Measuring rainbow trout by using simple statistics - Emerging Trends in Image Processing, Computer Vision, and Pattern Recognition - page 34

2 Experimental prototype
In this section, we describe our experimental prototype, which has been designed as part of
this research.
This novel prototype is essential to collect useful trout's 2D images; therefore, we have me-
ticulously designed it. Figure 2 shows our experimental prototype, which has evolved from
a traditional squared glass fish-cube (prototype version 1). We observed relevant issues from
our first prototype and that knowledge was experimentally analyzed to obtain our second
model. Note that our two prototypes have been experimentally evaluated in a trout farm; so,
we have gathered special knowledge about handling the rainbow trout.
FIGURE 2 Our experimental scenario for measuring rainbow trout. (a) Statistical approach
within a personal computer, (b) vision system, (c) canalization system, (d) illumination system,
(e) specimen to be measured, and (f) database.
Then, as observed in Figure 2 , our experimental prototype consists of three main compon-
ents: canalization, illumination, and vision which are aim to collect RGB trout images using a
standard personal computer.
2.1 Canalization System
We have design a novel canalization system based on opaque-glass within our prototype. This
canalization system poses two main properties. The first property is regarded to its trapezoid-
al shape, which has been decided according to the digital camera's vision field principle. As
long as such trapezoidal shape avoids reflection to be captured when taking a digital image.
As its second property, we can mention that this is a two-canal tray, which prevents occlusion
by taking only one fish per canal and it allows capturing two rainbow trout images per shot.
2.2 Illumination System
To assist our vision system, we have integrated an illumination system, which distributes light
in a uniform way at the botom of the canalisation system. To do this, a light source is located
to an appropriate high to distribute light uniformly over an acrylic diffuser. The light-source's
high was defined by using a bisection approach and measuring the light intensity projected
into the diffuser with photo resistors. We integrated this diffuse illumination to increase con-
trast into the image and highlighting the trout's body.
2.3 Vision System
In order to explore economical technology for our classification system, we have used a stand-
ard 2D LifeCam Studio [ 13 ] camera in this experimentation. This camera is able to capture
RGB-images with a maximum resolution of 1920 × 1080 pixels. In this prototype, this RGB
camera is located at the top of the canalization system to capture downward-view images of