Machine Vision for the Life Sciences

Transcription

1 Machine Vision for the Life Sciences Presented by: Niels Wartenberg June 12, 2012 Track, Trace & Control Solutions

2 Niels Wartenberg Microscan Sr. Applications Engineer, Clinical Senior Applications Engineer on Microscan's Clinical Team and regular instructor of identification technology courses, Mr. Wartenberg has been part of the Microscan Team since Prior to joining Microscan he gathered over 8 years experience implementing solutions in clinical laboratory systems.

3 Machine Vision is increasingly adopted as an effective means of automating critical processes and increasing laboratory throughput

4 Faster More Repeatable

5 Machine Vision and Auto ID Converge Microscan legacy: 30+ years in Auto ID 30+ years in Machine Vision Read bar codes, PLUS: Measure Vials Check Fill Level Verify Cap Alignment and More

6 Agenda Machine Vision Basics Definitions Uses in the Life Sciences Software Tools Image Processing Image Analysis System Configurations Smart Cameras PC-based Systems Typical Applications Identification Inspection Measurement Robotic Guidance

7 MACHINE VISION The automatic extraction of information from digital images.

8 Examples of Useful Information Presence/Absence of a Component Location/ Orientation of an Object Reading of a Human or Machine Readable Code Non-Contact Measurement of a Dimension

9 Application Examples Reading 1D symbols on microplates Reading 2D symbols on vials/racks Detecting correct orientation of slides Inspecting print quality on tubes Inspecting drops of dispensed liquid Guiding a lab robot to pick & place specimen tubes Check presence/absence of consumables (e.g. pipette tips, vials or other labware)

10 System Configurations

11 COMMUNICATION PROCESSING SENSOR LENS LIGHTING PART PART

12 Lighting Proper lighting is essential to a successful machine vision application Reveals features we want to detect/analyze Minimizes everything else Key choices Type of light Light placement with respect to the part and camera Surface geometry & texture of part are key factors in determining lighting Generally, if the feature cannot be seen, it cannot be analyzed

13 Lens Gather light & deliver to the image sensor Determine: Focal Point Field of View (FOV) Depth of Focus Lens & extension tubes

14 Lens Configurations Fixed, interchangeable lenses C-Mount standard Used with standard or smart cameras Autofocus lenses Mechanical or liquid lens autofocus Used in fully integrated imagers

15 Image Sensors Captures light and converts it to a digital image More pixels = more detail Higher resolution required when: Resolving the narrow line in a small bar code Seeing small defect on a part Making a precise dimensional measurement.3mp sensor Sensor is inside the camera 2MP sensor

16 Machine Vision Cameras Digital cameras Most modern machine vision cameras Alternative standards Camera Link Firewire (IEEE 1394) USB (2.0 and 3.0) GigE

17 GigE Vision Standard GigE Vision standard Developed by the Automated Imaging Association (AIA) Adopted by industry Advantages over other standards GigE Vision advantages High bandwidth for fast transfer of large images Uncompromised transfer up to 100 meters Standard h/w & cables for easy, low cost integration Standard h/w to connect multiple cameras to single/multiple computers Highly scalable to follow Ethernet bandwidth to 10GigE & beyond

18 Software Tools

19 Vision Processing Steps Acquire Image Image Processing Image Analysis Decision Logic Communicate Results Modify the image to make features stand out Extract features from the image Measure features and compare to specification Communicate Pass/Fail decisions and other data

20 Image Processing vs. Image Analysis Tools Image Processing Original Image -> New Image Used to make image easier to interpret or analyze Image Analysis Image -> Features Typical features include an edge, line, object, etc.

21 Image Processing Tool Examples Image arithmetic Image warping Binary & grayscale morphology

22 Image Warping Often used prior to OCR (Optical Character Recognition) Rotate text viewed at an angle Unwrap text printed on an arc or a circle

23 Morphology Transforms the image to make certain features stand out Use to expand, separate, merge, clean Does not extract features Erode black pixels: Create separation and then count Dilate white pixels: Increase Data Matrix cell size

24 Image Analysis Tool Examples The Blob Tool Edge Detection Pattern Matching 1D & 2D Symbols OCR & OCV Dynamic ROI Location Measurement Resolution

25 The Blob Tool A blob is a group of connected pixels within a size range similar color (shade of gray) differ from surrounding area Typical applications Count number of parts Locate position of a part Measure size of a part Compare to a tolerance Count: Verify that all wires are installed Measure: Check size to ensure parts are not broken

26 Edge Detection Edge tools scan an image along a user-specified direction Detect transitions between two regions of different intensity Fit a line, circle or ellipse to the edge data Applications Detect or locate an object Measure a distance Locate a corner Measure an angle

27 Vector Edge Detection Edges can be detected along user defined directions (vectors)

28 Edge Tool Usage Locate: Use two Edge tools to find a corner Measure: Measure fill level of a container or detect cap tightness Locate: Check label placement

29 Finding Patterns in an Image Normalized correlation based template matching Scans template across image and identifies best match Affected if part of what is in the template is missing from the image being analyzed Geometric edge pattern matching Matches patterns of edges in the image and the template Is not affected if part of the template is missing in the image or if the polarity of the image is reversed

30 Pattern Matching Pattern matching tools learn the outline of a part of pattern Scans across image and identifies best match Locates pattern with subpixel accuracy Intellifind Tool Example

31 1D & 2D Symbols Linear (1D) Bar Code Symbols Limited data storage Height provides redundancy Requires higher contrast 2D Symbols (ie, Data Matrix) Data encoded in both height & width Readable 360 Contrast as low as 20%

32 OCR - Optical Character Recognition Decodes human readable text Can handle dot matrix & dot peen printing Noisy backgrounds Uneven lighting Trainable Neural Network based Character addition/deletion Tolerates scale changes

33 OCV Optical Character Verification Typical Application Checking correctness and legibility of a printed label or text Detects printing defects

34 OCR vs. OCV Terms often used incorrectly NOT the same OCR: Optical Character Recognition OCV: Optical Character Verification OCR an automatic identification tool Intended to decode human readable information OCV a print quality inspection tool Intended to flag & reject poor quality text

35 Dynamic ROI Location Relocating regions of interest (ROI) to compensate for part position and rotation ROIs Positioned ROIs After Part Motion ROIs Corrected For Part Movement

36 Nominal Measurement Resolution Example: Example: Paper Clip = 1.0 " Image width = 1280 Pixels Resolution = 1.0 " /1280 Pixels =.0008 " /Pixel Car = 13.5 Image width = 1280 Pixels Resolution = 13.5 / 1280 Pixels 162 /1280 Pixels =.127" /Pixel

37 Actual Measurement Resolution Depends on: Sensor resolution Field of View(FOV) Sub-pixel capability of vision processing tools

38 Typical Applications

39 Automatic Identification Applications Specimen ID Reagent ID Tube carrier ID Microtiter plate ID Microtiter vial ID

40 Cap Color Detection

41 Tube & Cap Inspection Application requirements Rack location Bar code reading Tube/cap inspection Solution Camera Custom GUI

42 Absence/Presence with compact Smart Camera

43 Vision Guided Motion Applications Application requirements Determine object location & orientation Calibration in real world units Robust & accurate pattern matching Application examples Robotic tube pick & place Colony picking

44 Machine Vision Automate critical processes Increase throughput Achieve verifiable process repeatability

45 Thank You If you have questions regarding this webinar or topic, please an to For further information about machine vision, visit our website at