Design And Implementation Of Anomaly Target Recognition System Based On Historical Data Mining

Various kinds of hidden maritime security issues are increasing and maritime situation awareness has become a major issue since the increase of maritime traffic mobility.Nowadays,multiple sensors,such as radar,HD cameras and Automatic Identification Systems(AIS)are applied extensively in the fields of marine environment monitoring.However,the utility of this widely use is limited,mainly due to the only propose is graphical demonstration of raw detection data.How to analyze the maritime target behaviors,understand anomaly events and provide early warning potential maritime risks based on the raw data from sensors is the key to the functional development of maritime situation awareness system.Although at present there are studies on the automatic detection of anomaly behaviors of maritime targets,most of them are based on the data-driven method,which is to identify the abnormal trajectory through studying the AIS trajectory and the target normal motion model.Such method is based on only detection method and it cannot be applied when there is AIS spoofing or silence occurring;it is difficult to explain globally the target behavior only by studying the target trajectory;target normal model learning and abnormal behaviors detection are performed offline which severely restricts the use in real-time monitoring applications;there is lack of user involvement and the use of sophisticated detection techniques during the testing process and these techniques are usually difficult for operators to understand and the classification results are difficult to interpret at the management level.For the above problems,this article designs and implements a set of knowledge-driven anomaly target recognition system based on historical data mining by integrating multiple monitoring methods,multi-dimensional features of targets and multiple time dimensions,based on modelling human knowledge,integrating the semantic traj ectory,semantic attributes and semantic event information of maritime targets in the form of semantic rules.The historical data mining module can extract the multi-dimensional feature information,construct the target information map and establish the target historical knowledge base based on the raw data from multiple sensor while the anomaly target recognition module can fuse the target information in real time,detect the target anomaly behavior and alert the potential maritime risk by combining the target abnormal behaviors based on the semantic rules.The system designed in this paper can be used for real-time maritime situation awareness,which can assist the monitoring personnel to understand the target semantic behavior,pay attention to key abnormal targets and respond to maritime threating events.