Hochperformante Inline-3D-Messung

Similar documents
Machine Vision in Austria

Depth estimation using light fields and photometric stereo with a multi-line-scan framework

Lecture 19: Depth Cameras. Kayvon Fatahalian CMU : Graphics and Imaging Architectures (Fall 2011)

A NOVEL VISION SYSTEM-ON-CHIP FOR EMBEDDED IMAGE ACQUISITION AND PROCESSING

Ricoh's Machine Vision: A Window on the Future

Multi-aperture camera module with 720presolution

e2v Launches New Onyx 1.3M for Premium Performance in Low Light Conditions

THE VISIONLAB TEAM engineers - 1 physicist. Feasibility study and prototyping Hardware benchmarking Open and closed source libraries

Hyperspectral imaging (HSI) goes embedded All rights reserved Max Larin, 1

Imaging with hyperspectral sensors: the right design for your application

SOLAR CELL INSPECTION WITH RAPTOR PHOTONICS OWL (SWIR) AND FALCON (EMCCD)

brief history of photography foveon X3 imager technology description

NELA Brüder Neumeister GmbH

Image sensor combining the best of different worlds

Image Sensor and Camera Technology November 2016 in Stuttgart

A NEW NEUROMORPHIC STRATEGY FOR THE FUTURE OF VISION FOR MACHINES June Xavier Lagorce Head of Computer Vision & Systems

It Takes Two to Tango

Practical Image and Video Processing Using MATLAB

Fraunhofer Institute for High frequency physics and radar techniques FHR. Unsere Kernkompetenzen

WHITE PAPER. Sensor Comparison: Are All IMXs Equal? Contents. 1. The sensors in the Pregius series

Model-Based Design for Sensor Systems

CMOS Today & Tomorrow

Light-Field Database Creation and Depth Estimation

Introduction. Lighting

sensors & systems Imagine future imaging... Leti, technology research institute Contact:

Remote Sensing Platforms

High Resolution Spectral Video Capture & Computational Photography Xun Cao ( 曹汛 )

Evaluation of laser-based active thermography for the inspection of optoelectronic devices

Background. Computer Vision & Digital Image Processing. Improved Bartlane transmitted image. Example Bartlane transmitted image

Active Stereo Vision. COMP 4102A Winter 2014 Gerhard Roth Version 1

Single Camera Catadioptric Stereo System

FastPass A Harmonized Modular Reference System for Automated Border Crossing (ABC)

INTRODUCTION TO VISION SENSORS The Case for Automation with Machine Vision. AUTOMATION a division of HTE Technologies

How does prism technology help to achieve superior color image quality?

The chemical camera for your microscope

STEINBICHLER 25 YEARS OF INSPIRING INNOVATION ABIS OPTICAL SURFACE INSPECTION

Generalized Assorted Camera Arrays: Robust Cross-channel Registration and Applications Jason Holloway, Kaushik Mitra, Sanjeev Koppal, Ashok

SMARTSCAN Smart Pushbroom Imaging System for Shaky Space Platforms

SUPRA Optix 3D Optical Profiler

Vision with Precision Webinar Series Augmented & Virtual Reality Aaron Behman, Xilinx Mark Beccue, Tractica. Copyright 2016 Xilinx

Improved sensitivity high-definition interline CCD using the KODAK TRUESENSE Color Filter Pattern

23270: AUGMENTED REALITY FOR NAVIGATION AND INFORMATIONAL ADAS. Sergii Bykov Technical Lead Machine Learning 12 Oct 2017

Eight Tips for Optimal Machine Vision Lighting

Polaris Sensor Technologies, Inc. SMALLEST THERMAL POLARIMETER

Digital transformation of education & training in photonical measurement engineering & quality assurance (PMQ)

Combined expertise in the field of optical 3D gauging technology ensures market advantage

Institute of Computer Technology

Hyper-spectral, UHD imaging NANO-SAT formations or HAPS to detect, identify, geolocate and track; CBRN gases, fuel vapors and other substances

E90 Project Proposal. 6 December 2006 Paul Azunre Thomas Murray David Wright

CMOS Image Sensors in Cell Phones, Cars and Beyond. Patrick Feng General manager BYD Microelectronics October 8, 2013

Sensor system of a small biped entertainment robot

The EDA SUM Project. Surveillance in an Urban environment using Mobile sensors. 2012, September 13 th - FMV SENSORS SYMPOSIUM 2012

VisionMap Sensors and Processing Roadmap

PRODUCT BROCHURE PROFILER R. Tactile sensor for roughness measurement on Leitz CMMs

Dr. Ralf Freiberger. TEMA GmbH / Mühlbauer Group

Embedded Sensors. We can offer you complete solutions for intelligent integrated sensor systems.

Colour image watermarking in real life

RealNano & ACINTECH Projektbeispiele für Nanotechnologie in der Mikroelektronik

Advances in slabs defects inspection with Conoscopic Holography. Ignacio Alvarez, J.M. Enguita (UniOvi) J. Marina, R.

A NOVEL HIGH SPEED, HIGH RESOLUTION, ULTRASOUND IMAGING SYSTEM

Open Access The Application of Digital Image Processing Method in Range Finding by Camera

TECHNICAL DATA OPTIV CLASSIC 432

Parallel Mode Confocal System for Wafer Bump Inspection

Fiber-optic temperature measurement solves HV challenges in e-mobility Tech Article

Optotop. 3D Topography. Roughness (Ra opt, Rq opt, and Rz opt) Height Distribution. Porosity Distribution. Effective Contact Area

Machine Vision Basics

Recent Developments in Multifunctional Integration. Stephan Guttowski, Head of Technology Park»Heterointegration«, Fraunhofer FMD

SEE MORE, SMARTER. We design the most advanced vision systems to bring humanity to any device.

CMP for More Than Moore

International Journal of Innovative Research in Engineering Science and Technology APRIL 2018 ISSN X

Sensory Fusion for Image

Optical basics for machine vision systems. Lars Fermum Chief instructor STEMMER IMAGING GmbH

Autoliv Night Vision System Safety Application Automotive IR Camera

Xenon-Zirconia 3.3/92

White Paper Focusing more on the forest, and less on the trees

OLYMPUS Digital Cameras for Materials Science Applications: Get the Best out of Your Microscope

Congress Best Paper Award

Multispectral imaging and image processing

Remote Sensing Platforms

Notes and Thoughts By Tony Giovaniello, President, Shasta EDC

Development of intelligent systems

Hardware for High Energy Applications 30 October 2009

INNOVATIVE CAMERA CHARACTERIZATION BASED ON LED LIGHT SOURCE

ALMALENCE SUPER SENSOR. A software component with an effect of increasing the pixel size and number of pixels in the sensor

Optimizing throughput with Machine Vision Lighting. Whitepaper

Low Cost Earth Sensor based on Oxygen Airglow

The Challenge of. SYSTEMS FOR MACHINE VISIONComplexity. Pioneering vision.

Applying Automated Optical Inspection Ben Dawson, DALSA Coreco Inc., ipd Group (987)

Changyin Zhou. Ph.D, Computer Science, Columbia University Oct 2012

A Review of Related Work on Machine Learning in Semiconductor Manufacturing and Assembly Lines

THE ULTIMATE DOCUMENT EXAMINATION SYSTEM STATE-OF-THE-ART SPECTRAL ANALYSIS FORENSIC LABS SECURITY PRINTERS IMMIGRATION AUTHORITIES

Macro Varon 4.5/85. Key features. Applications. Web and surface inspections

Digital Photographic Imaging Using MOEMS

High-Speed 3D Sensor with Micrometer Resolution Ready for the Production Floor

Applications for cameras with CMOS-, CCD- and InGaAssensors. Jürgen Bretschneider AVT, 2014

Edge-Raggedness Evaluation Using Slanted-Edge Analysis

HIGH RESOLUTION COLOR IMAGERY FOR ORTHOMAPS AND REMOTE SENSING. Author: Peter Fricker Director Product Management Image Sensors

Time of Flight Capture

LINCE5M 5.2 MEGAPIXELS, 1 INCH, 250FPS, GLOBAL SHUTTER CMOS IMAGE SENSOR. anafocus.com

Distinctive Feature: International well situated

Transcription:

Hochperformante Inline-3D-Messung mittels Lichtfeld Dipl.-Ing. Dorothea Heiss Deputy Head of Business Unit High Performance Image Processing Digital Safety & Security Department AIT Austrian Institute of Technology

Bundesministerium für Verkehr, Innovation und Technologie 50,46% Industriellenvereinigung 49,54% 1300+ employees Budget: 120 Mio Future Networks and Services Intelligent Vision Systems Highly Reliable SW and Systems > 100 experts 2

Intelligent Vision Systems (IVS) for Process Automation & Inspection (Industry 4.0) Autonomous & Assistive Systems Surveillance & Protection Worldwide fastest vision sensor technology Robust and flexible 3D vision technology 3

Intelligent Vision Systems (IVS) for Process Automation & Inspection (Industry 4.0) Autonomous & Assistive Systems Surveillance & Protection 4

Process Automation & Inspection Example: Banknote Inspection World technology leader 100% inline inspection of newly printed banknotes KEY FACTS Speed 11 m/s (44 banknotes/s) Resolution 0,1 mm/pixel Color Multispectral inspection (color, infrared & UV) Multiview International 100% inspection of front-, backside and transmission Worldwide standard for banknote printing shops CURRENT RESEARCH FIELDS True color Hyperspectral imaging for true color inspection New security features Inspection of holograms and optical variable features 3D information for tactile elements Increasing optical resolution for micro text inspection 5

Projects and Solutions Electronics Infrastructure Print & Paper Metal Process Automation & Inspection world leading linescan PCB street surface & 3D PCB assembly failure classification wafer rail surface all in focus coin texture & 3D coin identification micro pores tube cracks banknotes packaging flex inspect tactile features OVI quality inspection 3D inspection classification photometric stereo light field high speed stereo line scan deep learning computational imaging 6

Computational Imaging Several images of an object under different conditions Redundancy and computation are used to enhance image quality (depth of field, resolution, contrast, sharpness, dynamic range, signal to noise ratio, ) and/or generate useful information, especially depth / 3D Examples: in-camera computation of digital panoramas, high-dynamicrange images, light field cameras, photometric stereo, Computing power is getting faster and cheaper far quicker than precise optical devices, stronger illumination etc. (Moore s law) 7

Photometric Stereo Object is acquired under different illumination angles Applications Surface inspection 3D reconstruction (surface normal, depth map) Material classification (Bidirectional reflectance distribution function) 8

Surface normals Depth maps Example: Coins (Photometric Stereo) 9

10

Photometric Stereo Advantages Sensitive to very fine details Information about glossiness / material classification Also suited for glossy surfaces No surface texture required Disadvantages Metrically not correct Globally not correct No measurement of steep edges 11

Light Field Object is acquired under different observer / camera angles Describes reflected light in each point of space (x, y) in each direction (u, v) State of the art CMOS chip with micro-lens array (e.g. Lytro) Multi-camera array Applications Computational refocusing / extended depth-of-field Changing viewpoint 3D reconstruction 12

Advantages of Light Field compared to Stereo Robustness: several acquisitions contain more information than two Less sensitive to specular reflections Less occlusion because of smaller stereo baseline Flexible number of views and viewing angles Additional methods for depth estimation possible, e.g. Multi-view block matching Depth from refocus Structure tensor in EPI domain Not only depth estimation, but also All-in-focus, Noise reduction, Deblurring, Super resolution, High-dynamic range, 13

Example: Coins (Light Field) 14

15

Light Field Advantages Metrically correct Globally correct Measurement of steep edges All in focus / enhanced depth of field Disadvantages Not sensitive to fine details Needs surface structure Not suited for glossy surfaces Depth map is incomplete 16

Light Field Photometric Stereo Advantages Metrically correct Globally correct Measurement of steep edges All in focus Disadvantages Not sensitive to fine details Needs surface structure Not suited for glossy surfaces Depth map is incomplete Advantages Sensitive to very fine details Information about glossiness / material classification Also suited for glossy surfaces No surface texture required Disadvantages Metrically not correct Globally not correct No measurement of steep edges Combination would be ideal 17

Requirements of Industrial Inspection Combination of Light Field and Photometric Stereo Good price/performance ratio Small installation space High speed Reproducibility Rough environment (dirt, heat) Robust against flutter, vibrations Minimal down-time Easy configuration Dome plus multi-camera array Costly Bulky Many components Area scan - not inline Many acquisitions Computationally expensive complicated Real-time fusion not solved Combination not ideal for industrial inline applications? 18

AIT Multi-line-scan Light-field Camera Camera 1 Line 1 Camera Line 2 2 Line 3 Camera 3 Light Field from Motion Fast CMOS area scan chip Multi-line scan acquisition 3D light field from motion Transport direction Pros Single sensor / single lens Flexible number of views Simple adjustment Easy data transfer Compact form factor Lower costs Inline Fast: 10.000 Hz 19

Light Field Acquisition Line Scan 20

Epipolar Planes Light field method: the more an epiolar plane is slanted, the higher the disparity, the nearer the object point to camera, the higher the object point 21

AIT Multi-line-scan Light-field Camera - Revisited Different observer angle light field AND different illumination angle photometric stereo in each object point (at a given point of time) Fusion is inherent in setup Only 1 camera, 1 optics, 1 illumination Angles interdepend! None of the known methods applicable 22

Epipolar Planes Overexposure Is it a bug or a feature New photometric method: finding line (=view) with the intensity peak Light field method: determining depth from slant New fused method: finding view with the intensity peak along the epipolar line Patents pending 23

Example: Euro Depth maps (derived from slant) light field alone Surface normals (derived from peak) photometric stereo alone Surface normals (derived from peak, but considering the slant) photometric stereo with light field 24

Example: Canadian Silver Dollar Depth maps (derived from slant) light field alone Surface normals (derived from peak) photometric stereo alone Surface normals (derived from peak, but considering the slant) photometric stereo with light field 25

Comparison: Photometric Stereo (Inline) and Light Field with Photometric Stereo (Inline) 26

Comparison: Photometric Stereo (Dome) and Light Field with Photometric Stereo (Inline) 27

xposure fastest line scan camera computational imaging High-Performance Vision Fusion of technologies and expertise deep learning High-Performance Vision Optimized Solutions for Real-time Embedded Vision Systems embedded parallel hardware 28

xposure - Worldwide Fastest Line Scan Sensor 600.000 lines per second monochrome, or 200.000 lines per second RGB color, or 50.000 light field acquisitions per second Digital output 1,2 Giga-Pixel / second (10 bit) Flexibility because of random access to 60 pixel lines Developed with Fraunhofer IMS 29

xposure - Inspiring Speed and Resolution Example m/s km/h Resolution @ 600kHz mm Earth observation satellite 7200 25920 12,0 Moon around earth / bullet 900 3240 1,5 Aircraft 300 1080 0,5 ICE train 90 324 0,15 Print inspection 10 36 0,017 Metal surface inspection 1,8 6,5 0,003 enabling a new dimension of high speed image processing applications 30

xposure Camera Launch at Vision fair 2016 in Stuttgart Focused on key requirements for quality inspection High speed High resolution High signal to noise ratio Small: 8x8x8 cm FPGA for control and preprocessing Sensor manufactured by Fraunhofer IMS Automotive certified Long term availability Made in Germany 31

AIT Multi-line Scan Light Field and Photometric Stereo Setup Advantages High speed acquisition and and fast algorithms (real-time) Small, few components, good price / performance ratio Inline (line-scan approach) Fine surface details with metrically and globally correct depth map Results robust to material properties and lighting conditions All in focus, enhanced image quality Flexible: accuracy vs. speed can be chosen dynamically Technology protected by several patents 32

AIT Austrian Institute of Technology your ingenious partner Dipl.-Ing. DOROTHEA HEISS Deputy Head of Business Unit High Performance Image Processing Digital Safety & Security Department AIT Austrian Institute of Technology GmbH 2444 Seibersdorf Austria T+43(0) 664 8251177 dorothea.heiss@ait.ac.at