Research On Moving Object Detected Technique For PTZ Cameras

Moving Object Detection plays a very important role in intelligent surveillance, it has been used in static cameras to assist staffs for monitoring. Comparing to static cameras, PTZ cameras has many advantages; it can be used for wide area surveillance. Because of the camera motion, Moving Object Detection for PTZ cameras is hard to solve, and there are no satisfactory solution. In the thesis, we research on the problem of moving object detection for PTZ cameras, mainly do work as follows.(1) Proposed a framework for solving the moving object detection problem for PTZ cameras. In the framework, we remove the motion of cameras contained in videos by Motion Estimation technique, and then set up the Panoramic Background Model, and the detect the moving object by background subtraction, at the same time, update the Panoramic Background Model.(2) We did a lot research on Panoramic Background Modeling and Updating methods, proposed a new method on the basis of Median Filter method for the characteristics of PTZ cameras. The propose method can generate a Panoramic Background Image containing no moving objects, but with features keep well. We also improved updating method for model.(3) We researched a lot on the background subtraction moving object detection method. For the sake of abiding error cause by Motion Estimation and Background Modeling, we proposed a new method named Neighbourhood-Redundance Background Subtraction, It keeps the good detect result in the new framework.In this thesis, we did sufficient experiments with the video in and out door taking by PTZ cameras, the results show well proving that our framework and improved method effective.