Abstract: A new idea with GM（1，1）model to predict the change of target feature is proposed in this paper. The method divides the target area into blocks，and calculates the sum of the pixel in the block area. At the same time，the short time series are used to cumulativly process the pixels sum sequence to generate a new sequence. The target prediction model is obtained with GM（1，1）model. GM（1，1） pixel prediction model has a strong predictive ability for the target，and has a better predictive ability for the short-term characteristics change. In this paper，a new effective tracking algorithm is proposed，which combines the feature change with the prediction. After testing the proposed method，it is found that the tracking performance is improved significantly compared with the traditional tracking algorithm.

Key words: GM（1, 1）, pixel prediction, time series, target tracking