Automated System for Detection of Suspicious Activity in Examination Hall

One among the curses in the education institution of the country is malpractice during written examination, either online or off line. Some of the suspicious activities during written examination includes, copying from neighbouring candidates, other written materials, exchange of answer scripts, etc. These activities lead to physical movement of the face and other parts of the body than the normal writing posture of the candidate(s) seen at the initial stage of the examination. Any deviation to normal posture can be an indicative of suspicious activities. Main objective of this work is to build a real time edge computing-based video analytics and techniques to identify the suspicious activities relating to malpractice case booking as evidence and proofs during written examination. The proposed model includes a low-cost webcam video frame capturing and edge computing for analytics to detect the face and behaviour recognition using the following techniques. Firstly, the CNN is used to detect human faces. Secondly, candidates are recognised using YOLO and features are extracted using Local Binary Patterns Histogram (LBPH). After feature extraction, the activity classification is performed and is used to detect whether the activity is suspicious or not based on the face movement, hand contact detection and body posture using SVM. Lastly, the system will set an alert based on the threshold set for movement pattern of the candidate(s) in the given area focussed by the web cam or any other camera used in an examination centre or for individual online examination for suspicious activities.