DOI QR코드

DOI QR Code

Implementation of a Wi-Fi Mesh-based Fire Detection System using Multiple Sensor Nodes

  • Received : 2024.09.19
  • Accepted : 2024.10.24
  • Published : 2024.11.29

Abstract

In this paper, we propose a Wi-Fi Mesh-based fire detection system for fire detection and rapid response. Previous fire detectors had the problem that it is difficult to distinguish between fire and non-fire based on a single sensor, and since multiple detectors operate independently, there is a lack of interconnectivity. In this paper, we present a fire detection system based on a K-NN classification model using a multi-sensor based fire detector. Also, by constructing a mesh network for fire detection, detectors within a spatial range can be interlinked to detect fire. Looking at the performance evaluation results of the implemented system, it was confirmed that the TPR(True Positive Rate) of fire classification was 96.1%, the FPR(False Positive Rate) was 0%, and the F1-Score, which corresponds to the harmonized mean value of accuracy and reproduction rate of fire and non-fire classification, was 98.01%, and the prediction accuracy ACC(Accuracy) showed excellent performance of 98.05%. In the future, we intend to develop it into an intelligent fire detector system through mesh network monitoring and multi-sensor self-diagnosis functions.

본 논문은 화재 감지 및 신속한 대응을 위한 Wi-Fi Mesh 기반의 화재 감지 시스템을 제안한다. 기존 화재감지기는 단일 센서 기반으로 화재와 비화재에 대한 판별이 어렵고 다수의 감지기가 독립적으로 동작하기 때문에 상호 연계성이 부족한 문제점을 가진다. 본 논문에서는 다중 센서 기반의 화재감지기를 통하여 K-NN 분류 모델 기반의 화재 감지 시스템을 제시한다. 또한, 화재 감지용 메시 네트워크 구축을 통하여 공간 범위내에 있는 감지기가 상호 연계되어 화재를 감지할 수 있게 한다. 구현된 시스템의 성능평가 결과를 살펴보면, 화재 분류의 TPR(True Positive Rate)는 96.1%, FPR(False Positive Rate)는 0%, 화재와 비화재 분류의 정밀도와 재현율의 조화 평균값에 해당하는 F1-Score는 98.01%로 확인되었고 예측 정확도 ACC(Accuracy)는 98.05%의 우수한 성능을 보였다. 향후, 메시 네트워크 모니터링 및 다중 센서 자가진단 기능 등을 통하여 지능형 화재감지기 시스템으로 발전시키고자 한다.

Keywords

References

  1. The Kyunghyang Shinmun, "Fire trucks dispatched with alarms sound, 29 out of 30 times 'in vain'", Sep. 2023. https://www.khan.co.kr/national/national-general/article/202309191603001
  2. S. F. Sulthana, C. T. A. Wise, C. V. Ravikumar, R. Anbazhagan, G. Idayachandran and G. Pau, "Review Study on Recent Developments in Fire Sensing Methods," in IEEE Access, Vol. 11, pp. 90269-90282, Aug. 2023. DOI : 10.1109/ACCESS.2023.3306812
  3. D. -L. Nguyen, M. D. Putro and K. -H. Jo, "Lightweight Convolutional Neural Network for Fire Classification in Surveillance System," in IEEE Access, Vol. 11, pp. 101604-101615, Aug. 2023. DOI : 10.1109/ACCESS.2023.3305455
  4. Ryu Jinkyu, and Kwak Dongkur, "A Study on Flame and Smoke Detection Algorithm Using," Convolutional Neural Network Based on Deep Learning, J. Korean Soc. Hazard Mitig., Vol. 20, No. 1, pp.223~232, Feb. 2020. DOI : 10.9798/KOSHAM.2020.20.1.223
  5. Han-Sol Bang, Joo-Hyung Lee, and Jin-Ku Choi, "Implementation of Fire Detection with Color Filters and Deep Learning," Journal of KIIT. Vol. 21, No. 10, pp. 137-145, Oct. 2023. DOI : 10.14801/jkiit.2023.21.10.137
  6. Yun-ji Kim, Hyun-chong Cho, "Detecting Location of Fire in Video Stream Environment using Deep Learning," The Transactions of the Korean Institute of Electrical Engineers, Vol. 69, No. 3, p.474-479, March 2020. DOI : 10.5370/KIEE.2020.69.3.474
  7. M. F. Ruzaij Al-Okby, T. Roddelkopf and K. Thurow, "Low-cost IoT-based Portable Sensor Node for Fire and Air Pollution Detection and Alarming," 2024 IEEE Sensors Applications Symposium (SAS), Naples, Italy, pp. 1-6, Aug. 2024. DOI : 10.1109/SAS60918.2024.10636455
  8. Chai, Y., Zeng, XJ. & Liu, Z., "The future of wireless mesh network in next-generation communication: a perspective overview," Evolving Systems, Vol. 15, pp. 1635-1648, Aug. 2024. DOI : 10.1007/s12530-024-09583-8
  9. ESPRESSIF, SDKs, https://github.com/espressif/esp-mesh-lite
  10. M. Pujari, C. Awati and S. Kharade, "Efficient Classification with an Improved Nearest Neighbor Algorithm," 2018 Fourth ICCUBEA, Pune, India, pp. 1-5, April 2019. DOI : 10.1109/ICCUBEA.2018.8697515.