In the three-dimensional measurement based on the linear structured light projection, the extraction precision of light stripe center affects the ultimate measurement accuracy of system directly. Aiming at the problem of broken line defect which exists in stripe center acquired from current light stripe center extraction algorithms, an proposed approach based on curve fitting of the linear structured light stripe center sub- pixel extraction is proposed. The praposed approach uses the erosion thinning method to get the stripe′s basic skeleton. The normal direction of each point on the skeleton line is computed by using the average squared gray gradient. The initial light stripe central points are obtained by weighted barycenter method. Sectional cubic polynomial curve fitting method is used to achieve smooth light stripe central positions in sub-pixel. The experimental results show that method can improve the broken line defect and the accuracy of light stripe center extraction.