Examination room video monitoring abnormal behavior feature identification method

Abstract

The invention provides an examination room video monitoring abnormal behavior feature identification method. The application of multi-view examination room abnormal behaviors is considered; abnormal behaviors of the examination room are identified through video data of two visual angles; a traditional 3D convolutional neural network is improved; the size of the convolution kernel is modified; a network structure is adjusted, and a two-way 3D convolutional neural network model is designed; the double-camera environment of the examination room is combined; two paths of video data are used as input; video data of different visual angles are utilized so that the defect of a single visual angle is overcome; by use of the trained two-way 3D convolutional neural network model, the behavior characteristics of the examinees in the examination room are extracted, then the two paths of behavior characteristic vectors are fused, the fused characteristics are classified and recognized through libSVM, accurate examination room abnormal behavior recognition is achieved, and the method is very easy to achieve through software and can be widely popularized to various Chinese university campuses.

Description

technical field [0001] The invention relates to a method for identifying abnormal behavior characteristics of examination room video monitoring, which belongs to the field of visual computing and computer vision, uses machine learning and deep learning technology, and improves the network structure on the basis of traditional 3D convolutional neural networks. Background technique [0002] At present, intelligent video surveillance is already an important development direction of video surveillance, and examination room surveillance has also been extended to various middle and high school campuses across the country. Intelligent examination room monitoring will play an increasingly important role, achieving the functions of alarm during the examination and retrieval and retention after the examination. [0003] The main research directions of intelligent video surveillance include: motion detection, target classification, human body tracking, behavior understanding and descri...

AI Technical Summary

Problems solved by technology

However, in the environment of the examination room, the examinee’s movement range when answering the test is small. Even if there is abnormal behavior, the movement scale is generally not too large. The effect of the traditional recognition method is not very good. In the recognition of small-scale behavior There are still some shortcomings

Images