In line structured light vision measurement, the extraction accuracy and precision of the light stripe center point directly affects the final accuracy of the measurement results. The contradiction between high capacity of anti-interference, good robustness and intensive computation exists in optical center extraction algorithm, so a cross-correlation stripe center extraction algorithm is proposed. The gradient threshold is used to separate the effective stripes region, and correlation is applied between cross-correlation coefficient and corresponding gray scale. The stripe with cross-correlation′s extreme value is selected as the initial center, and then the curve fitting method is used to refine the accurate stripe center. Here the cross-correlation′s value is used to evaluate whether the stripe center exists or not, and the similarity constraints of gray scale and position in the neighborhood stripes are employed to eliminate the noise. The experimental results show that the algorithm has high stripe extraction accuracy and calculation speed. Compared with current algorithms, it is simple and practical, and also has good robustness and the ability of anti-noise repair on disconnected stripes.