MVPCC-Net: Multi-View Based Point Cloud Completion Network for MLS data

In this paper, we introduce a novel multi view-based method for completing high-resolution 3D point clouds of partial object shapes obtained by mobile laser scanning (MLS) platforms. Our approach estimates both the geometry and color cues missing or incomplete segments, projecting input cloud multiple virtual cameras, performing 2D inpainting in image domains different views. contrast to existing state-of-the-art methods, our can generate consisting variable number points, depending on detailedness measurement, which property highly facilitates efficient processing MLS data with inhomogeneous density. For training quantitative evaluation proposed method, provide new dataset that consists synthetic four street objects accurate ground truth, real measurements partially fully scanned vehicles. The qualitative experiments provided demonstrate surpasses approaches reconstructing local fine geometric structures as well estimating overall shape pattern objects.