The invention discloses a workpiece surface detection method. An input image of a workpiece is converted into a grayscale image, and is subjected to binary processing, then an outline of the workpiece is extracted, the grayscale image of the workpiece is extracted, through filter processing of the grayscale image, a strip-shaped texture which is in an included angle of five degrees in the vertical direction and the horizontal direction of the surface to which the workpiece belongs is filtered out, edge sharpening processing and self-adaptive binary processing are conducted, then an outline image is obtained, a patch with a preset size is utilized to be subjected to outline detection processing, accordingly a corresponding characteristic vector is obtained, the characteristic vector is substituted into a beforehand trained SVM model for judgment calculation, a result of the judgment calculation is output, and thus an image of a corresponding range of the workpiece can be accurately and fast extracted. The calculation amount of a subsequent processing process is drastically reduced, meanwhile, interference can be avoided, the precision of judgment is improved, and the robustness is improved at the same time.