Point cloud modeling based on user-provided seed

摘要

In one embodiment, a semi-automated technique is used to reconstruct a vector-based model from a point cloud that represents a physical subject. At least a portion of the point cloud is displayed. A selection of a seed of points that correspond to a particular physical feature of the physical subject is received from a user. The selection is received with a flashlight virtual tool having a field of view. An automatic detection of a geometric primitive is performed based on the seed of points. The automatic detection uses an iterative algorithm that detects an initial geometric primitive based on the seed of points, uses the initial geometric primitive to determine an expanded set of points, and detects a refined geometric primitive based on the expanded set of points. A vector-based model is produced using the detected geometric primitive.