Aiming at watermark defects of solar silicon wafers,a detection method based on time-frequency domain hybrid analysis was proposed.On this basis, the hardware system of the solar silicon wafers surface defect detection was designed.First image preprocessing was executed in time domain for collected silicon images,then frequency spectrum images were obtained by real fast Fourier transform,and the optimal Gaussian filter was designed for filtering, and then reconstructed images were obtained through the real fast inverse discrete Fourier transform,and the difference images were obtained by difference method in time domain. Finally,the defect areas were acquired by threshold and morphological operation.Experimental results show that this method has high accuracy on defect detection for solar silicon wafers samples including watermark defects and it has high practical value.