Moving Object Detection by a Ghost Cancel Method on Indoor Images
Hiroyuki Utsunomiya, Shigeru Fujita · 2010
There are many behavioral-recognition camera for people or pets and such as street surveillance camera became to use camera at many place. However, the amount of information is increasing, and in the manual analysis of all things and cause cost and oversight, to collect data for learning to act like a machine can recognize, it is difficult to collect large, the necessity of analyzing a camera automatically is increasing. In this paper, we propose a method to extract moving object from room of camera image. The dynamic background subtraction methods have been proposed as a method to extract it, but there has several technical problem that the person to be buried with things they accelerate the rate of updating the background that the false detection is prolonged when the speed of the background updating is slowed down. In this paper, the estimation and the ghost or what caused the change in daylight or motion detection area, according to the results, or the latest update of the background rate determined for each background detection area, another area of dynamic background update propose a method. The evaluation experiment, giving rise to changes in daylight and the opening and closing the blinds in 328 video frames, 92% and the recognition of motion in the frame of things confirmed this approach is effective.