A Real-time Fire recognition technique using a Improved Convolutional Neural Network Method

Traditional fire-detection systems may be replaced in future smart cities with vision-based systems that use developing technologies and deep learning to build fire safety in society. It was our goal to build an early warning system for fires that can detect even the tiniest of sparks and sound an alert. An Improved Convolutional Neural Network (ICNN)along with LGBM Classifier was used for detecting fire areas. The reliability of model with fire area forecasts was first tested using the CNN technique. Unfortunately, this strategy of detecting fire incidents failed to provide the expected results in several studies. The classic model was upgraded by adding LGBM in final layer and expanding the amount of training data dataset using data augmentation methods for real-time fire catastrophe monitoring. When used under a variety of weather conditions, the suggested technique effectively recognized and alerted the public to the occurrence of devastating fires by altering the network structure via automated color enhancement and parameter reductions, among other things. In tests, the suggested technology proved to be effective in protecting smart cities and detecting fires in the urban environment. Finally, we tested the classification results produced by our system against those of previously published fire-detection methods that used commonly used performance matrices.