CNN-LSTM Based Smart Real-time Video Surveillance System

Modern surveillance technologies are intended to be precise, cost-effective, and run without the need for human interaction. In this regard, few topologies were employed in the previous decade which treated human movement as traditional object detection. Because of this approach, suspicious behavior was also tackled as an object, which made the existing human activity recognition (HAR) systems slow and inefficient when it came to classification for real-time application. In these recent years, machine learning algorithms have made significant advancements in the time-dependent classification of events. This research presents an HAR system that employs a Convolution Neural Network (CNN) to extract spatial information along with a Long Short-Term Memory (LSTM) approach for the rapid and precise sequential tracking of an identified object. This CNN-LSTM technique not only lowers the model's complexity but also improves its accuracy which allows it to be executed in real-time. Therefore, the proposed CNN-LSTM approach can detect suspicious activities in real-time at 10-13 FPS and obtain the best tracking performance in any circumstance while implemented on Raspberry Pi which works as a standalone system.