bean counting

1
bean counting
Hemina Patel Sai-Ming Law Themis Toache Tony
Girardi
2
image processing

• an image is a 2 dimensional function, eg. f(x,y)
• image processing is the analysis, interpretation,
  and manipulation of images

3
the problem

• picture acquisition
• filtering/thresholding algorithms
• shrinking/separating algorithms
• counting algorithms

4
Why????

• http//www.wellcome.ac.uk/en/wia/gallery.html?imag
  e20
• http//www.visualsunlimited.com/images/watermarked
  /899/899108.jpg
• http//faculty.mc3.edu/jearl/ML/ml-5-2.htm
• http//www.bone-net.de/textgut/ecoli.htm
• http//www.nirgal.net/graphics/e_coli.jpg

e.coli
sarcina lutea bacteria
5
first try
rgb
gray scaled image
after thresholding
after shrinking
6
problems with our first try
problems with double counting
problems with connected beans
7
more problems with our first try
Blurry images
Fragmented beans
8
new counting algorithm

• - check the waiting list for elements.
• traverse the image until a certain pixel is
• found (in the waiting list)

• find the first pixel A and add it to
• the waiting list.

• After adding the elements of As neighborhood
• to the waiting list, we remove A from the
• waiting list, change its color to white and add
  
• it to the visited list.

• Once A is in the waiting list we check its
• neighborhood for more elements and add them
• to the waiting list.

9
new counting algorithm cont.
5. Since B is the first element in the
Waiting List, we add the neighbors of B are
not in the list.
6. After that, we take B out from the list,
and add it to the Visited List.

• 8. Then we add the size of the bean, i.e, total
  number
• of elements in Visited List to the Size List.

7. We follow the same procedures until the
Waiting list is empty.
10
new shrinking/separating algorithm

• works by finding the edge of each bean, and then
  repeatedly subtracting the outer edge from the
  bean
• Shrinking/separating algorithm needs good
  thresholding

11
new filtering and thresholding algorithm
Grayscale
Blue
filtered image

• We used the difference in the red green and blue
  images to achieve separation of the beans

12
lentils
1
2
3
4
13
lentil results
14
(No Transcript)
15
mms
16
counting by color
Blue
Yellow
17
mm results
18
(No Transcript)
19
rice
1
2
3
20
rice results
21
(No Transcript)
22
for the future

• more testing of our algorithms
• apply new filtering separating techniques
• apply our algorithms to new objects

counting red blood cells