The Detection And Alarm Technology Research Of The Infrared And Visible Integration For Key Monitoring Area

For the increasingly tensing of security situation, intelligent video surveillancesystem has become an important means of preventing public places from terroristattacks. The technology has a great potentiality and economic value in environmentalmonitoring of battlefield, safeguard of important regions and traffic surveillance. Toovercome those problems familiar in current application systems, such as weakness ofreal-time process, singleness of sensor types and low level of intelligence decision, thispaper attempts to take advantage of infrared and visible sensors, and explores the alarmstrategy of abnormal event and intelligent monitoring for special surveillance area.The main contents are as follows:Firstly, aiming at the characteristic of monitoring and detecting abnormal event inspecial surveillance area, we propose an event detection and analysis framework withcontext knowledge. In the framework, we focus on key elements in special surveillancearea and construct the contextual knowledge with spatial context, temporal context andobject context. Then we analyze and refine five typical abnormal events and givecontext representation of two of them for surveillance.Secondly, we design a general process for video abnormal events detection.Combined with typical methods of target detection and tracking, we complete thefeatures extraction of moving target in the fixed scene with fixed viewpoint. Combinedwith context knowledge of different abnormal events, we design feasible fuzzymembership function for different event elements according to corresponding eventunits respectively, and make on intelligent soft-decision of abnormal events based onfuzzy multi-factor comprehensive judgment method.Lastly, we show a small software demonstration system which makes full use ofinfrared sensor and visible sensor to perform a specific monitor and alarm task. Thesystem has two trigger mechanisms, which are human-computer interaction alarm andintelligent alarm. It applies data fusion on the pixel level to assist human visionjudgment. For intelligent alarm, we use previous work to detect abnormal eventsthrough context knowledge. In addition, our system performs with appropriate alarmaccording to comprehensive reliability value.Above methods’ feasibility was verified by real data downloaded from the web andreal scenes recorded.