Due to the exponential increase in video data, an automated examination of videos has become essential. A significant requirement is the capability of the automated video summarization process, which is helpful in vast application areas from surveillance to security. It assists in monitoring the user application with reduced memory and time. Therefore, this paper designs an automated deep learning-based video summarization approach for forest fire detection (ADLVS-FFD). The ADLVS-FFD technique aims to summarize the captured videos and detects the existence of forest fire in it. In addition, the ADLVS-FFD technique involves different subprocesses such as frame splitting, feature extraction, and classification. Besides, a merged Gaussian mixture model (MGMM) is used to extract keyframes and features. Moreover, the long short-term memory (LSTM) model is employed to detect and classify input images into normal and forest fire images. To ensure the better performance of the ADLVS-FFD technique, a comprehensive experimental validation process takes place on a benchmark video dataset. The resultant experimental validation process highlighted the supremacy of the ADLVS-FFD technique over the recent methods.
Deep Learning-based Video Summarization
