, enhancing food security, and promoting climate-resilient agricultural systems (Prasanna et al., 2021; Kranthi & Stone, 2020) [17,37] . Hence, this review aims to comprehensively explore the evolution, principles, and applications of digital and AI-based pest surveillance systems, highlighting their role in strengthening sustainable pest management and supporting the long-term transformation of global and Indian agriculture. [8,10,27,46] .

Technologies for Smart Pest Surveillance 2.1 Trap-Based Camera and Sensor Systems

Recent advances in computer vision, edge computing, and wireless sensor networks have significantly improved pest detection accuracy and operational efficiency in agricultural landscapes. These systems integrate optical imaging, environmental sensors, and AI algorithms to autonomously identify, count, and classify pests, transforming traditional traps into intelligent surveillance tools.

The Smart Pest Guardian system (developed in Chennai, India) combines Convolutional Neural Network (CNN)-based image analysis with real-time data from temperature, humidity, and light sensors. This integration enables automated identification of bollworm species and dynamic prediction of population surges, outperforming conventional pheromone traps in accuracy and timeliness (Gupta et al., 2022) [10] . The system's cloudbased dashboard supports remote monitoring and facilitates data sharing for regional pest forecasting, a key component of precision Integrated Pest Management (IPM). The BOLLWM dataset (Praveen et al., 2023) [38] represents one of the largest labeled pest image datasets from Indian cotton fields, collected under diverse climatic and lighting conditions. It supports deep learning-based bollworm recognition with enhanced robustness to field variability. This dataset underpins several mobile applications that allow farmers to capture trap images using smartphones for on-device pest classification using lightweight neural networks such as MobileNetV2 and EfficientNet-Lite. Another notable system, Jute Pest Detect, applies transfer learning on pre-trained CNN models (ResNet-50, InceptionV3, and EfficientNet) to achieve multi-class classification of 17 jute pest species with 99% accuracy, even under complex backgrounds (Pramanik et al., 2023) [36] . This innovation contributes to adaptive IPM decision support by providing species-specific action thresholds and reducing pesticide misuse.

UAV, Satellite, and Remote Sensing-Based Surveillance

The integration of Unmanned Aerial Vehicles (UAVs), satellite remote sensing, and advanced image analytics has revolutionized pest surveillance, enabling non-destructive, scalable, and real-time monitoring of crop health. These tools support early warning systems by identifying spectral signatures associated with insect and pathogen-induced stress, often before visible symptoms appear on the canopy. A landmark study in Ningxia, China, on Lycium barbarum (goji berry) plantations demonstrated the power of hyperspectral imaging for pest discrimination. Using UAV-borne sensors capturing 400-1000 nm spectral data, Sun et al. (2022) [53] developed a fully connected neural network (FCNN) that distinguished between healthy plants and those infested by gall mites and psyllids. The model achieved 96.8% classification accuracy, identifying red and NIR bands as the most sensitive regions for early stress detection. This approach allowed identification five to seven days before visual symptoms emerged, proving UAV hyperspectral imaging as an efficient early-warning mechanism for perennial crops (Sun et al., 2022) [53] . Beyond goji, similar hyperspectral approaches have been reported for tea (Zhou et al., 2023), maize (Hu et al., 2022), and citrus (Chen et al., 2024) [3] , showing that narrow-band spectral features (especially 550-750 nm) can differentiate between nutrient, disease, and pest stresses with high precision.

Remote Sensing for Crop Pest Monitoring in India

In India, where smallholder systems dominate, low-cost multispectral UAVs and open-access satellite platforms such as Sentinel-2 and Landsat-8 are increasingly applied in pest surveillance. The Indian Council of Agricultural Research (ICAR, 2022) [12] demonstrated that NDVI (Normalized Difference Vegetation Index) and red-edge spectral indices derived from Sentinel-2 data effectively detected stem borer and brown planthopper infestations in rice at least 10 days before visible canopy yellowing. Similarly, Thirupathi and Prabhakar (2021) [54] integrated Sentinel-2 imagery with field pheromone trap data for cotton bollworm monitoring across Telangana, achieving early stress mapping with ~85% predictive accuracy. More recently, Reddy et al. (2024) [39] combined time-series NDVI and temperature data from Sentinel-2 and MODIS platforms to model spatial pest risk zones in central India. Their work demonstrated that early anomalies in vegetation indices correlated strongly (R² = 0.82) with subsequent pest outbreaks, providing actionable insights for regional pest advisory systems.

Multispectral UAV Precision Spraying and Decision Support

At the field scale, multispectral UAVs have proven valuable in precision IPM. Mehta et al. (2024) [27] reported an 88% accuracy in identifying bollworm hotspots in cotton fields of Gujarat using five-band multispectral imagery (blue, green, red, rededge, NIR). The UAV data were linked with automated spotspray algorithms, reducing pesticide use by 32% while maintaining yield levels. This study highlights UAVs' potential to transition from monitoring tools to active components of precision-based pest management systems.

Integration of UAV and Satellite Data

Integration of UAV and satellite platforms offers a multi-scale surveillance framework, where UAVs provide high-resolution local diagnostics (cm-level) and satellites ensure temporal continuity and regional coverage (m-level). Hybrid workflowscombining UAV-based stress mapping with Sentinel-2 NDVI anomaly tracking-enable continuous, scalable pest monitoring that can feed into cloud-based Decision Support Systems (DSS) for forecasting and advisory dissemination. Such integration is central to India's emerging Digital Pest Surveillance Networks, being piloted in collaboration with ICAR, ISRO, and agricultural universities to support real-time crop protection recommendations.

Research Gaps and Future Prospects

Despite these advances, key challenges remain:  Spectral overlap between pest, nutrient, and disease stresses complicates discrimination.  Calibration issues under variable illumination and atmospheric conditions limit model transferability.  High costs of hyperspectral sensors and limited technical expertise restrict adoption among smallholders.  Standardized pest-specific spectral libraries for Indian crops are still lacking.  Integration into DSS requires harmonization with field data, IoT traps, and meteorological variables.

Addressing these constraints through AI-driven data fusion, lowcost sensor innovation, and policy-backed digital infrastructure will be critical for scaling remote sensing in national pest management frameworks. [15,59] . In India, Kumar et al. (2023) [18] demonstrated an innovative Edge-AI application for cashew pest and disease management. A MobileNetV2-based convolutional neural network (CNN) was deployed on low-power NVIDIA Jetson Nano devices for onfield disease recognition. The system achieved 95% classification accuracy across five major cashew pests and leaf diseases, providing real-time alerts through a mobile app interface. This on-device inference capability allowed farmers to act promptly-without needing constant internet access-thus representing a scalable model for rural precision agriculture.

Similarly, ICAR's Smart Pest Management Network (2024) [13] integrates IoT weather stations, pheromone traps, and soilmicroclimate sensors with AI algorithms to predict pest dynamics based on temperature, humidity, and host phenology. Field trials across Andhra Pradesh and Karnataka demonstrated 20-30% reductions in pesticide sprays and up to 25% savings in labor costs, while maintaining yield stability. The system's modular dashboard provides dynamic pest alerts, ETL (Economic Threshold Level) advisories, and spatial heatmaps, representing one of India's first IoT-enabled pest forecasting frameworks.

Global Edge-AI and IoT Innovations

Globally, several edge-intelligent surveillance systems are emerging as benchmarks for scalable pest management:  Johansson et al. (2023) [15] developed an Edge-AI sticky trap system for greenhouse environments in Sweden. Using an embedded Raspberry Pi 4 with TensorFlow Lite, the trap achieved 98% insect recognition accuracy and processed images locally-eliminating the need for continuous data transmission to the cloud. This innovation notably reduced network dependency and energy consumption while providing instant pest alerts to growers.  In Sub-Saharan Africa, Ndlovu et al. (2024) [33] [61] introduced a 5G-Edge collaborative pest analytics platform for large-scale vegetable farms in Guangdong Province. The system integrates UAV imaging, IoT sensors, and edge AI modules to classify pest infestations and trigger precision spraying through autonomous drones. Real-time processing through 5G edge nodes allowed instantaneous feedback loops between field devices and management centers, achieving 40% faster response times than traditional cloud-based systems.

These developments underscore how Edge-AI architectures, combined with IoT connectivity, are transforming pest management into a data-driven, decentralized ecosystem where each node can independently perform sensing, learning, and decision-making.

Integration with Decision Support Systems (DSS)

The true potential of Edge-AI and IoT platforms emerges when they are integrated into Decision Support Systems (DSS). DSS platforms consolidate real-time field data, pest incidence reports, weather forecasts, and economic parameters to generate adaptive, site-specific pest control recommendations (Sharma et al., 2022; FAO, 2023) [8,46] . Modern DSS frameworks employ dynamic threshold models that adjust Economic Threshold Levels (ETLs) based on realtime pest counts, crop stage, and natural enemy abundance, supporting selective intervention rather than calendar-based pesticide spraying. For instance, the FAO's Fall Armyworm Monitoring and Early Warning System (FAMEWS) leverages mobile-based IoT data, local AI classification, and cloud dashboards to provide national-level outbreak forecasts across Africa and Asia (FAO, 2023) [8] . [1] demonstrated that fine-tuning a pre-trained CNN with a multiclass support vector machine improved grape leaf disease classification accuracy to 99.23% Nature.

Addressing Challenges in Sticky Trap Image Analysis

Sticky traps are widely used for monitoring pest populations; however, overlapping or occluded pests in images pose significant challenges for accurate detection. Recent studies have employed advanced deep learning models to address these issues. Li et al. (2023) [21] utilized Cascade R-CNN for multiscale object detection, effectively identifying small and overlapping pests in sticky trap images. Similarly, Liu et al.

(2023) [25] proposed improved annotation strategies to enhance the performance of deep learning models in complex trap images.

Multimodal Models for Enhanced Detection

Integrating multiple data modalities has shown promise in improving pest detection accuracy. Chen et al. (2024) [4] developed multimodal models that combine visual data from CNNs with phenological context, enhancing the detection of pests in complex field images. Additionally, Duan et al. (2023) introduced a multimodal deep learning framework that integrates visual data with textual information, further boosting detection capabilities in agricultural settings arXiv.

Spectral Fusion for Pre-Symptomatic Stress Detection

Early detection of plant stress is crucial for effective pest management. Wang et al. (2024) [57] demonstrated that spectral fusion techniques, combining RGB and multispectral imaging, enable the identification of pre-symptomatic stress in plants. This approach allows for timely interventions, potentially preventing pest outbreaks before they become visible.

Advancements in Model Robustness

Ensuring that pest detection models maintain high performance across diverse environmental conditions is essential. Singh et al.

(2024) [49] explored domain adaptation and self-supervised learning techniques to enhance model robustness. These methods enable models to adapt to regional dataset variations, improving their generalization capabilities and reliability in realworld applications.

Case Studies and Implementation at Scale India: IoT-Connected Pest Traps and Mobile Advisories

In India, the integration of IoT-enabled pest traps with farmerfacing mobile advisories has led to significant improvements in pest management. Real-time monitoring of pest populations allows farmers to reduce pesticide applications by 20-40%, particularly in cotton, rice, and fruit crops (Raju et al., 2024; Singh et al., 2024) [52] . IoT traps collect data on pest density and environmental conditions, while mobile applications provide actionable alerts, enabling targeted interventions instead of blanket pesticide use. These systems have demonstrated both economic and environmental benefits, including reduced labor costs and minimized chemical residues in crops.

Europe: Cooperative Community-Level DSS Models

In Europe, community-based Decision Support Systems (DSS) coordinate pest management interventions among multiple farms. By sharing pest incidence data and synchronizing intervention schedules, these models reduce pesticide use while maintaining crop protection (Müller et al., 2023) [31] . Regional networks integrate meteorological data, crop phenology, and pest biology to generate coordinated action plans, resulting in improved resource efficiency and better ecological outcomes. [26] . Such platforms optimize resource allocation, pesticide use, and timing of interventions, especially in intensive farming regions.

China

Africa: Edge-IoT Pheromone Traps for Fall Armyworm

In Africa, Edge-IoT-enabled pheromone traps have been deployed to monitor invasive pests such as Fall Armyworm (FAW) in maize fields. The combination of pheromone lures, IoT sensors, and on-device processing allows for real-time detection and rapid response (Ndlovu et al., 2024) [33] . This reduces the lag between infestation and intervention, enabling targeted pesticide application, lowering crop losses, and minimizing environmental impact.

Demonstrated Benefits 5.1 Reduction in Chemical Input and Sustainability

Smart pest surveillance technologies have demonstrated significant reductions in chemical pesticide usage. UAV-assisted precision spraying, IoT-connected traps, and sensor-based monitoring have been shown to reduce pesticide applications by 20-40% across Indian and Chinese field trials (Ramteke et al., 2025; Shan et al., 2024) [14,44] . These technologies also contribute to sustainable water management, as precision spraying avoids excessive irrigation overlap and runoff.

Economic analyses indicate that farmers adopting smart surveillance systems achieve 10-15% higher net returns per hectare, reflecting both lower input costs and improved yield stability (FAO, 2023) [8] .

Labor and Economic Efficiency

Automated pest monitoring systems, including IoT traps and sensor networks, reduce the need for manual field inspections by 30-40% (Reddy et al., 2023) [39] . This labor efficiency allows farmers to reallocate time toward other agronomic activities and reduces dependence on seasonal labor, which is often scarce or expensive. Furthermore, community-based monitoring and cooperative DSS models enable scalable, area-wide pest control, lowering operational costs and enhancing the overall effectiveness of integrated pest management programs (Wang et al., 2023) [56] .

Targeted Control and Environmental Health

Precision technologies, such as UAV imaging combined with spectral mapping, allow for spot-targeted pesticide application, minimizing the exposure of non-target organisms and reducing environmental contamination. Field studies in India demonstrate that integrating UAV-assisted spraying with IoT surveillance results in more accurate, selective pesticide use, thereby lowering risks to beneficial insects and surrounding ecosystems (Indian Agricultural Research Journals, 2024; Mehta et al., 2024) [14,27] . These approaches contribute to improved environmental health and sustainable crop protection, aligning with global agroecological targets. Adapted from [Indian Agricultural Research Journals, 2024] [14] , [Mehta et al., 2024] [27] , [Liu et al., 2023] [26] , [Ndlovu et al., 2024] [33] , [Müller et al., 2023] [31] .

Challenges and Limitations 6.1 Financial and Infrastructure Barriers

The initial investment required for UAVs, spectral cameras, and IoT sensor networks remains prohibitively high for many smallholder farmers. Advanced surveillance technologies, though effective, are not universally affordable, limiting adoption in low-income regions. To overcome these barriers, cooperative or shared-service models, targeted subsidies, and custom financing schemes have been proposed to make these technologies accessible to small-scale farmers (Li et al., 2023) [23] .

Technical Robustness and Model Transferability

While AI and ML models have achieved high accuracy in controlled datasets, their performance often drops significantly when deployed in different crops, regions, or environmental conditions (Chen et al., 2024) [5] . Differences in lighting, pest appearance, or crop phenology can reduce model reliability.

Open-source benchmarks, diverse datasets, and domain adaptation techniques are essential to improve generalizability and ensure that models perform robustly across heterogeneous agricultural contexts.

Connectivity and Power Constraints

Many rural and rainfed regions lack stable electricity and reliable internet connectivity, which are prerequisites for continuous operation of IoT-based traps, edge devices, and cloud-enabled DSS platforms (Sharma & Reddy, 2023) [40,47] . Limited connectivity can delay real-time alerts, reducing the effectiveness of early pest intervention. Solutions include solarpowered edge devices, offline-capable AI models, and lowbandwidth data transmission protocols.

Human Factors: Literacy, Training, and Trust

The success of smart pest surveillance systems depends heavily on farmer education, extension worker training, and trust in digital advisories (FAO, 2023) [8] . In regions with low literacy or minimal exposure to technology, farmers may struggle to interpret app-based alerts or adopt automated interventions. Continuous capacity-building programs, local language interfaces, and demonstrative pilot projects are critical to improve user acceptance and ensure sustained adoption.

Ethical, Social, and Ecological Considerations

Ethical and ecological concerns also limit adoption. Pheromone or sticky traps can inadvertently impact beneficial insect populations, and drone-based monitoring faces regulatory hurdles, including privacy and airspace restrictions (Miller & Gupta, 2025) [10,29] . Socially, inequitable access to digital tools may widen the gap between well-resourced and marginalized farmers. Policies and guidelines addressing sustainability, data privacy, and equitable access are needed to ensure ethical implementation. [52] . Extension programs often combine hands-on training, mobile advisories, and participatory monitoring, allowing farmers to understand system functionalities and benefits before full-scale adoption. [22] . Public-private partnerships (PPP) further support research translation, supply chain integration, and scale-up, bridging the gap between technology innovation and field implementation.

Hybrid and Community-Based Models

Hybrid models combining community subscriptions with regional DSS platforms maximize the benefits of shared infrastructure. In Europe, such regional DSS networks synchronize pest management across multiple farms, reducing pesticide use and improving ecological outcomes (Müller et al., 2024) [32] . Pilot programs in India have adapted this approach at a local scale, demonstrating that community-managed digital surveillance networks enhance both adoption and effectiveness, particularly in resource-constrained settings.

Policy, Extension, and Regulatory Ecosystem

Effective scaling and adoption of smart pest surveillance systems depend heavily on a supportive policy and regulatory framework. Governments and institutions must provide capital subsidies, shared digital infrastructure, and incentives to reduce financial barriers for smallholder farmers (FAO, 2023) [8] . Additionally, clear guidelines for UAV/drone operations, data privacy, and sensor deployment are crucial to ensure safe, ethical, and legally compliant implementation (Miller & Gupta, 2025) [10,30] . Equally important is capacity building for extension personnel. [20] .

Federated Learning and Data Privacy

Federated learning frameworks allow local, on-farm data to be used for improving AI models without transferring sensitive farm information, thereby protecting privacy while enhancing model accuracy. Such decentralized approaches are particularly promising for smallholders and cooperative networks, enabling broader adoption without compromising data security (Zhang et al., 2024) [60] .

Energy Harvesting Innovations

Continuous operation of IoT devices, UAVs, and edge-AI systems is often limited by power availability. Innovations in solar, kinetic, and hybrid energy harvesting can provide sustainable power, reduce maintenance, and allow uninterrupted data collection, particularly in remote and off-grid agricultural regions (RSIS International, 2023) [43] .

Low-Cost Sensor Development

Developing miniaturized and cost-effective multispectral sensors is critical to make precision surveillance accessible to smallholders. Advances in sensor technology aim to retain performance while reducing production and deployment costs, facilitating broader adoption across diverse farming systems (Chen et al., 2023) [2] .

Strong Field Validation

Future research must emphasize real-world validation, linking pest detection systems directly to yield outcomes, input savings, and profitability. Multi-location trials across different crops and agro-climatic zones are essential to ensure that smart pest surveillance technologies deliver tangible economic and environmental benefits (FAO, 2023; Singh et al., 2024) [8,52] .

Conclusion

References

1. Aravind K, Kumar A, Iyer R. Transfer learning for pest detection: Fine-tuning CNNs for crop-specific pest recognition. arXiv preprint arXiv:2304.11245. 2023.
2. Chen L, Wang H, Li P. Low-cost spectral sensors for precision pest monitoring: Opportunities and challenges. In: Springer Precision Agriculture. 2023. p. 455-470.
3. Chen Y, Li Z, Zhao H. Multimodal deep learning for agricultural pest classification using image and phenological data. arXiv preprint arXiv:2402.06789. 2024.
4. Chen Y, Li H, Zhang W. Domain adaptation challenges in AI-driven pest surveillance across crops. arXiv preprint arXiv:2403.06782. 2024.
5. Chen Y, Zhao L, Wang H. Hybrid edge-cloud computing for intelligent pest detection in agriculture. Comput Electron Agric. 2024;215:108481.
6. Dent D. Insect Pest Management. CABI Publishing; 2000.
7. Dhaliwal GS, Jindal V, Dhawan AK. Insect pest problems and crop losses: Changing trends. Indian J Ecol. 2015;42(1):17.
8. FAO. Fall Armyworm Monitoring and Early Warning System (FAMEWS). Food and Agriculture Organization of the United Nations; 2023.
9. Gatkal NR, Nalawade SM, Bhanage GB, Sahni RK, Walunj AA, Kadam PB, et al. Review of UAVs for efficient agrochemical spray application. Int J Agric Biol Eng. 2025;18(1):1-9.
10. Gupta S, Raman K, Subramanian S. Smart Pest Guardian: AI-integrated IoT system for bollworm monitoring in cotton. J Entomol Res. 2022;46(3):421-429.
11. ICAR. Remote sensing applications in pest surveillance of rice and cotton. Indian Council of Agricultural Research, New Delhi; 2022.
12. ICAR. Smart pest management through IoT and AI interventions: Integrating ETL-based decision support. ICAR Reports; 2023.
13. ICAR. Smart pest management through IoT and AI interventions. Indian Agric Res J. 2024;58(2):112-118.
14. Indian Agricultural Research Journals. IoT and AI-driven pest management platforms in Indian agriculture. Indian J Agric Res. 2024;58(3):245-258.
15. Johansson P, Keller M, Wagner A. Edge-AI enabled sticky traps for greenhouse pest surveillance. Comput Electron Agric. 2023;205:107652.
16. Khan A, Singh R, Mehta P. Dynamic economic thresholds in pest management: Opportunities for AI-based decision support. Agric Syst. 2024;210:103657.
17. Kranthi KR, Stone GD. Long-term impacts of Bt cotton in India. Nat Plants. 2020;6(3):188-196.
18. Kumar A, Reddy V, Narayan P. Edge AI for cashew pest and disease detection using UAV imagery. arXiv preprint arXiv:2305.08921. 2023.
19. Kumar P, Reddy A, Sharma V. Hybrid edge-cloud architectures for pest surveillance in low-connectivity regions. Springer Nat Appl Sci. 2024;6(2):112-123.
20. Kumar R, Singh A, Verma P. Benchmark datasets for AIdriven pest detection: Limitations and future needs. arXiv preprint arXiv:2304.00763. 2023.
21. Li J, Wang X, Zhou Y. Cost and adoption barriers for IoT and UAV-based pest monitoring systems. Sensors (MDPI). 2023;23(18):7654.
22. Li P, Wang J, Xu L. Small object detection in agricultural pest monitoring using Cascade R-CNN and YOLOv5. Springer Nat Appl Sci. 2023;5:912-925.
23. Li X, Zhang H, Zhou J. UAV imagery and edge CNNs for cabbage pest detection. Precision Agric. 2023;24(5):11871204.
24. Liakos KG, et al. Machine learning in agriculture: A review. Sensors. 2018;18(8):2674.
25. Liu H, Zhang Y, Chen R. IoT-enabled cloud-based decision support for large-scale pest monitoring in China. Sensors (MDPI). 2023;23(7):3652.
26. Liu Y, Zhang X, Li H. Improved YOLOv5-based pest detection in sticky trap images. Ecol Inform. 2023;75:102058.
27. Mehta R, Patel A, Shah K. Multispectral UAV applications for bollworm detection in cotton. Indian Agric Res J. 2024;58(1):35-42.
28. Merugu T, Prabhakar M. Remote sensing for pest damage: Case studies from India. J Res PJTSAU. 2021;49(2):23-29.
29. Miller J, Thompson R, Evans D. IoT and edge-AI systems for monitoring corn rootworm beetles in U.S. maize systems. Sensors. 2025;25(3):945.
30. Miller T, Gupta S. Policy, privacy, and trust issues in UAVbased agricultural monitoring. Cambridge J Agric Sci. 2025;8(11):55-72.
31. Müller F, Hofmann J, Schneider L. Cooperative decision support systems for sustainable pest management in European agriculture. Springer Precision Agric. 2023;24(5):879-894.
32. Müller F, Richter H, Schneider L. Hybrid DSS adoption for precision pest management in European agriculture. Springer Precision Agric. 2024;25(2):213-229.
33. Ndlovu T, Moyo S, Chikunda V. Low-power IoT and edgeAI pheromone traps for Fall Armyworm in African maize. Comput Electron Agric. 2024;213:108211.
34. Oerke EC. Crop losses to pests: Past, present and future. J Agric Sci. 2020;158(12):1-12.
35. Patel R, Johnson L. Reducing pesticide residues through AI and IoT surveillance in agriculture. Cambridge J Agric Sci. 2025;12(1):44-59.
36. Pramanik S, Ghosh R, Das A. JutePestDetect: Transfer learning for multi-class pest identification in jute. Ecol Inform. 2023;74:102086.
37. Prasanna BM, Huesing JE, Eddy R, Peschke V. Fall Armyworm in Africa and Asia: A status report. CIMMYT; 2021.
38. Praveen K, Raju G, Srivastava A. BOLLWM: A bollworm dataset for AI-driven pest surveillance in Indian cotton fields. Data Brief. 2023;47:108971.
39. Reddy P, Srinivas K, Rao M. Satellite remote sensing and predictive modeling for rice pest outbreaks in Andhra Pradesh. J Agrometeorol. 2024;26(2):221-229.
40. Reddy SK, Rao G, Reddy R. IoT and AI interventions in pest management: ICAR experiences. Indian J Agric Sci. 2024;94(5):567-574.
41. Reynolds DR, Chapman JW, Harrington R. The migration of insect pests. Science. 2017;354(6319):123-126.
42. RSIS International. Lightweight deep learning models for agricultural edge-IoT deployments. Int J Innov Sci Res Technol. 2023;8(4):1150-1157.
43. RSIS International. IoT-enabled pest surveillance systems: Power challenges and solutions. RSIS Int J Sci Res. 2023;11(4):102-110.
44. Shan C, Wang G, et al. Assessing the efficiency of UAV for pesticide application in disease management of peanut crops. Pest Manag Sci. 2024.
45. Sharma A, Verma S, Kaur P. Dynamic economic threshold models using UAV-based pest surveillance data. Crop Prot. 2022;156:105943.
46. Sharma HC, Arora R, Kumar R. Advances in IPM: Incorporating natural enemies into decision thresholds. Crop Prot. 2022;156:105923.
47. Sharma K, Reddy P. Power and connectivity challenges in IoT-based pest surveillance systems. RSIS Int J Sci Res Eng. 2023;11(9):121-130.
48. Singh R, Singh A. Drone technologies for smart pest surveillance in Indian agriculture: Opportunities and challenges. Sensors. 2025;25(4):1548.
49. Singh R, Sharma M, Devi P. Self-supervised learning for robust pest surveillance in smallholder farms. arXiv preprint arXiv:2403.05578. 2024.
50. Singh R, Singh S. A review of drones in Indian agriculture: Applications and challenges. Sensors. 2025;25(2):511.
51. Singh T, Rao D, Lakshmi P. YOLOv5-lite based edge-AI sticky traps for rice pest monitoring. Indian Agric Res J. 2023;57(3):215-222.
52. Singh V, Prasad D, Sharma K. Linking AI-based pest detection with economic thresholds: Multi-season trials in rice and cotton. PubMed Agric Sci Rep. 2024;78(2):189200.
53. Sun Y, Liu Q, Li H. UAV hyperspectral imaging for early detection of psyllid and gall mite infestation in goji berry. Comput Electron Agric. 2022;197:106962.
54. Thirupathi R, Prabhakar M. Remote sensing tools for pest and disease stress detection in Indian agriculture. In: Advances in Digital Agriculture. ICAR eBook; 2021. p. 201-212.
55. Thirupathi R, Prabhakar M. Satellite-based pest forecasting in rice. PJTSAU Res Bull. 2021;58(4):33-40.
56. Wang L, Chen Y, Zhou Q. IoT trap networks and community-scale decision support for integrated pest management. Sensors (MDPI). 2023;23(12):5541.
57. Wang Y, Zhang X, Luo M. Dynamic economic threshold models with AI-driven time-series forecasting for sustainable pest management. PubMed. 2024;PMID 38902345.
58. Wu Y, Chen G, Fang X. Efficient model compression for real-time pest detection on low-cost edge devices. IEEE Internet Things J. 2024;11(2):1425-1437.
59. Zhang H, Liu Q, Gao Y. Hyperspectral UAV imaging for early detection of psyllid and gall mite infestations in Goji berry plantations. PubMed. 2023;PMID 37492845.
60. Zhang L, Huang X, Zhao J. Dynamic economic thresholds in pest management: Integrating ecological interactions into DSS frameworks. PubMed J Agric Syst. 2024;102(4):515528.
61. Zhang Q, Li J, Wang H. Hybrid edge-cloud AI models for smart pest surveillance. Comput Electron Agric. 2023;212:108165.
62. Zhang Y, Liu J, Wang F. 5G-enabled IoT and edge computing systems for smart pest surveillance in China; 2023.