During the service of vacuum chamber blanket module of the fusion reactor, cracks may occur on the surface of tungsten layer due to the high heat load, posing potential safety hazards and significantly affecting the structural integrity and service life of the fusion reactor. Therefore, implementing effective nondestructive testing for surface cracks in the tungsten layer of water-cooled blanket modules is crucial. Numerical simulations were conducted to investigate the response characteristics of Ka-band microwave detection signals in defect-free and cracked regions of the tungsten layer. Based on the simulation results, the experimental platform for microwave detection of tungsten surface cracks in water-cooled blanket modules was established. A coaxial probe operating 38 ~ 40 GHz was used for planar scanning of tungsten specimens with cracks of different depths. At each scanning point, an amplitude cepstrum method and a clutter-suppression method based on singular value decomposition were applied to process the detection signals, enabling effective crack imaging and localization. The localization error was within 0.5 mm. Furthermore, time-domain signal features were utilized to establish a mapping relationship between signal characteristics and crack depth. The assessment error for crack depth was within 2%. The results demonstrate the effectiveness of microwave reflection methods and the proposed approach in imaging and quantitatively detecting surface cracks in the tungsten layer of water-cooled blanket module.