SLIDE 1
M O C H A
M O CHA 1
In tr o d uctio n
M O CHA - M inimal O ptical Coffee Height A nalysis M achine Vision Java
M O CHA 2
M O C H A M O CHA 1 Java M achine Vision M O CHA - M inimal O - - PDF document
M O C H A M O CHA 1 Java M achine Vision M O CHA - M inimal O - - PDF document
M O C H A M O CHA 1 Java M achine Vision M O CHA - M inimal O ptical Coffee Height A nalysis In tr o d uctio n M O CHA 2 M a ch in e V isio n Digital Image Capture Cost vs Q uality Used relatively cheap digital camera Image
SLIDE 2
SLIDE 3
M a ch in e V isio n
Digital Image Capture
Cost v’s Q uality Used relatively cheap digital camera
Image manipulation using filters and operators
M edian, threshold, max, edge detection, skeletonizing, ...
Inference
Pulling know ledge from a digital image Coffee volume Hough transform Classic detects straight lines G eneral detects precalculated arbitrary shapes
M O CHA 3
SLIDE 4
Im a g e Filter s - Th r esh o ld
M O CHA 4
SLIDE 5
Im a g e Filter s - Skeleto n iz e
M O CHA 5
SLIDE 6
C o ffee Vo lum e - H o w it sh o uld h a ve b een
M O CHA 6
SLIDE 7
C o ffee Vo lum e - H o w it is
M O CHA 7
SLIDE 8
H o ug h Tr a n sfo r m
M O CHA 8
SLIDE 9
C la ssic H o ug h Tr a n sfo r m
Detects straight lines
M O CHA 9
SLIDE 10
C la ssic H o ug h Tr a n sfo r m
Find edges
M O CHA 10
SLIDE 11
Ex p la n a tio n o f C la ssic H o ug h Tr a n sfo r m
Each point in an image votes for all the lines that go through it
y = mx + x- s(
M O CHA 11
SLIDE 12
C la ssic H o ug h Tr a n sfo r m
Hough space
M O CHA 12
SLIDE 13
Ex p la n a tio n o f C la ssic H o ug h Tr a n sfo r m
M O CHA 13
SLIDE 14
G en er a l H o ug h Tr a n sfo r m
Detects arbitrary shapes
W ill add ability to detect rotated and scaled shapes
M O CHA 14
SLIDE 15
Sh a p es
Shapes must be built up from an image
M ust have a reference point Stores shape as an array of x, y difference pairs
Each point in a digital image votes to say that it may be part of a shape
M O CHA 15
SLIDE 16
Sh a p e o f C o ffee Po t
M O CHA 16
SLIDE 17
G en er a l H o ug h Sp a ce
M O CHA 17
SLIDE 18
Ex p la n a tio n o f G en er a l H o ug h Tr a n sfo r m
Voting array is thresholded A ny points remaining above the threshold limit indi- cate the likely presence of the shape in the original image
They actually represent reference points These points can be used to map the shape back into image space
M O CHA 18
SLIDE 19
Sh a p e M a p p ed Ba ck o n to Im a g e
M O CHA 19
SLIDE 20
C o n clusio n s
Image manipulation toolkit in Java M achine vision can be done on the cheap Shape building functions W orking classic and general Hough transform im- plementations
N eed to play w ith to determine parameters for successful useage
M O CHA 20