Image-based Security Monitoring Algorithm Research

Moving object detection is one of the most important processes in computer vision and moving image coding and is widely used in many fields. It is a key algorithm especially for the intrusion alarming of security surveillance. Moving object detection involves two phases of processes. One is the verification of an object in image sequences and locating it precisely. As the background of surveillance is complex with swaying vegations and changing illumination and vibration of the camera in the scene, the precise detection of moving-objects are still a challenging problem.Some common methods in the field of object detection, such as Temporal Difference method, Background Subtraction method and Optical Flow method can be used. Each method has its advantages and scopes, and they will be introduced in more detail in this paper.Image sequence background can be divided into two categories: one is still, the other is moving. There are many reports in the first category, and these reports are mostly based on Background Subtraction. According to the characteristic of the indoor surveillance, this paper gives an adaptive mixture background model for moving object detection, which is insensitive to changing illumination, and the moving distance analysis is computed in order to detect abnormity. All these processes can be fast performed.A preliminary investigation is made to object expression and description. Firstly we researched several main object expression methods and intruduced several object description methods such as area, perimeter, etc, which can be used to analyse and track moving-objects and reduce false alarms.