Human Activity Recognition in Video Surveillance Using Long-Term Recurrent Convolutional Network

Identifying and classifying human activitiesbased on data collected from sensors such as accelerometers and gyroscopes is known as human activity recognition (HAR). HAR has potential applications in various domains, including healthcare, sports, and surveillance. Public safety and security have become increasingly important in today's society, leading to a growing demand for automated human activity detection systems that utilize real-time video surveillance footage. These systems can help detect and prevent potential threats such as vandalism, theft, violence, or trespassing, allowing for a prompt response to incidents. These systems must be able to handle complex scenes, such as crowded areas, occlusions, and changes in lighting conditions. It is a complex problem because of the large number of observations produced each second. Another challenge is that prediction of movements is a complex activity. It is challenging to monitor public spaces continuously; therefore, intelligent video surveillance is required to monitor human activity, classify it as usual or unusual, and generate an alert. The proposed system will use footage from the CCTV camera to monitor human behavior and warn of any suspicious event. This article uses computer vision and artificial intelligence to automatically identify and flag potentially suspicious or dangerous behavior in real-time video surveillance footage. The combination of two neural networks - CNN and RNN in LRCN enables the recognition of both spatial and temporal features of human activities, making it an effective tool for detecting abnormal events. Thus, the proposed model implements principles of human activity recognition by using video footage as input and identifies the activity being performed in the input video.