The line-structured light vision sensor is widely used in 3D reconstruction and scanning, quality inspection and other fields. In this technique, the segmentation of light stripe image is an important preprocessing steps, which separates light stripe and extract ROI from an original image. The segmentation of light stripe is the basis of centerline extraction and also affects the efficiency of centerline extraction. In our work, the 3D reconstruction of wood surface needs to be realized through using line-structured light sensor, and then the 3D defects need to be detected. The measured wood has a curved, rough and scattering surface. The acquired light stripe is relative complex. Several algorithms of image segmentation are studied, and the segmentation performance and operation efficiency are compared and analyzed, so as to help us find the optimum technique.