Information Technology Reference
In-Depth Information
purchasing a product [
1
,
2
]. The main challenge for these inspection systems is to
combine the image quality results with lower costs and losses in the process,
especially considering the accuracy and reliability of the process. Because of this,
each system should be developed and configured properly in order to consider its
reproducibility and traceability, making it a challenge for the industry, requiring a
greater interface between the professionals of the fields of Engineering, Metrology
and Computation.
This work has the purpose of discussing the main parameters that influence
color measurement in images using computer vision techniques.
4.2 Computational Vision Systems
Computer Vision is the science responsible for the study and application of
methods that enable computers to understand the contents of an image and
interpret important features extracted from this image for a particular purpose [
3
].
The development of Computer Vision Systems (CVS) requires an input data
that is usually obtained from sensors, cameras, or videos, which is an image. The
image is then processed and transformed into some sort of expected information.
Even though a CVS should be organized according to its application, there are
typical steps for all CVS, which can be summarized as: image acquisition, pre-
processing, segmentation, feature extraction, and processing (analysis); all of these
are shown in the block diagram of Fig.
4.1
.
In the image acquisition step, the image of a real object or a scene is trans-
formed into a digital image using an acquisition device (digital cameras, scanners,
videos, etc.). To represent and manipulate these digital images it is necessary to
create mathematical models suitable for this purpose. These are constructed from
an image of the real object, which undergoes a transformation in order to be
verified. This scanning process generates a continuous mapping of the actual
image, which is discretized at various points, called pixels. A matrix is then
formed in a way that each position (x, y) that has information on the gray level or
color associated with f(x, y). The color is represented by color systems and the
most widely used systems are the RGB and HSL.
The preprocessing is the step prior to feature extraction, which aims to improve
the acquired image. It can enhance visibility and the separability between the
background and the objects, without adding information to the image. Among the
techniques of preprocessing, it is possible to highlight the transformation to
grayscale, as well as thresholding and filtering.
After the preprocessing there is the segmentation step whose purpose is to
divide an image into homogeneous units, considering some of its essential char-
acteristics, for example, the gray level of the pixels and texture contrast. These
units are called regions and may correspond to objects in a scene, and are formed
by a group of pixels with similar properties. Through this process of dividing an
image into regions, which will simplify and/or change its representation, it is
Search WWH ::
Custom Search