In order to rapidly detect the space debris in the Geosynchronous Earth Orbit (GEO) image with optical observation, a grading detecting method based on Signal-to-noise Ratio(SNR) was proposed. First, the real measured data are analyzed, and the characteristics that image velocity of different GEO debris are same is acquired. An improved dynamic programming method is proposed based on the velocity information obtained from high SNR debris, which can detect low SNR debris effectively. The results of real image show that the detection performance of the grading detecting method is the best when the search window of recursive equation is 5 pixel×5 pixel, and the detection time is reduced by more than 87% as compared with that of original algorithm when the cumulative number is greater than 5. With the advantages of high detection rate, low false alarm rate and low computational cost, the proposed method is suitable for the detection of space debris in GEO optical observation images.