Robust laser stripe extraction for three-dimensional reconstruction based on a cross-structured light sensor

Cross-structured light sensor architecture for 3D reconstruction was established. An improved center of mass method was proposed for laser stripe extraction. For each initial laser stripe center point, the center of mass method was performed along the normal direction that was calculated using the Hessian matrix. The normal directions can be used to divide the laser stripe center points into two categories. Laser stripe extraction experiments showed that the proposed method is fast and robust. 3D reconstruction of a cylinder was used to analyze reconstruction accuracy, with relative accuracy of less than 0.15 mm. 3D reconstruction of a shoe last showed that cross-structured light sensors can obtain more abundant information than single-structured light sensors.