The invention discloses a wafer surface defect detection method based on Gabor features and random dimension reduction. According to the method, a CCD camera is used to collect a wafer surface image,and then the image is preprocessed; 40 Gabor filters used for acquiring wafer surface texture features are designed, and then convolution operation is performed on the 40 Gabor filters and the image to obtain 40 feature images; and random dimension reduction is performed on the 40 feature images, wherein threshold partitioning is performed on the images obtained after dimension reduction, an objective function for a partitioning threshold is constructed, the objective function is solved to obtain a final partitioning threshold, the final partitioning threshold is used to partition the images into a foreground and a background, the partitioning threshold is determined, and a wafer surface defect is accurately detected at last. Through the method, the wafer surface defect can be well recognized and positioned, and recognition efficiency is greatly improved.