Automatic background subtraction algorithm for 3D object by using connected-component labeling algorithm

Abstract

This paper proposes an automatic background subtraction algorithm for 3D object. The objective is to apply the algorithm to quickly pick out the 3D object from background when the position is roughly fixed. The method consists of 3 main steps as follows. First step creates a box for hold all 3D data. Second step projects the whole 3D points on a horizontal plane which is align on the middle of the box, then converts the 3D projected points into 2D black and white image. Final step applies connected component labeling algorithm for classifying black pixels, in order to find the group which satisfies the object constraints and to render associated 3D points. The experimental results shown that the whole process takes 1.6 sec on average, with 90 % accuracy.