Study On Background Estimation And Updating Used In Video Detection System

Video detection system is a technology which is using image processing technology detect, identify, track vehicle in image sequences and gain the traffic flow information. Common methods are based on comparison with the background. So background estimation and real-time video detection of vehicle are very important.Common algorithms for background estimation such as average, statistics, block and gaussian mixture model, etc. For bigger traffic flow often can't gain a satisfactory result. This paper propose an algorithm of background estimation which is combing statistic histogram and average method. The algorithm firstly find several maxima from the histogram, and using the average method to determine the believe degree of the several maxima, and then select the highest believe degree as background. Through lots of experiments, including the lighter and bigger traffic flow conditions. The results indicate that this algorithm can gain satisfactory background in these conditions. The proof can be applied in the real time.Actually, the environment is the dynamic changes, such as light, the sun light changes and shadow area changes. In these conditions it need for real-time background updating. This paper propose a method of background updating based on block. Algorithm according to judge whether the block matching in stable condition in a certain period, then use block mean and variance decided whether this block in the continuous stable state, and decide whether the block need updating. Through a lot of experiments, including the condition in the tunnel, the light changes in the tunnel and the shadow position changes, etc. The results indicate that background updating have better effect, and The proof can be applied in the real time.Through a lot of experiments, the algorithms proposal in this paper can gain satisfactory results and can be applied in the real time.