1. Germ Reader (08 June, 2013) Jaehwan Kim (edenkim519@gmail.com)
Bangyong Song
Adviser: Prof. Dong Seon Cheng

2. Contents
Introduction
Methodology
Validation
Implementation
Limitation

3. Introduction
In the Dept. of Food Engineering, research and analysis of the bacteria have been proposed.

4. Introduction

5. Introduction They count bacteria manually.
-> inaccurate and time consuming

6. Introduction
1min * 60 times = 1 hour
-> it takes a lot of time!!

7. Introduction
If there is a system to count germs automatically and correctly ?
-> it will be much help to the research.

8. Introduction
dilluted 10^6 times
dilluted 10^7 times

9. Methodology User Interface Expression Image Path Resource Loading
1. Detect Red range
Image Processing
2. Detect White range
3. Red & Not-White
Binarization
Labeling
Counting
Shows the calculated result

10. Methodology Color Recognition (1)
Change RGB to HSV for extracting the exact value
RGB
HSV

11. Methodology Color Recognition (2)
Th6 = threshold hue of ^6
Ts6 = threshold saturation of ^6
Tv6 = threshold value of ^6
Th7 = threshold hue of ^7
Ts7 = threshold saturation of ^7
TsW = threshold saturation of white
Color Recognition (2)
Detect not-white -> not(lowsat AND MAXvalue)
not-White = not( ( (saturation ≤ TsW) AND (value = MAXVAL) ) )
Detect red -> lowhue
^6 : (hue ≤ Th6) AND (Ts6 ≤ saturation) AND (value ≤ Tv6)
^7 : (hue ≤ Th7) AND (Ts7 ≤ saturation)
red
&
not-white

12. Methodology
Connected Component
8-neighbors

13. Methodology
Connected Component labeling

14. Validation Precision-Recall Curve Precision Recall = TP / (TP+FP)
= TrueDetected / Result
Recall
= TP / (TP+FN)
= TrueDetected / Manual
the source of the image :

15. Validation Precision-Recall Curve - ^6 (1)
Th6 = threshold hue of ^6
Ts6 = threshold saturation of ^6
Tv6 = threshold value of ^6
TsW = threshold saturation of white
Precision-Recall Curve - ^6 (1)
(Ts6=62, Tv6=250, TsW=76.5)
Th=23
Th=23
-> Experimental proof that Th6=23 is best for ^6

16. Validation Precision-Recall Curve - ^6 (2)
Th6 = threshold hue of ^6
Ts6 = threshold saturation of ^6
Tv6 = threshold value of ^6
TsW = threshold saturation of white
Precision-Recall Curve - ^6 (2)
(Ts6=62, Tv6=250, TsW=76.5)
Th=23
Th=23
-> Experimental proof that Th6=23 is best for ^6

17. Validation Precision-Recall Curve - ^7 (1)
Th7 = threshold hue of ^7
Ts7 = threshold saturation of ^7
TsW = threshold saturation of white
Precision-Recall Curve - ^7 (1)
(Ts7=85, TsW=76.5)
Th=22
Th=22
-> Experimental proof that Th7=22 is best for ^7

18. Validation Precision-Recall Curve - ^7 (2)
Th7 = threshold hue of ^7
Ts7 = threshold saturation of ^7
TsW = threshold saturation of white
Precision-Recall Curve - ^7 (2)
(Ts7=85, TsW=76.5)
Th=22
Th=22
-> Experimental proof that Th7=22 is best for ^7

19. Validation

20. Implementation
MFC
MATLAB
Visual Studio 2008
OpenCV

21. Implementation A: Image B: Mode C: Execute D: Mask scroll E: Result
F: ListBox
G: Export *.xls

22. Implementation

23. Limitation
Depending on the performance of the scanner, the quality of image is different.
-> not flexible

24. Thank you.
감사합니다.