Abstract: Fire outbreak is the common issue happening everywhere and the damage caused by this type of incidents is tremendous towards nature and human. Vision based fire detection system have recently gained popularity as compared to traditional sensor-based fire detection system. However, the detection process by image processing technique is very tedious.
We proposed a fire detection algorithm using Convolutional Neural Networks (CNN) to achieve high-accuracy fire image detection, trained on a dataset consisting of 3696 fire and 541 non-fire images, totaling 4237 images. Out of these, 2857 images were used for training and 1921 for testing. The model employs convolution, activation functions, and max pooling operations. Through experimentation with different batch sizes and epoch values, we achieved a model accuracy of 94%, correctly predicting 1817 out of 1921 test images.
Our study also reviews smoke and fire detection techniques, emphasizing early detection's importance. We discuss various methods, including RGB and HSI models, aiming to minimize false alarms through optimized technologies.
Furthermore, we address human-animal conflicts in forest zones by developing a monitoring system using cameras to detect intrusions, classify them, and notify relevant authorities.
Copyright Form