In This Research has been studied tracking moving object with filter Kalma. That is based on combination of the forground model in each pixel and Collection of mono hypothesis the background models, all of them are based on the size, position, speed and color distribution of the objects. In order to obtain estimating of object position, filter Kalman combines the estimating of object position based on previous measurements and the estimating of object position based on current measurements. This combination the minimum variance of obtaines, on the other hand, has the best accuracy. A comparable model of speed and side of normal moving is used for primer estimating of object speed. This model usually is an extensive Kalman filter.System works with video frame rate (between 24 and 40). Supposing that separatibility is enough and the size of the object is not greater than the size of object in primer measurement. In this study, two models are investigating that consist of girl dancing and vehicles tracking and autonomous conclusion model is proposed, that An has got the factors of both models,with geting the first frame of an image select algorithm tracking related to that image based on block diagram.