Illegal Logging Detection from Acoustic Sensors

Abstract

Illegal logging is a major concern for forests such as deforestation and biodiversity loss. In this presentation, we will be introducing an acoustic-based monitoring system which detects chainsaw and tree-cutting sounds and filters out everyday forest sounds like birds, wind, rain and insects. The system uses INMP441 MEMS micros and a pre-trained model for sound classification, YAMNet. It integrates with ESP32, GPS module and the IoT communications to offer instant warnings of suspicious activity. The proposed approach intends to facilitate the forest monitoring, manual patrol and assist in the conservation of forest resources. Illegal logging is responsible for the destruction of countless square kilometers of forest land, yet the robust acoustic monitoring system is often overlooked. Illegal logging for example is destroying enormous amount of square meters/kilometers of forests, in which case, the audio monitoring system, which is robust and capable is rarely being utilized.

Introduction

Forests are vital natural resources that maintain ecological balance, regulate climate, preserve biodiversity, and provide habitats for wildlife. However, illegal logging poses a serious threat by causing deforestation, habitat destruction, soil degradation, and increased carbon emissions. Traditional forest monitoring primarily relies on manual patrols by forest officials, which require significant manpower, time, and financial resources. Since illegal logging is often detected only after considerable environmental damage has occurred, there is a need for automated systems capable of continuously monitoring large and remote forest areas.

Recent advances in Artificial Intelligence (AI), acoustic sensing, and Internet of Things (IoT) technologies have enabled the development of intelligent forest monitoring systems. Acoustic monitoring is particularly effective because illegal logging activities generate distinctive sounds such as chainsaw operation, tree cutting, vehicle movement, and axe strikes. Deep learning models can accurately distinguish these sounds from normal forest sounds, allowing suspicious activities to be detected in real time. Early detection enables forest authorities to respond promptly, minimizing environmental damage and reducing dependence on manual surveillance.

This study proposes an Illegal Logging Warning and Detection System (IWDS) based on acoustic monitoring. The system integrates an INMP441 MEMS microphone, an ESP32 microcontroller, a GPS module, and a pretrained YAMNet audio classification model. The microphone continuously captures environmental sounds, while YAMNet analyzes the audio waveform and classifies it into normal forest sounds or suspicious logging activities. When the probability of illegal logging exceeds a predefined confidence threshold, the ESP32 retrieves the GPS coordinates and sends an alert containing the detected sound category, time, and location to forest authorities through GSM or Wi-Fi communication. Since data are transmitted only when suspicious activity is detected, the system minimizes power consumption and unnecessary communication while enabling continuous monitoring.

The study is intended for deployment in protected forests, wildlife sanctuaries, reserve forests, and other ecologically sensitive regions where unauthorized tree cutting is a major concern. Acoustic sensor nodes are strategically installed throughout these areas to continuously monitor surrounding sounds. This distributed monitoring approach increases coverage in large and inaccessible forests while reducing the need for constant human intervention.

The audio dataset used for training consists of both publicly available environmental sound datasets and manually recorded forest sounds. The recordings are categorized into two classes: normal forest sounds, including birds, insects, flowing water, wind, and animal calls, and illegal logging sounds, including chainsaws, axe strikes, vehicle movements, and other human activities associated with unauthorized logging. All recordings are carefully labeled, converted into a standardized format, and divided into training, validation, and testing datasets for model development and evaluation.

The proposed methodology follows a standard YAMNet workflow. Captured audio undergoes preprocessing to reduce background noise and improve sound quality before being supplied directly to the pretrained YAMNet model. Rather than manually extracting acoustic features, YAMNet internally generates the required audio embeddings. Transfer learning is then applied by retraining the final classification layer using the labeled forest audio dataset, enabling the model to accurately distinguish between natural forest sounds and illegal logging activities.

System performance is evaluated using standard machine learning classification metrics, including accuracy, precision, recall, and F1-score. Accuracy measures overall classification performance, precision evaluates the proportion of detected illegal logging events that are correct, recall assesses the system's ability to detect all actual illegal logging incidents, and the F1-score provides a balanced measure of precision and recall. The trained model is validated using an independent testing dataset to assess its capability to classify normal forest sounds, recognize chainsaws and logging activities, perform under varying environmental conditions, and minimize false positive and false negative detections.

The proposed system is also compared with conventional manual forest monitoring based on several criteria, including detection response time, monitoring coverage, human effort, operating cost, continuous monitoring capability, and effectiveness in detecting illegal logging. Continuous reliability testing is performed under varying weather conditions, ambient noise levels, and different forest environments to evaluate system stability and robustness. Additional performance indicators include false alarm rate, communication reliability, power consumption, and response time.

The expected results indicate that the proposed acoustic monitoring system will accurately distinguish illegal logging sounds from normal environmental sounds with high classification performance using the YAMNet transfer learning model. Integration of the INMP441 microphone, ESP32 controller, GPS module, and IoT communication enables continuous, autonomous forest monitoring and rapid location-based alerts to forest authorities whenever suspicious activities are detected. Compared with traditional manual patrols, the system is expected to reduce monitoring costs, decrease manpower requirements, shorten response time, and improve surveillance coverage. Overall, the proposed AI-enabled acoustic monitoring system offers an efficient, reliable, and sustainable solution for protecting forest resources, conserving biodiversity, and supporting long-term forest management through early detection and timely intervention against illegal logging.

Conclusion

In this study, an illegal logging detection system based on Acoustic technology and IoT is proposed, which will be integrated with Environmental sound classification technology, to enhance the forest monitoring system. The system can detect the sound of chainsaw and tree cutting and differentiate the sound of the chain saw and tree cutting from the forest noises, such as birds, wind, rain and insects. The proposed system is designed to send alerts in real-time whenever the presence of suspicious logging is detected by incorporating the pretrained YAMNet model with an INMP441 MEMS microphone, ESP32, GPS module, and wireless communication. This can help lower the need for constant manual patrolling of the forest, facilitate quick identification of illegal logging activities, and enhance the effectiveness of forest surveillance. Proposed system can potentially enable sustainable forest management and help conserve forest resources and biodiversity with future field testing and further improvements.

References

[1] U. R. Prajna, P. Adhishri, M. V. Gaonkar, K. R. Prajwal, and K. Sathwik, “Acoustic Based Real Time Monitoring and Alert System for Illegal Logging Detection,” 2026. [2] K. Sivanandam, A. Suryanarayanan, V. Tharun, K. Srisaran, and S. Vignesh, “Monitoring System for Protection of Endangered Trees Against Illegal Cutting and Smuggling Using IoT & GSM,” 2025. [3] P. R. Rajeev, M. Siddarth, V. Suresh, and A. R. Jayan, “Sensor Fusion Based Approach for the Detection of Illegal Tree Felling,” 2025. [4] A. L. Kotian, Ashritha, A. H. Shetty, Anush, A. Hafeeza, and A. V. Suvarna, “Guardians of the Green: A Review of AI, ML and IoT Innovations Against Illegal Deforestation and Wildlife Poaching,” 2025. [5] V. Singh, K. C. Ray, and S. Tripathy, “Real-time Monitoring of the Illegal Logging Events with Intelligent Acoustic Sensors Nodes,” 2024. [6] D. Simiyu, A. Vikiru, H. Muchiri, F. Gu, and J. Butime, “Chainsaw-Sound Recognition Model for Illegal Logging in Forest,” 2024. [7] J. Svatos and J. Holub, “Impulse Acoustic Event Detection, Classification, and Localization System,” 2023. [8] S. F. Ahmad and D. K. Singh, “Detection of Tree Cutting in Forest by Using Acoustic Properties Automatically,” 2022. [9] A. Andreadis, G. Giambene, and R. Zambon, “Monitoring of Illegal Tree Cutting Using Ultra-Low-Power Smart IoT Devices,” 2021. [10] G. A. Mutiara, N. Suryana, and O. Mohd, “Multiple Sensor on Clustering Wireless Sensor Network to Tackle Illegal Cutting,” 2020. [11] G. A. Mutiara, N. S. Herman, and O. Mohd, “The Use of Long-Range Wireless Sensor Network to Monitor the Illegal Cutting Log,” 2020. [12] I. Mporas, I. Perikos, V. Kelefouras, and M. Paraskevas, “Acoustic Surveillance of Forest for Illegal Logging Detection,” 2020. [13] G. Sharma, K. Umapathy, and S. Krishnan, “Trends in Audio Signal Feature Extraction Methods,” 2020. [14] M. Magno, F. Vultier, B. Szebedy, et al., “A Bluetooth-Low-Energy Sensor Node for Acoustic Monitoring of Small Birds,” 2020. [15] M. M. Faraji, S. B. Shouraki, E. Iranmehr, and B. Linares-Barranco, “Wide-Range Outdoor Sound Source Localization Based on Distributed Sensor Network,” 2020. [16] M. J. Baucas and P. Spachos, “Large Scale IoT-Based Urban Sound Classification with Cloud and Fog Computing,” 2020. [17] V. Yaremenko, M. R. Azimi-Sadjadi, and J. Zacher, “Unattended Acoustic Sensor Systems for Source Detection, Classification, and Tracking,” 2019. [17] J. T. Chen, C. B. Lin, J. J. Liaw, and Y. Y. Chen, “Improvements on the Implementation of Sensor Nodes for Illegal Logging Detection,” 2018. [18] D. C. Prasetyo, G. A. Mutiara, and R. Handayani, “Illegal Logging Detector System Using Chainsaw Vibration and Sound,” 2018. [19] A. Gaita, G. Nicolae, A. Radoi, and C. Burileanu, “A Spectral Haar Coefficients Based Chainsaw Sound Detection Method,” 2018. VII

Copyright

Copyright © 2026 Smitha B N, Divyashree K, Preksha M, Priyanka H M, Uma T R, Zaina Shariff. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.