fire and smoke tracking and detection in videos based on pyramid convolutional deep learning

Nowadays, jungles are burning into fire due to climate changes. detection and tracking any smoke will be prevent any burning, so it needs pattern recognition in images. in this research we propose a developed method for real-time and synchronous fire and smoke tracking and detection in videos. in this approach, at first, we apply a pre-processing phase to enhance the image frames and then deep learning technique based on pyramid convolutional neural network apply for data training and testing based on fractal model for image segmentation and feature extraction. simulation done in matlab platform which results indicated good results in terms of smoke and fire tracking and detection. also we use some evaluation criteria such as accuracy, sensitivity, specificity, area under curve (auc) with 98.91%, 93.54%, 93.17% and 0.8719 respectively. we use two different image and video dataset in this research and both of them have good performance in terms of accuracy in comparison to recent methods.