Student Surveillance System for Detecting Abnormal Incidents in the Campus using Deep Learning

Video surveillance has evolved as a challenging research area for video processing and analysis. As a result, a substantial study has been undertaken to increase performance in the detection of anomalous activity. These video surveillance systems can also be used for building applications to monitor the students in schools and colleges to detect and notify any abnormal incidents happening to them so that immediate care can be taken. The proposed model is to monitor students in their campus for detecting any abnormal conditions like fainting, bike or car accidents, and dog biting. The primary purpose of this work is to render unusual incident identification from video sequences of individuals and groups to identify various types of abnormal conditions among students. An image dataset related to abnormal incidents is prepared and used to train the CNN model. Images are gathered by labeling each frame with a set of student actions. This model will capture the movements of students when any motion is detected. The captured movements of students will be classified according to the type of abnormal incident that occurred. As soon as any abnormal incident is detected, the model alerts with an immediate warning message to the health care center in the campus.