What question did this study set out to answer?

This research aims to improve soil oxygen monitoring using an RFID-based sensor and deep learning algorithms.

March 15, 2026Langmuir0 citations

RFID-Based ZnO/TiO 2 /SnS 2 Soil Oxygen Content Sensor Coupled with 1D-CNN-GRU Model: Classification for Predicting Soil Oxygen Content System

FMFengjuan MiaoQiqihar University FWFan WuQiqihar University BTBairui TaoQiqihar University

Key Points

This research aims to improve soil oxygen monitoring using an RFID-based sensor and deep learning algorithms.
Developed an RFID sensor utilizing ZnO/TiO2/SnS2 nanocomposite materials.
Integrated a 1D-CNN-GRU model for classification and prediction of soil oxygen levels.
Conducted experiments to measure response linearity and monitoring accuracy under varying oxygen concentrations.
Achieved a response linearity (R2) increase from 0.97 to 0.99 for oxygen levels of 5-25%.
Overall accuracy of oxygen level classification reached 98.4%, with high precision, recall rate, and F1 score values.
Demonstrated a dynamic monitoring accuracy improvement of 3.16% over traditional methods.

Abstract

In greenhouse pot cultivation, the oxygen content in the soil is a key factor influencing root respiration and nutrient absorption of crops, directly affecting the yield and quality of the crops. To achieve stable and precise soil oxygen monitoring, this study proposes an RFID sensor based on ZnO/TiO2/SnS2 nanocomposite materials, combined with intelligent algorithms to overcome the shortcomings of traditional monitoring methods in terms of cost, stability, accuracy, and anti-interference ability. The sensor is based on FR-4 copper-clad laminate as the substrate, integrates multilayer nanocomposite materials, and adopts environmental multifrequency encoding technology to achieve synchronous perception of multiple parameters and signal separation. To further improve the sensor performance, this study introduces a 1D-CNN-GRU classification prediction model combining one-dimensional convolutional neural network and gated recurrent unit. This algorithm significantly optimizes the data-processing capability and output quality of the sensor by extracting the local features of the sensor sequence and modeling its long-term temporal dependency. Experiments show that with the algorithm's support, the response linearity (R2) of the sensor within the oxygen concentration range of 5-25% has increased from 0.97 to 0.99, and the real-time dynamic monitoring accuracy has increased by 3.16% compared to traditional methods. In the classification task of four levels of oxygen intervals (5-10%, 10-15%, 15-20%, 20-25%), the overall accuracy of the system reaches 98.4%, with precision, recall rate, and F1 score reaching 98.8%, 97.4%, and 97.7%, respectively, and the sample error is less than one level, with a response time of less than 0.2 h. This study verifies the significant advantages of the sensor and algorithm collaboration in soil oxygen monitoring, providing reliable data support and methodological basis for soil health management, growth optimization, and intelligent regulation of greenhouse pot crops.

Demander à l'IA

Bookmark

View Full Paper