Design And Implementation Of Regional Intrusion Detection Algorithm Based On IPC

For the sake of security and management it is neccssary to carry Out intrusion detection in some region.Intrusion detection methods are mainly sensor-based and IPC-based,Compared with traditional sensor-based intrusion detection technology,intrusion detection technology based on IPC has more advantages.For example,the alert region can be delimited according to the user needs,it can achieve higher detection rate,it can early warning,When the intruder approaches the alert region.Therefore.this disertation studies an intrusion detection algorithrn based on IPC.This algorithm processes the video sequence transmitted by IPC,when a person or vehicle is found to enter the warning area and the intrusion analysis condition is triggered,the alarm inform the staff.Algorithum includes moving object detection and intrusion analysis. The main research contents of this dissertation are as follows:1.The common moving,object detection algorithm is studied.It is difficult to detect moving object accurately in scenes with changing illumination,dynamic back ground or shadows.To solve these problems,a moving object detection algorithm based on improved Gauesian Mixture Model and shadow removal is proposed in this dissertation.On the basis of traditional GMM algorithm,adjustable learning rate is introduced to update background model,combined weight is introduced to carry out background/foreground segmentation,the method based on color and texture features is introduced to carry out shadow removal,finally morphological processing is carried out to remove the noise.2.The common intrusion analysis mechanism is studied.Intrusion analysis mechanism includes two aspects:warning region and intrusion analysis rule.In order to realize drawing of warning region of any polygon,this dissertation determines the vertex of warning region with the help of OpcnCV library function.In order to increase vigilance when the intrusion object approaches the warning region,this dissertation sets two warning forms:warning and alarm,and judge whether the invasion object enters or leaves the warning area.3.An IPC of HIKVISION is used as video input to analyze the performance of intrusion detection algorithm in this dissertation.CDW-2014 dataset,OTB50 dataset,WallFlower dataset and surveillance video of Xi'an university of technology are used to verify the ability of the intrusion detection algorithm in this dissertation to deal with various complex video scenarios.The experimental results indicate that:The intrusion detection algorithm in this dissertation can not only set up various warning areas according to user requirements,but also accurately warn and alarm even if there is interference such as illumination change,dynamic background and shadow in video scene.In the case of the test,the accuracy rate of intrusion detection is 95.0%,and the false detection rate is 3.9%,which meets the requirements of high detection rate and low false alarm.