Advance fire and smoke detection in trains
| dc.contributor.author | Manu M 1NH21EC090; Nandeesh Gowda B 1NH21EC102; P Hemanth Kumar Reddy 1NH21AI127; D Rahul 1NH21AI128 | |
| dc.date.accessioned | 2025-05-14T08:50:39Z | |
| dc.date.available | 2025-05-14T08:50:39Z | |
| dc.date.issued | 2025 | |
| dc.description.abstract | Fire and smoke detection systems play a vital role in ensuring safety and reducing losses caused by fire-related incidents. However, conventional systems that rely on heat and smoke sensors are often susceptible to false alarms and may fail to detect fires at their earliest stages. This paper presents a modern, cost-effective, and intelligent approach to fire and smoke detection, utilizing the ESP32-CAM microcontroller combined with a Convolutional Neural Network (CNN). The ESP32-CAM, a compact and affordable IoT-enabled microcontroller with an integrated camera, serves as a reliable platform for real-time image capture and processing. When paired with CNNs, a cutting-edge deep learning architecture for image recognition, the system provides a robust and highly accurate solution for detecting fire and smoke. By integrating IoT and AI technologies, the proposed system bridges the gap between affordability and advanced functionality, ensuring timely and precise alerts. The system architecture comprises three core components: the ESP32-CAM for image acquisition, the CNN for image-based analysis, and a communication interface for transmitting alerts. The ESP32-CAM continuously captures images or video streams, which are then preprocessed to enhance detection accuracy. Preprocessing techniques, such as resizing, normalization, and noise reduction, prepare the visual inputs for the CNN. The trained CNN identifies critical features in the data to detect fire and smoke reliably. At the heart of the system lies the CNN, optimized specifically for deployment on resource-constrained devices. Its lightweight design ensures compatibility with the ESP32-CAM’s limited computational capabilities while maintaining high accuracy. Optimization techniques like model pruning, quantization, and transfer learning help reduce computational demands without compromising performance. Moreover, the modular nature of the system allows for easy integration with existing fire safety setups and future upgrades. Enhancements such as incorporating additional sensors (e.g., temperature or gas sensors) or adopting more advanced machine learning models can further improve detection capabilities. | |
| dc.identifier.uri | http://192.168.75.5:4000/handle/123456789/18911 | |
| dc.language.iso | en | |
| dc.publisher | NHCE | |
| dc.title | Advance fire and smoke detection in trains | |
| dc.type | Learning Object |