DefinePK

Research Objective

To develop and demonstrate an AI-powered IoT system for real-time monitoring and detection of illegal logging activities in forests.

Methodology

The methodology involves collecting audio data (Silence, Axe, Chainsaw, Fire), preprocessing it using Mel-filter bank energy features (MEF), and training a Convolutional Neural Network (CNN) model. The trained model is then deployed on an ESP32 microcontroller as a sensor node. These nodes communicate via LoRa to a central ESP32 server node, which integrates with Firebase for real-time data storage and a mobile application for notifications.

Methodology Flowchart

                        graph TD
    A[Data Collection & Preprocessing] --> B[Feature Extraction: MEF];
    B --> C[Train CNN Model];
    C --> D[Export Model to ESP32];
    D --> E[ESP32 Sensor Node: Audio Recording & Classification];
    E --> F[LoRa Communication];
    F --> G[ESP32 Server Node];
    G --> H[Firebase Database Integration];
    H --> I[Mobile Application: Real-time Notifications];
    I --> J[Enforcement Agency Intervention];                    

Discussion

The research highlights the effectiveness of combining AI and IoT for combating illegal logging. The use of audio classification via CNNs trained on MEF features allows for the detection of specific sounds associated with illegal activities. The low-power ESP32 microcontroller and LoRa communication enable widespread, real-time monitoring in remote forest areas. The integration with cloud platforms like Firebase and a user-friendly mobile application facilitates prompt intervention by enforcement agencies.

Key Findings

The developed AI-powered IoT system achieved an accuracy of 96.2% on the validation set and 82.57% in real-world testing scenarios. The system demonstrated efficient inferencing (21ms) with minimal resource usage (10.6Kb RAM, 32.6Kb Flash) on the ESP32 microcontroller, making it suitable for deployment in resource-constrained forest environments.

Conclusion

The AI-powered IoT system presents a viable and effective solution for combating illegal logging, contributing to biodiversity preservation and sustainable forest management. The system's ability to provide timely detection and intervention capabilities makes it a valuable tool for safeguarding forests and their ecosystems.

Fact Check

1. Accuracy: The model achieved 96.2% accuracy on the validation set and 82.57% in testing scenarios.
2. On-device performance: Inferencing time was 21ms, with peak RAM usage of 10.6Kb and Flash usage of 32.6Kb on the ESP32.
3. Data collection: The dataset comprised 75 samples for each of the four audio classes (Silence, Axe, Chainsaw, Fire).

Mind Map