A novel method to correct the bias of curvilinear light stripe center

Abstract

Light stripe center extraction is an important operation for the line-structured light measurement system. Common methods use Gaussian kernels to smooth the image before extracting, but it changes the grayscale distribution of curvilinear light stripe, which causes the bias towards the center of curvature. A novel method is proposed to correct the bias of curvilinear light stripe center. The method involves two steps: first, the new images similar in nature to the linear light stripe are formed by the gray values of the light stripe center before correction and its surrounding pixels. Second, the first derivative zero-crossing detection is applied to the new images to obtain the corrected center points. Simulations and experiments show that the bias of curvilinear light stripe center caused by the Gaussian smoothing can be effectively corrected by this method.