Research On Moving Object Detection Based On Independent Component Analysis

Video moving object detection is a important part of intelligent monitoring, movement image analysis and human-computer interaction technology. The purpose of target detection is to find the motion information in video sequences, and its location. This process can establish the foundation for identification and analysis of moving target tracking, and simplify the difficulty. There are two different conditions in the background of moving object detection. One is dynamic background, and the other is static. The research object of this paper is carried on in the static background. So far, the focus of research scholars is mainly on studying a moving object detection algorithm which is stable, reliable, and robust.The traditional object detection methods are frame difference, background subtraction and optical flow method. These methods have advantages and disadvantages, and the scope of application are subject to certain restrictions. In this paper, the moving object detection method is based on independent component analysis(ICA).This method regards the moving object and static background as two independent source signals, separates the observed signals using ICA estimation by maximization of non-gaussian, and then find the moving target. The advantage of this method is that it has good stability, reliability, and robustness in complex noising environment.Video moving object detection system requires us not only to detect the moving target, but also to locate the target. In the case of complex noising environment, there will be a large number of noise in the foreground images which are separated by ICA. In addition to these simple multiplicative noise and additive noise, there are a lot of noise introduced by calculation. If we do not remove these noises, they will create great difficulties for targeting. So far, the most commonly image denoising methods include median filter denoising, wavelet denoising, morphological filter denoising, segmentation denoising based on edge extraction, and so on. But we can't get a good result by using these methods in this paper. So we analyze the noise sources, and propose a new image denoising method based on image segmentation and the combination of median filter.