Moving-object Detection Of Adaptive Gaussian Mixture Model For Varied Background

In the course of moving-object detection, the background models are crucially important for the object extraction, the results of moving-object detection accuracy directly affect the next tracing accuracy. At present, the mature and widely used method in moving-object detection is Intro duced-frame Difference, Optical Flow and Background Subtraction,. The Gaussian mixture model (GMM) is one of popular and robust methods in the background models.Gaussian mixture model has a better adaptability to the interference of noise, shadow and light change in the target detection. It can detect a more accurate target quickly, but may have mistakes when testing conditions are special or extremely complex. This paper has study in improving Gaussian mixture model for moving object detection in complex environments.This thesis mainly focuses on GMM detecting background in complicated weather situation. For reducing measuring error of GMM detecting background in complicated weather situation, based on the different motion characters of background and foreground, this paper proposes a new improvement method. It adopt different learning rate to update parameter for the background and foreground regions respectively by means of using pixel analysis to distinguish for the background and the foreground. The experiences results show that the proposed method can not only improve the detection accuracy in unsatisfactory weather and reduce the miscarriage of justice for the foreground of background, but also greatly reduce some noise.This thesis also focuses on GMM detecting background under the influence of light factors. In the course of moving-object detection, the background models are crucially important for the object extraction, the Gaussian mixture model (GMM) is one of popular and robust methods in the background models. But under the influence of light factors it leads to larger error. For the different luminance of the background and the moving-object, Extract the luminance in HSI color space, differentiate them using automatic threshold, reduce the error detection caused by light source change. The results show that the method can improve the detection accuracy under the influence of lighting change and reduce the miscarriage of justice for the foreground and background.