Abstract：The traditional template matching method has low efficiency and low speed. This paper proposed a fast template matching algorithm. In the beginning, just a small part of points were involved in template matching, and gradually more and more points got involved in template matching. Through the comparison of correlation coefficients, it was determined whether to increase the number of matching points or to abandon the current matching position and move to a new position for a new match. When calculating the correlation coefficient, we just calculated with the new points, and then merged it with the original correlation coefficient to get a new correlation coefficient. This greatly reduced the computation of the algorithm. The points involved in calculating the correlation coefficient always were distributed on the template uniformly, which ensured the accuracy of method. The proposed algorithm has high accuracy and high speed, and it can satisfy the request of real-time.