Research On Wireless Embedded Video Monitoring System And Method Of Feature Extraction

With the increasing safety requirements, the superiority of the video surveillance systemhas also been a great demonstrated. an important standard measure of a monitoring system isthe ability to get better security through effective monitoring. Therefore, the development offlexible and reliable, cost-effective wireless video surveillance system is of great significance.Based on the needs of intelligent video surveillance, first designed a video surveillancehardware platform based on the i.MX27, ARM Linux, use H.264full-duplex codec functionsof i.MX27, do video transmission via Wi-Fi module to meet the demand of multi-channelvideo capture and remote real-time video monitoring.About anomaly detection in the intelligent monitoring, do corresponding processing andanalysis on collected video at the application layer. Including the moving target detectiontracking and motion characteristics extraction on the video. In moving target detection usingan improved Gaussian mixture background modeling combined with the backgroundsubtraction method to detect Moving Target Extraction; In moving target tracking, usingKalman filter tracking algorithm to realize the target tracking based on projections bycomputing the centroid distance targeting and calibration; In moving target feature extraction,the study object is abnormal behavior in front of the ATM machine, use the characteristics ofthe trajectory feature extraction method. To verify the effectiveness of the extraction method,based on results of sample video feature extraction, with classification of training data as areference, recognize behavior by the SVM method. The results show that, in moving targetdetection and track, has great improvements on integrity and real-time to moving target, thetrajectory feature extraction in the identification of abnormal behavior has certain significant.Due to time constraints, failure to transplant the application layer procedures to theembedded platform, did not complete a complete hardware and software systems, furtherresearch is needed.