Intelligent Agricultural Product Quality Monitoring System Leveraging Sensor Network Data
Summary
No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.
Summarise with VIRAT Free to try with your Science Catalogs account.Abstract
To address the significant differences in monitoring requirements across agricultural production scenarios, such as greenhouses, fields, and fruit trees, and the difficulty of traditional time series algorithms in effectively capturing spatiotemporal characteristics, this paper designs a multi-scenario adaptive intelligent agricultural product monitoring system and proposes a spatiotemporal attention fusion LSTM (STA-LSTM) algorithm. The system comprises three core modules: the sensor network layer is deployed according to specific scenarios (50 m2/node in greenhouses, 20 m2/node in fields, and a combination of "trunk and ground" deployment for fruit trees). It utilizes devices such as the SHT30 temperature and humidity sensor (accuracy: ±0.5°C) and the JWSK-600 soil sensor (accuracy: ±0.1 pH). The data transmission layer adapts ZigBee (for greenhouses), LoRa (for fields), and 4G (for long-range networks), employing AES-128 encryption and CRC-16 checksum to ensure data security. The intelligent analysis layer, based on the STA-LSTM algorithm, integrates six core monitoring data types (temperature, humidity, soil parameters, etc.) using a time period adjustment factor (α = 0.2) and a spatial weighting model to improve prediction accuracy. The experiment used 12,600 samples (7,200 for tomatoes and 5,400 for cucumbers) collected over 60 days at a planting base in Jiangsu Province, comparing them with traditional LSTM, GRU, and CNN-LSTM algorithms. Results show that STA-LSTM achieved an RMSE of 0.032 (a 28.9% reduction compared to traditional LSTM) and an Acc of 96.8% (an 8.2% improvement) in tomato monitoring, and an RMSE of 0.035 (a 26.3% reduction) and an Acc of 95.6% (a 7.5% improvement) in cucumber monitoring. The system also achieved an abnormal response time of 0.8s, with accuracy dropping only 2.1% when resistant to 10% noise. Regarding system performance, the system achieved a response time of 5.01s under a high load of 10,000 data items, a node optimization lifespan of 98-111 days (energy consumption reduced by 44%-45%), and an average daily failure rate of ≤0.3 in a 30-day stability test, meeting the real-time and reliability requirements of precision agricultural monitoring.
Fields of Study & Subjects
Smart Agriculture and AI
Advanced Sensor and Control Systems
Water Quality Monitoring Technologies
Publication details
- Journal / source
- IEEE International Conference on Power, Electronics and Green Energy (ICPEGE)
- Pages
- 1797–1802
- Publication date
- 28 January 2026
- Language
- EN