Analysis method of cheating behavior in examination room based on motion feature enhancement and long-term time series modeling

Abstract

The invention discloses a method for analyzing cheating behavior in an examination room based on motion feature enhancement and long-term time series modeling, which belongs to the field of video behavior recognition and deep learning. First, a data set is collected, then the data is labeled with behavior categories, and a video stream is extracted as Image frame, based on the method of motion feature enhancement to improve the model's ability to capture moving objects, information fusion between frames is performed by feature spectrum shift, and long-term time series relationship is modeled based on the method of time series pyramid, and the recognition model is completed. Then, the Xavier method is used to initialize the behavior recognition classification model according to the images obtained from the data set, and the sampling sequence of video frames is obtained by segmental extraction, and the loss function based on the classification model is iterated to the preset number of iterations to complete the model. Finally, use the video frame sequence obtained by sampling for inference testing to obtain specific behavior category results.

Description

technical field [0001] The invention belongs to the field of behavior recognition and deep learning, and relates to a method for analyzing cheating behavior in an examination room based on motion feature enhancement and long-term time series modeling. Background technique [0002] Behavior recognition is a research field that has received much attention in the field of computer vision. Its purpose is to identify the behavior category of the current person in the video, so it is considered to be an important basis for video understanding. In recent years, with the improvement of computer computing power, the development of deep convolutional neural networks has made many remarkable achievements in video behavior recognition tasks. [0003] Nowadays, behavior recognition models with superior performance are all based on deep convolutional neural networks, and their powerful feature extraction capabilities have brought great development to various tasks in computer vision. Com...

AI Technical Summary

Problems solved by technology

[0006] The purpose of the present invention is to: provide a method for analyzing cheating behavior in an examination room based on motion feature enhancement and long-term time series modeling, and solve the problem that in samples dominated by motion information, motion features will have an important impact on the results of behavior recognition models, and Static feature information basically does not contribute to the improvement of recognition accuracy. The current general-purpose model cannot directly emphasize motion-related features from the feature spectrum of RGB frame sequences and suppress scene features, resulting in low accuracy.

Images