Research On Detection Method Of Human Abnormal Behavior In Micro-geographic Environment

Micro-geographic environment is the environment which contains rich semantic information connected outdoor and indoor of three-dimensional topology environment. Criminal acts such as ATM robbery, burglary, theft, torture, prison Pa often happened in micro-geographic environment, so it is a challenge to manage the urban society and government well. In recent years, bad effect has been made by cases like "Zhao Zuohai injustice ","Zhang Gaoping-Zhang Huishu injustice ","Li Huailiang injustice". It is urgent for public security bureau to protect the legitimate rights and interests of suspects, and take full advantage of the technological means of miscarriages of justice is becoming a trend. With the development of intelligent video analysis, computer vision and3DGIS technology makes it is to achieve real-time dynamic display and analysis of human behavior in the micro-geographic scene. Technological police is the development strategy of wisdom police. With the advantage of the intelligent video analysis, GIS, GPS, big data and cloud computing, makes it possible for public security police working with modern, intelligent and process-oriented visualization. As an important part of the wisdom police, the wisdom police room makes interrogation becomes more intelligent, open and share. Prevention of torture is the key part of the wisdom police which is a useful tool to protect the legitimate rights and interests of criminal suspects.Based on background above, this paper try to study the abnormal behavior which happens in interrogation room. As the target of abnormal behavior detection, extort confessions by torture was studied in micro-geographic environment. The results are as follows:(1) A priori knowledge library of abnormal human behavior detection of micro-geographic environment. The geometry topological relations and the scene semantics was analyzed, and abnormal behavior in the interrogation room was classified as heavy violence, light violence and soft violence, after which each detection methods of violence was also classified. Information of environment, abnormal behavior and its detection method was summarized as three main aspects of the existing abnormal human behavior detection to achieve the construction of detection method priori knowledge.(2) Spatial characteristics of interrogation room scene and the vision technology were used to build three-dimensional reconstruction in a human micro-geographic scene. Three-dimensional pose of human body and body silhouette were used to estimate the unit specifications of human three-dimensional model. Combined with single image measurement technology, geometry parameters of human body and geospatial location were determined which can be used to build the three-dimensional alert protection of human bounding volume, which provide the environment constraints for abnormal behavior detection and recognition.(3) Based on the priori knowledge of interrogation room, algorithms about human abnormal behavior detection and recognition which take environment in account was proposed. The detection process was divided into two levels by three-dimensional space constraints:yellow warning and red alarm. The detected behaviors were divided into heavy violence and light violence by behavior constraints. Body contour, moving history, optical flow, and kinetic energy characteristics were used to express abnormal human behavior. The algorithm was designed and implemented, and experiments were made using data taken from an interrogation room.(4) Based on research results above, a prototype video surveillance system was designed and implemented for micro-geographic environment. The system achieves three-dimensional of body modeling, detection and identification of abnormal behavior.