Moving Objection Detection Based On Adaptive Mixture Gaussian Models

Nowadays, intelligent control are used more and more widely in various occasions, and is very important in intelligent video surveillance. Moving target detection has become a research focus, and it influent the moving target classification, tracking, understanding behavior, and so on. It has great theoretical significance, and also important practical application value. It has fusion of image processing, computer vision, artificial intelligence, pattern recognition and other multidisciplinary technology. It has so much methods, but no matter what method used, people are always looking for an accuracy, speed and good robustness algorithm. This adaptive mixture Gaussian models, whose purpose is to solve the moving object detection in complexing and changing scene.In this paper, use static camera video in complex scenes as the research object, committed to target calibration, remove the shadow, target estimate, and fog rain pollution image restoration. In the paper, the following mainly work is:1, Comparison of the several basic image processing algorithms, the median filtering has the calculation speed and better effect. And combined it with other algorithms to get an improved algorithm. Not only good to remove the salt and pepper noise, but also good to remove the Gaussian noise.2, Because the background difference method is better than other algorithms in stable scene,and by repeatedly setting the mixed Gaussian models parameters to obtain a algorithm in unstable scene.3, Improving the mixed Gaussion models, and get a fast algorithm, In order that it can be used in practical applications. And aim to some of the actual situation, combined with the corresponding algorithm to get improved mixture Gaussian models.