Human abnormal behavior detection using CNNs in crowded and uncrowded surveillance – A survey

Abstract

The demand for surveillance networks is increasing universally on account of decreasing the faith in people. This leads to monitor the people during working, roaming, traveling, and shopping, etc. Thus, a surveillance system is needed to monitor human behaviors as a third eye in crowded and uncrowded indoor and outdoor areas. This system records the incidents that contain the patterns of various human behaviors. The video is checked to identify an incident manually is time-consuming. Hence, an automation system is needed for processing lengthy videos. The growth of graphics processors and Convolutional Neural Networks (CNNs) addresses video processing challenges to identify the incidents. This paper examines human abnormal behaviors using various CNNs to recognize the abnormal behaviors in the video. This study observed that 3D Convolutional Neural Network is performing better than machine learning algorithms. The comparison showed the various CNN model's performance to identify the various human abnormal behaviors with diverse datasets.

Keywords

• Human abnormal behavior
• Falling
• Vehicle theft
• Loitering
• Patient’s abnormal behavior
• Violence and panic