Recovery method of a contaminated multi-laser for stereo-vision-based handheld multi-laser scanner under complicated...

A robust and accurate recovery method for contaminated multi-laser stripes is promoted in this paper. First, a noise detection method is employed to locate contaminated laser stripes in an image. This process is mainly aimed at dividing an image containing multiple laser stripes into multiple images, including a single laser stripe to prepare for further analysis. Then, a DBSCAN clustering algorithm is introduced in our methods. It helps us find outliers in an interrupted laser stripe in terms of its features (stripe widths, normals and gray values). And the interrupted laser stripe in an image can thus be divided into noise and non-noise parts. Then, a centerline extraction method for laser stripes is promoted based on the improved Shepard and RBF (Radial Basis Function) interpolation to implement accurate centerline extraction in non-noise parts. Finally, the 3D information obtained through left and right cameras is fused to recover the damaged features. The experiments verify the effectiveness of the algorithm proposed in this paper.