Extracting The Essence of Video Content: Movement Change based Video Summarization using Deep Neural Networks

The security and safety of the general public is ensured by the use of intelligent video surveillance systems in a variety of applications to monitor events and evaluate activities. One or more monitoring cameras generate, analyze, and store a large volume of data for security reasons. Video surveillance is an important part of maintaining security. Yet, since there will only be motion for a small duration of the picture, it may be difficult to identify any motion or movement in a replay version. Analyzing the video would take a lot of time, and pinpointing the exact frame at which change took place wouldn't be easy every time. At present the state of art results are having different issues basis on the different events of the video these proposed model captures the main content better than already existing methods. Researchers investigate Open CV for key frame extraction and recognizing distinct frames to summarize the video. This study uses a DNN model, which is trained by using a Mobile-Net SSD Caffe model with proto-txt to recognize objects, and then that same DNN is used to identify humans. Due to the complexity of routine indoor and outside monitoring as well as the processing rate required for real-time detection, the accuracy of behaviour identification gets improved.