A Survey on Deep Learning based Video Surveillance Framework

Technologies that are used in automated video surveillance give the capacity of automatically identifying security breaches or other potentially dangerous occurrences taking place inside the field of view of the cameras. These technologies have applications in the areas of surveillance as well as the detection of intrusions along perimeters. Recognizing anomalous behavior in crowded environments in a quick and efficient manner using technology is a very successful method for boosting public safety. Several different automated and real-time surveillance technologies for use in security applications are discussed in this study. The fact that public locations cannot be manually monitored is the most important factor in determining their level of safety and security. The algorithms for strange behavior have made efforts to increase their effectiveness, resistance to pixel occlusion, generalizability, computational cost, and execution speed. In a manner analogous to the current state of the art in anomalous behavior identification in crowded settings, researchers broadly divided techniques into distinct categories such as tracking, classification. It has been discovered that hybrid learning approaches and deep learning methods provide more satisfying outcomes during the categorization stage.