Increasing the coherent accumulation time is an effective method to improve radar detection capability. However, when target velocity is high, long-time coherent accumulation will lead to the range migration effect of the target, reduce the Signal-to-Noise Ratio, and affect the radar's detection power. Aiming at the range migration problem, Keystone Transform(KT) correction algorithm is given. The performance of three KT implementation methods are simulated and evaluated. Then, a parallel implementation algorithm based on Chirp Z Transform(CZT) on the Graphics Processing Unit(GPU) is proposed and implemented. Experimental results verify the real-time performance and effectiveness of the proposed method in combination with the passive radar system.