The Algorithm Of Moving Object Detection Based On Background Modeling

In recent years, moving object detection has become a hot topic in the interdisciplinary fields of applied mathematics and computer vision. Moving object detection, which picks up the interesting moving object, is one of the key technologies in intelligent monitoring system and it directly affects the following technology such as moving object recognition and tracking.However, it is a challenge to detect moving object accurately in complex dynamic scene with the background disturbances such as waving trees、moving water and fountain. So, this paper focuses on the problem of overcoming the frequent changes in background areas and proposes an effective method of moving object detection based on background modeling. The main contributions are summarized as follows:Firstly, a background modeling algorithm of moving object detection based on dynamic distribution of model numbers is proposed. The main idea is that when the total model numbers of image pixels are fixed, the pixels in frequent changing areas can "borrow" models from the pixels in relatively stable areas if the model numbers of each pixel and the total model numbers of image pixels are both not exceed their upper limits. Thus, background model numbers are dynamically distributed and the parameters are updated by deleting, increasing or replacing models based on the model transformation rule to achieve the generation of background models, which makes the background model numbers adjust the dynamic scene. And then, background modeling is combined with spatial and temporal information to estimate the background or foreground pixels, in order to eliminate the effect of dynamic background disturbance.Secondly, in the processing procedure of dealing with the illumination, an improved feature based on HOG(Histograms of Oriented Gradients) can effectively describe the shape feature of objects. Then, a foreground verification approach based on Bhattacharyya coefficient is proposed to overcome the false detection caused by changing illumination.Thirdly, the algorithm in the paper is analyzed and experimented, and the experimental results under various scenes illustrate that the method can detect moving object accurately in complex dynamic scene with the background disturbance such as waving trees, moving water, fountain and changing illumination. Comparing with traditional moving object detection methods, the method is more accurate and real-time.Finally, difficult points in the algorithm of moving object detection based on background modeling are discussed and summarized which need to be considered in the future research of background modeling.