Automatic Fire Detection Algorithm In Roads And Forests Using Convolutional Neural Network

Automatic Fire Detection Algorithm In Roads And Forests Using Convolutional Neural Network

In our research paper, we propose using image processing techniques alongside convolutional neural networks to detect fires at night on roads and in forests. Fires, which generate thermal and light energy through oxidation, pose significant environmental and personal safety risks. The visual characteristics of flames, such as shape and color, vary based on the combustibles involved. Our study used a Kaggle dataset and additional night-time fire images from the internet. We enhanced these images through preprocessing methods like brightness adjustment and noise reduction to aid our neural networks in recognizing fire features under low-light conditions. By employing transfer learning, we utilized pre-trained models to improve detection accuracy and model generalization across different fire scenarios. Our validation tests confirmed the effectiveness of this approach, demonstrating its potential in early fire detection systems to mitigate risks associated with nocturnal fires in varied environments.