Hand-eye calibration for 2D laser profile scanners using straight edges of common objects

Abstract

This paper presents a method for calibrating a 2D profile laser scanner mounted on an industrial articulated robot; a task also known as the hand-eye calibration problem. The challenge of recovering the transformation matrix, from the robot's flange coordinate system to the scanner's coordinate system, lies in the lack of sufficient 3D information, as only 2D data is available. The task is typically performed using precision calibration specimens such as spheres, disks, and planes or using additional external devices such as cameras and 3D sensors. Here, we present an approach based on detecting straight edges found in common objects. Points extracted from the same edge, under various robot poses, are used to solve the calibration problem using a two-phase least-squares strategy, where rotation is recovered first, followed by translation. The process is semi-autonomous, requires minimal laborious and error-prone manual operations; its setup effort is small, because common objects can be used instead of costly precision gauges or external devices; it does not require large number of samples and it is simple to reason about, implement and compute.