A Prospective Algorithm for Real Plane Identification from 3D Point Clouds and 2D Edges

We present a prospective algorithm to understand test scene for line base recognition and 3D pose estimation. Given 3D line segments, first, we set up reference plane using principle component analysis (PCA) for each line pair. Then, we measure the normal distance and line orientation for the remains of 3D segments to the reference plane and define coplanarity. And we initiate visibility test to prune out ambiguous planes amongst reference planes. After we finish visibility test, we patching these reference planes and define them as real plane candidates using plane sweep algorithm and find objects in the scene. We apply this algorithm to real images, and the results are found useful to identify table or door especially adjacent to walls.