Hardware Reference
In-Depth Information
The first sensing scheme is CCD camera-based. As described in Chap.
1
, CCD
cameras can be used in experiments to view the top sides of droplets simultaneously.
Based on images captured by the CCD camera, droplets can be automatically
located by the control software. The procedure for automatical search of the
droplets can be described as a “template matching” problem. Here a pattern can
be represented as the image of a “typical” droplet. During the matching process, we
move the template image to all possible positions in the image of the entire array
and crop a sub-image that has the same size as the template image. Then the control
software computes the correlation index, which measures the similarity between the
template and the “cropped image”. The correlation factor is calculated on a pixel-
by-pixel basis, and this process is shown in Fig.
2.3
a.
In the control software, all images are stored in grayscale form. These grayscale
images can be encoded as matrices or vectors. Suppose the template image is
represented in a 1-D array:
x
D
.x
1
;x
2
;:::x
N
/.Herex
i
represents the gravel level
of a pixel and N is the total number of pixels in the template image. Similarly,
the cropped sub-image to be compared with the template image can be written
as
y
D
.y
1
;y
2
;:::;y
N
/. Thus the correlation factor between these two images is
defined as:
i D1
N
.x
i
x/
.y
i
y/
cor
D
s
N
;
i D1
i D1
N
.x
i
x/
2
.y
i
y/
2
where x and y are the average gray level in the template image and cropped sub-
image, respectively. The range of correlation factor
cor
is a real number between
1
and
C
1. According to the definition of correlation, a correlation factor with larger
absolute value represents a stronger relationship between two images.
After deriving the correlation factors for all possible locations in the image of
the complete biochip, we obtain the correlation map between the template and the
original input image. Suppose there are droplets on the biochip. The locations of
droplets can be determined by searching for the largest correlation factors in the
correlation map. An example is shown in Fig.
2.3
b, c [
18
].
Figure
2.3
b shows the original input image of the whole chip and the pattern
image, and Fig.
2.3
c is the correlation map, where the best matching locations, i.e.
the coordinations of droplets derived by the control software are (77, 107), (77,
147) and (76, 208). Thus the control software automatically locates the droplets.
According to the image, the sizes and colors of droplets can be further analyzed.
In this manner, the volumes and concentrations of droplets can be acquired after
processing the image taken by the CCD camera.
Instead of searching for droplets in the complete image, we use imaging
techniques to check whether the droplets have been moved to the expected positions.
This procedure is implemented using the following steps:
