Unsupervised Intrinsic Calibration of Depth Sensors via SLAM

Abstract

Abstract-We present a new, generic approach to the calibration of depth sensor intrinsics that requires only the ability to run SLAM. In particular, no specialized hardware, calibration target, or hand measurement is required. Essential to this approach is the idea that certain intrinsic parameters, identified here as myopic, govern distortions that increase with range.We demonstrate these ideas on the calibration of the popular Kinect and Xtion Pro Live RGBD sensors, which typically exhibit significant depth distortion at ranges greater than three meters. Making use of the myopic property, we show how to efficiently learn a discrete grid of 32,000 depth multipliers that resolve this distortion. Compared to the most similar unsupervised calibration work in the literature, this is a 100-fold increase in the maximum number of calibration parameters previously learned.Compared to the supervised calibration approach, the work of this paper means the difference between A) printing a poster of a checkerboard, mounting it to a rigid plane, and recording data of it from many different angles and ranges -a process that often requires two people or repeated use of a special easel -versus B) recording a few minutes of data from unmodified, natural environments. This is advantageous both for individuals who wish to calibrate their own sensors as well as for a robot that needs to calibrate automatically while in the field.

Elastic Fragments for Dense Scene Reconstruction

“…To produce the idealized data, we process the perfect synthetic depth images using the quantization model described by Konolige and Mihelich [13] and introduce sensor noise following the model of Nguyen et al [18]. To produce the simulated factorycalibrated data, we add a model of low-frequency distortion estimated on a real PrimeSense sensor using the calibration approach of Teichman et al [31].…”

“…Reconstructions produced by different approaches on the stone wall sequence from Zhou and Koltun [35]. The top row shows results with original data from the sensor, the bottom row shows results with data that was processed by the calibration approach of Teichman et al [31], which reduces low-frequency distortion. (a) Extended KinectFusion [22] is unable to produce a globally consistent reconstruction due to drift.…”

A Mixture of Manhattan Frames: Beyond the Manhattan World

“…Besides calibrating the IR camera of the depth sensors using standard monocular camera techniques [17], there is work by Herrera et al [19] on the joint calibration of RGB and depth of an RGB-D sensor. Teichman et al [32] follow a different approach for depth-camera intrinsic and distortion calibration within a SLAM framework.…”

“…This provides us with a way of calibrating a depth scanner from a single depth image of any scene exhibiting MMF structure. Compared to other techniques [17,19,32] our proposed calibration procedure is much simpler.…”

OLT: A Toolkit for Object Labeling applied to robotic RGB-D datasets

“…Also needed are extrinsic parameters to relate the position of both cameras. The intrinsic parameters largely differ among RGB-D devices, so it is recommended to calibrate them through algorithms like [25]. Those extrinsic and intrinsic parameters can be conveniently introduced into a configuration file, and this component will set them throughout all the contained observations within the dataset sequence.…”