Inline Computational Imaging: Single Sensor Technology for Simultaneous 2D/3D High Definition Inline Inspection

Transcription

1 Inline Computational Imaging: Single Sensor Technology for Simultaneous 2D/3D High Definition Inline Inspection Svorad Štolc et al. AIT Austrian Institute of Technology GmbH Center for Vision, Automation & Control Vienna, Austria

2 AIT Inline Computational Imaging: Single Sensor Technology for Simultaneous 2D/3D Inline Inspection ICI Sensor System Single Sensor Technology for Simultaneous Light Field and Photometric Stereo Capture ICI Software Modules Generic Computational Imaging Library for 2D/3D Tasks + 2

3 AIT Inline Computational Imaging: Working Principle Camera Illumination Transport stage Inspected object 3

4 AIT Inline Computational Imaging: Working Principle Camera captures multiple observation angles simultaneously Object travels in front of the camera on the transport stage 4

5 AIT Inline Computational Imaging: Working Principle Multi-line scan sensor reads out only few sensor lines in high speed Light rays pass through a camera lens 5

6 AIT Inline Computational Imaging: Working Principle Every view is associated with a different viewing perspective Over time, each active sensor line produces an individual view 6

7 AIT Inline Computational Imaging: Working Principle Advanced computational imaging algorithms are used to compare and match the content of different views 7

8 AIT Inline Computational Imaging: Working Principle Multiple views of the light field allow for reconstruction of an accurate 3D model 8

9 AIT Inline Computational Imaging: Working Principle Photometric stereo increases sensitivity to fine surface details, even in texture-less areas 9

10 AIT Inline Computational Imaging: Working Principle Pixel-precise de-noised 2D texture can be extracted from multiple views And much more! Let s take a closer look 10

11 AIT Inline Computational Imaging: Simultaneous Capture of Multiple Viewing Perspectives Pin grid array Due to the stereo parallax, pins are moving in the opposite direction as the background Stereo disparity 11

12 AIT Inline Computational Imaging: 3D Sensing Pin grid array Precise 3D model using AIT approach: Fusion of Light Field and Photometric Stereo 3D model using state-of-the-art stereo algorithm 12

13 AIT Inline Computational Imaging: Robust to Material Reflectance Properties Color image Scene containing various material types Precise 3D model using AIT approach 3D model using state-of-the-art stereo algorithm 13

14 AIT Inline Computational Imaging: Sharper Images with All-in-Focus / 3D-TDI PCB board with protruding components All-in-focus image with reduced noise State-of-the-art image with motion blur 14

15 AIT Inline Computational Imaging: Computational Bright Field and Gloss Suppression PCB board with glossy components Computational bright field Computational gloss suppression 15

16 AIT Inline Computational Imaging: Computational Bright Field and Gloss Suppression Metal part with drilling and brushing Computational bright field Computational gloss suppression 16

17 AIT Inline Computational Imaging: Computational Bright Field and Gloss Suppression Circulation coin Computational gloss suppression Computational bright field 17

18 AIT Inline Computational Imaging: Fine Surface Structures via Photometric Stereo Circulation coin 2.5D surface structures 2D color image 18

19 AIT Inline Computational Imaging: Fine Surface Structures via Photometric Stereo Product packaging with braille script 2D color image 2.5D surface structures 19

20 AIT Inline Computational Imaging: Security Print and Optical Variable Devices Color image Banknote 10 Fine surface structures of intaglio print OVD / hologram detection 20

21 ICI Sensor System: Scalable Optical Configurations Configuration Examples Standard Scale Micro Scale Large Scale Optics 45mm f/4 Modified 10x industrial inspection microscope 20mm f/2.8 with NA=0.28 Working distance 128 mm 34 mm 570 mm Field of view 46 mm 1 mm 462 mm Lateral resolution 20 µm/pixel 1 µm/pixel 200 µm/pixel Depth resolution *) 20 µm 1 µm 100 µm Depth range 10 mm 300 µm 200 mm Min. / typ. viewing angles 3 / 11 3 / 30 3 / 11 Typical acquisition speed 100 mm/s 10 mm/s 1 m/s *) using Light Field and Photometric Stereo 21

22 ICI Software Modules: Generic Computational Imaging Library for 2D/3D Tasks Acquisition Stereo matching Depth refinement ICI Sensor System Camera Illumination Transport stage Light field data acquisition Camera, illumination, and transport control Rectification Camera lens rectification Shading correction Image processing High speed multiview / light field stereo matching Feature extraction Extraction of robust image features Extraction of local surface features Fast discrete and continuous depth regularization Fusion with additional depth cues Texture enhance All-in-focus image generation Computational bright field and dark field images OVD detection Detailed 2D/3D data Refined 3D depth model (depth map, point cloud) Depth measurement confidence Enhanced 2D texture images (all-in-focus, bright field / gloss suppression, OVD detection) Fine 2.5D surface map 22

23 AIT Inline Computational Imaging: Industrial Use Cases Electronic parts Metal parts Product packaging Coins Printed circuit boards Measurement Security print / OVD 2381μm 1064μm 23

24 AIT Inline Computational Imaging: Conclusions 24

25 Thank you! Svorad Štolc