Mobile Currency Reader March 12, 2008
Outline <ul><li>Introduction </li></ul><ul><li>Data flow </li></ul><ul><li>Platforms </li></ul><ul><li>Bill detection </li...
Introduction <ul><li>Mobile currency reader </li></ul><ul><li>Targeted to assist the visually impaired to properly identif...
Data flow
Data flow - Picture taken <ul><li>JME Multimedia API (MMAPI) provides a camera interface. </li></ul><ul><li>Returns an ima...
Data flow – Bill Auth <ul><li>The necessity to identify a bill is important. </li></ul><ul><li>UV rays can show the ID lin...
Data flow – Bill extraction <ul><li>The image taken will be passed into the image processing function which returns a fixe...
Data flow <ul><li>Image comparison takes in the M x N image, and runs it through the neural network to determine the value...
Data flow <ul><li>Depending on the result of the Image comparison block, the program will output a .wave file to the user ...
Platform and Libraries <ul><li>Targeted platform is the Java Micro Edition (original plan to also support Android has been...
Bill Extraction <ul><li>Aim: </li></ul><ul><ul><li>Extract the M x N image of bill and pass it for Image Comparison </li><...
Bill Extraction – Edge Detection <ul><li>Canny Edge Detection Algorithm  more .. </li></ul>
Bill Extraction - Steps <ul><li>Identify boundaries </li></ul><ul><li>Find corresponding pixels </li></ul><ul><li>from ori...
Value identication
Schedule <ul><li>3/8 – 3/12 – PDR (3/12) </li></ul><ul><li>3/13-3/19 – (spring break 3/17-3/23) </li></ul><ul><ul><li>Demo...
Risks and Assumptions <ul><li>Processing power of mobile phone relative </li></ul><ul><li>Ability to implement NN on JME <...
Upcoming SlideShare
Loading in …5
×

Mobile Currency Reader March 12, 2008

465 views

Published on

0 Comments
0 Likes
Statistics
Notes
  • Be the first to comment

  • Be the first to like this

No Downloads
Views
Total views
465
On SlideShare
0
From Embeds
0
Number of Embeds
7
Actions
Shares
0
Downloads
6
Comments
0
Likes
0
Embeds 0
No embeds

No notes for slide

Mobile Currency Reader March 12, 2008

  1. 1. Mobile Currency Reader March 12, 2008
  2. 2. Outline <ul><li>Introduction </li></ul><ul><li>Data flow </li></ul><ul><li>Platforms </li></ul><ul><li>Bill detection </li></ul><ul><li>Value identification </li></ul><ul><li>Bill authentication </li></ul><ul><li>Schedule </li></ul>
  3. 3. Introduction <ul><li>Mobile currency reader </li></ul><ul><li>Targeted to assist the visually impaired to properly identify US paper currency </li></ul><ul><li>Developing software that will run on a device that supports JME and has a camera (ie. Cell phone) </li></ul>
  4. 4. Data flow
  5. 5. Data flow - Picture taken <ul><li>JME Multimedia API (MMAPI) provides a camera interface. </li></ul><ul><li>Returns an image object </li></ul>Output Result to User Bill authentication Picture Taken Image Comparison Image Processing Bill extraction
  6. 6. Data flow – Bill Auth <ul><li>The necessity to identify a bill is important. </li></ul><ul><li>UV rays can show the ID line in a bill, but normal cameras cannot detect this. </li></ul><ul><li>(We are skipping this step and assuming that the bill we get is real.) </li></ul><ul><li>Input: </li></ul><ul><ul><li>Unprocessed camera image </li></ul></ul><ul><li>Output: </li></ul><ul><ul><li>Same camera image, but verified to have a bill </li></ul></ul>Output Result to User Bill authentication Picture Taken Image Comparison Image Processing Bill extraction
  7. 7. Data flow – Bill extraction <ul><li>The image taken will be passed into the image processing function which returns a fixed M x N image of the bill that is passed into Image Comparison </li></ul><ul><li>Input: </li></ul><ul><ul><li>Unprocessed camera image </li></ul></ul><ul><li>Output: </li></ul><ul><ul><li>M x N image of a bill </li></ul></ul>Output Result to User Bill authentication Picture Taken Image Comparison Image Processing Bill extraction
  8. 8. Data flow <ul><li>Image comparison takes in the M x N image, and runs it through the neural network to determine the value of the bill or if a result is inconclusive, returns value to the output block </li></ul><ul><li>Input: </li></ul><ul><ul><li>M x N Image of a bill </li></ul></ul><ul><li>Output: </li></ul><ul><ul><li>Value of bill, or error code </li></ul></ul>Output Result to User Bill authentication Picture Taken Image Comparison Image Processing Bill extraction
  9. 9. Data flow <ul><li>Depending on the result of the Image comparison block, the program will output a .wave file to the user with the denomination of the bill, or the inconclusive result </li></ul><ul><li>Input: </li></ul><ul><ul><li>Value of bill or error code </li></ul></ul><ul><li>Output: </li></ul><ul><ul><li>Plays wav file to the user with resulting value </li></ul></ul>Output Result to User Bill authentication Picture Taken Image Comparison Image Processing Bill extraction
  10. 10. Platform and Libraries <ul><li>Targeted platform is the Java Micro Edition (original plan to also support Android has been dropped due to concerns of lack of library support for the platform) </li></ul><ul><li>Initial tests of the Neural Network will be done in matlab and then translated into JME </li></ul><ul><li>WEKA (Waikato Environment for Knowledge Analysis) library for support on machine learning </li></ul>
  11. 11. Bill Extraction <ul><li>Aim: </li></ul><ul><ul><li>Extract the M x N image of bill and pass it for Image Comparison </li></ul></ul><ul><li>Input: </li></ul><ul><ul><li>Unprocessed camera image </li></ul></ul><ul><li>Output: </li></ul><ul><ul><li>M x N image of a bill </li></ul></ul>
  12. 12. Bill Extraction – Edge Detection <ul><li>Canny Edge Detection Algorithm more .. </li></ul>
  13. 13. Bill Extraction - Steps <ul><li>Identify boundaries </li></ul><ul><li>Find corresponding pixels </li></ul><ul><li>from original image </li></ul><ul><li>Rotate the image </li></ul><ul><li>Pass the image </li></ul><ul><li>for Image Comparison </li></ul>
  14. 14. Value identication
  15. 15. Schedule <ul><li>3/8 – 3/12 – PDR (3/12) </li></ul><ul><li>3/13-3/19 – (spring break 3/17-3/23) </li></ul><ul><ul><li>Demonstrate NN in matlab </li></ul></ul><ul><ul><li>Investigate WEKA compatibility </li></ul></ul><ul><li>3/20-3/26 – Status update </li></ul><ul><ul><li>Demonstrate NN on JME </li></ul></ul><ul><ul><li>Demonstrate basic edge detection with JME (on a solid colored surface) </li></ul></ul><ul><li>3/27-4/2 – CDR (4/2) </li></ul><ul><ul><li>Combine products </li></ul></ul><ul><li>4/3-4/9 – </li></ul><ul><li>4/10-4/16 – Status update </li></ul><ul><li>4/17-4/23 </li></ul><ul><li>4/24-4/30 – Final meeting </li></ul>
  16. 16. Risks and Assumptions <ul><li>Processing power of mobile phone relative </li></ul><ul><li>Ability to implement NN on JME </li></ul><ul><ul><li>Backup to NN is to use image registration </li></ul></ul><ul><ul><li>http://en.wikipedia.org/wiki/Image_registration </li></ul></ul><ul><li>We will start with the assumption that our bill is complete and does not have pen markings or rips and smudges </li></ul><ul><li>Once we have a working model of our bill detector and value identification we will work on more advanced cases </li></ul><ul><li>We assume that the user will take a picture of a reasonably sized bill. </li></ul>

×