Understanding Laser Sensor and its Calibration with Camera Youngjin Yoon / Yongseop Jeong Apr. 16, 2012 Robotics and Computer Vision Laboratory KAIST System Configuration Camera Laser Camera Laser Laser Range Finder • Device which uses a laser beam to determine the distance to objects Laser Range Finder: LMS151 • Specifications Using 905nm IR Field of view: 270º 50Hz Scanning Freq. 0.5º angular resolution – Operating Range: 0~50m – Max. range with 10% reflectivity: 18m – – – – Sample Laser Scanner Data Pattern Laser For Left camera, you need only left pattern Objectives [𝑹𝟏 |𝒕𝟏 ] 𝑲𝟏 𝑯 [𝑹𝟐 |𝒕𝟐 ] Camera 2 𝑲𝟐 Camera 1 [𝑹𝑳 |𝒕𝑳 ] Laser Scanner To estimate 𝑹𝑳 𝒕𝑳 , you should find 𝑲𝟏 first. [1,2] Find 𝑲𝟐 then estimate 𝑹𝟐 𝒕𝟐 (optional) What is Camera Calibration? • Estimating extrinsic relation between 3D world and 2D camera and the camera internal parameter 1) Intrinsic parameter K • Optical/geometrical property of a camera 2) Extrinsic Parameter [R|t] • Rotation and translation matrix Camera Calibration[1] • Estimate Homography • Intrinsic parameter estimation (K) • Extrinsic parameter estimation ([R|t]) • Stereo Calibration(Optional) Recommended Experiment Schedule • Week 1, 4/16(Mon.) – 4/22(Sun.) – Camera intrinsic parameter estimation • 𝑲𝟏 – Extrinsic parameter of Camera 1 • [𝑹𝟏 |𝒕𝟏 ] • [𝑹𝟏 |𝒕𝟏 ] is defined in world coordinates – Note that each image derives each [𝑹𝟏 |𝒕𝟏 ] – But 𝑲𝟏 is unique. Recommended Experiment Schedule • Week 2, 4/23(Mon.) – 4/29(Sun.) – Extrinsic parameter of Laser • 𝑹𝑳 𝒕𝑳 : Relative pose of the laser range finder w.r.t. the Camera 1 Recommended Experiment Schedule • Extra work – 𝑲𝟐 and 𝑹𝟐 𝒕𝟐 • Non-overwrapping stereo calibration • 𝑹𝟐 𝒕𝟐 is the relative pose w.r.t. the Camera1 • You’ll receive super-extra points Recommended Experiment Schedule • You don’t have to submit 2 reports – Just submit the result of 2 weeks until the due date. Given data • Pattern image – Pattern.zip – Includes pictures of patterns from camera 1 and 2 • Laser Data – Laser.zip Extra Points • Compute K, R and t without toolbox • Automatic feature extraction from pattern – Get each pixel point automatically • Optimization • Non-overwrapping stereo calibration – Extra work Submission • Due: 23:59:59, April 29(Sun.), 2012 • Youngjin Yoon (yjyoon@rcv.kaist.ac.kr) – Title: [RE510] ID - NAME • Example: [RE510] 20113573 – Lee Myungbak – Attachment file • Report (doc, docx, hwp or pdf) • Executable File • Source Code(compile-able) – All used libraries must be included in submitted package Submission: Report – Report should contain • The answers; 𝑲𝟏 , [𝑹𝑳 |𝒕𝑳 ], 𝑲𝟐 , [𝑹𝟐 |𝒕𝟐 ] – Justify your answer with any means » Calculating re-projection error or project laser ray on image sets • • • • Your progress to estimate them List of used toolbox(es) and libraries Referred resources(papers, books, etc.) Developing Environment – Version of OS, MSVS, MATLAB, … – Don’t attach whole source code in your report! • Detailed explanation and comment on source code is enough – Language: Korean, English Submission: Executable File • Remarks – Compiled(executable) file should be attached in a separate folder • With required dll, lib, data files and etc. • If you used MATLAB, m-files are OK – With a execution manual • If your program is not executable, you’ll lose a lot of points • Please check your program on another computer Submission: Source Code • Remarks – PLEASE add comments – Include all libraries you used • Your project should be compiled on our computer Grading Policy Item Maximum Points Understanding 40 Accuracy of the answers 20 Justifying answers(includes logical sense) 20 Completion of program 20 Extra Points Extra works(Non-overwrapping stereo calibration) 20 Without toolboxes 10 Automatic feature extraction 10 Optimization 10 Reference [1] Flexible Camera Calibration By Viewing a Plane From Unknown Orientations – Zhengyou Zhang, ICCV 1999 [2] Extrinsic Calibration of a Camera and Laser Range Finder (improves camera calibration) – Qilong Zhang, Robert Pless, IROS 2004 [3] Multiple View Geometry (2/E) – Hartley, Zisserman