Rubber O-ring seals are widely used in high-demand industrial sectors such as aerospace and petroleum energy, where surface defects directly affect sealing reliability. To address the low efficiency of manual inspection, the missed detection of micro-defects, and the difficulty of quantifying defect depth, a surface-defect inspection method integrating machine vision and laser line scanning is proposed. For flash defects, sector-shaped ROI segmentation, piecewise edge fitting, and abnormal-contour discrimination are employed. For pits, short shots, and adhesion defects, multi-light-source temporal differencing, Gaussian-high-pass composite filtering, and Sauvola adaptive thresholding are combined. For pit-depth evaluation, maximum-similarity template matching and least-squares curve reconstruction are used. Experiments were conducted on 75 samples containing flash defects and 106 samples containing pits, short shots, or adhesion defects. The detection rate for prominent flash defects was 100.00%, and the overall detection rate for pits, short shots, and adhesion defects was 97.24%. The relative errors of area measurement for all four defect types were below 10%. For ten repeated measurements of the same pit, the mean absolute deviation was 13.94 μm and the standard deviation was 1.21 μm. Double-sided inspection of five samples required approximately 2 min. The proposed method integrates micro-defect recognition, area measurement, and depth evaluation, providing a quantitative basis for O-ring quality grading