The Research And Application Of Head Movement Analysis In Examination Video Monitoring System

Computer intelligent video monitoring system is one of new arising high-tech applicantion fields,It high degree multi-disciplinary of integration computer science, artificial intelligence,machine vision,pattern recognition,image engineering etc. The role of the Intelligent video monitoring system change from "first record,latter evidence" to "abnormal warning, sub-shot", through the rapid and effective positioning, efficient processing and analysis, clear and orderly record, achieve the automatic record of the multi-objective abnormal behavior.This monitoring system is divided into three parts,first candidate required to detection and extract monitoring objectives, and also need detect and determine the normal body posture and scope of legitimate activities, second using pre-defined rules of cheating analyze, determine activities of every objects, and at last determine the suspected cheating goal and record the action.For the target extraction algorithm different mechanisms, analyzing and comparing motion detection methods on the foundation of background subtraction algorithm, frames difference algorithm. Using combination of background difference method inter-frame difference methods to achieve the extraction of moving targets, to improve the detection accuracy and image quality. Comparison for gaussian filtering, median filtering, bilateral filtering, and improved gaussian filtering algorithm , In ensuring the operational efficiency while suppressing noise and retains its edge well; Analysis a variety of basic methods of of the human target for extracting and combining the characteristics of this monitoring system environment, used of the human target extract algorithm based head detection, that is to use AdaBoost algorithm to establish the head detection cascade classifier, optimized and improved its classifier training process ; According to the examination room monitoring environmental stability and the target range of motion is small, non-linear characteristics , improve of the CamShift algorithm, make that have light robustness in the process of tracking;Used to the rules of cheating on the extract real-time target detection, and determine the target whether have suspected of cheating ultimately.