Online examination cheating identification method based on multi-instance learning

Abstract

The invention discloses an online examination cheating identification method based on multi-instance learning, and relates to the technical field of examination cheating identification, and the method comprises the following steps: obtaining video information, and extracting video features; continuously sampling the extracted video features, and determining abnormal videos in the video features; generating a fragment label for each abnormal video by using a multi-instance generator for the abnormal videos; through the implementation of the cheating identification method, cheating behaviors can be accurately and easily found, the detection efficiency is relatively high, specific judgment standards exist, the accuracy is relatively high, few cheating detection data can be accurately positioned, the problem of difficulty in collection is solved, suspicious positions in a space can be quickly found, and the accuracy of cheating detection is improved. And the detection problem in a weak supervision mode can be solved, the burden of personnel is reduced, the cheating detection and identification efficiency and accuracy are improved, and the personnel can use the method conveniently.

Description

technical field [0001] The invention relates to the technical field of examination cheating identification, in particular to an online examination cheating identification method based on multi-instance learning. Background technique [0002] Exam cheating means that when the invigilator examines the knowledge and skills mastered by the examinee through written, oral questions or practical operations, the examinee participates in the examination through improper means, and the assessment process is within the scope that is not allowed in the assessment. Seeking or attempting to seek answers is contrary to the principles of fairness and impartiality. [0003] The prior art simply calculates whether the examinee's facial data is greater than a certain threshold. For example, in Chinese Invention Patent No. 112464793, the method of eye tracking is used to determine the deviation angle of the line of sight. Similarly, Chinese Invention Patent No. 110837784 uses the tilt angle of ...

AI Technical Summary

Problems solved by technology

[0004] The purpose of the present invention is to solve existing technology existing in the prior art and just simply calculate whether the examinee's facial data is greater than a certain threshold, such as: the direction of the line of sight, the inclination angle of the head, etc., and the false alarm rate is very high; Face recognition technology can detect whether candidates have left the seat, multi-person cooperation and other behaviors, but cannot further judge behaviors such as turning their heads or prohibited items; and in most online examinations, it is not easy to rely solely on a specific human body information The detection of cheating behaviors, such as using mobile phones, flipping books, etc., is mainly focused on the movement of the upper body, especially the head. The above-mentioned methods for estimating the posture of the whole body skeleton are too broad, the detection efficiency is low, and there are no specific criteria for judging. The accuracy is not high, and it is not practical, and a method for identifying cheating in online exams based on multi-instance learning is proposed

Images