The marine clutch control mechanism operates under high-frequency alternating stress, with fatigue strength being the key factor affecting the service life of its components. As a complex and irregularly shaped part, the clutch control mechanism lacks specific design criteria and testing methods for fatigue strength. During high-frequency fatigue testing, it is challenging to measure the stress state of the component. To address the difficulty in testing the stress field distribution of components during high-frequency fatigue loading, this paper proposes a high-frequency fatigue strain field testing technique based on parallel matching. Traditional digital image correlation (DIC) methods are too slow for real-time high-frequency fatigue testing. Therefore, this paper presents a fatigue strain measurement technique based on parallel matching 3D-DIC, with key research areas including an initial value search algorithm based on GPUsift, a CUDA-accelerated IC-GN sub-pixel search algorithm, and a strain field parallel computation method based on local least-squares fitting. The proposed method not only meets the real-time requirements of high-frequency fatigue testing for the marine clutch control mechanism but also significantly improves computational efficiency while maintaining high measurement accuracy. This provides crucial support and practical value for the safety and reliability of marine power systems, offering new insights and methods for technological optimization and development in related fields.