Algorithms for Fitting Cylindrical Objects to Sparse Range Point Clouds for Rapid Workspace Modeling

Abstract

Current methods for construction site modeling employ large, expensivelaser range scanners that produce dense range point clouds of a scenefrom different perspectives. While useful for many purposes, thisapproach is not feasible for real-time applications, which wouldenable automated obstacle avoidance and semi-automated equipmentcontrol, and could improve both safety and productivity significantly.This paper presents human-assisted rapid environmental modeling algorithmsfor construction, and focuses on cylindrical object fitting algorithms.The presented algorithms address construction site material of cylindricalshape. Experiments were conducted to determine: (1) the effect ofthe ratio of length to diameter of the cylinder to the accuracy ofthe results, (2) the effect of the angle of view to the accuracyof the results, (3) the minimum number of scanned points requiredto give adequate modeling accuracy for cylinders of various lengthto diameter ratios. The results indicate that the proposed algorithmscan model geometric primitives used in a construction site rapidlyand with sufficient accuracy for automated obstacle avoidance andequipment control functions.