07-10-2016, 02:55 PM
1458213960-imageprocessing.pdf (Size: 644.77 KB / Downloads: 17)
Abstract: This paper proposes a method to calculate the angular frequency of a wind turbine by using Artificial
Vision techniques. For a given video in which several wind turbines appear, the set of frames that compose it are
taken out. Then, a certain frame is analyzed to determine the wind turbines presented in the image. To deal with
that, a preprocessing is carried out in order to improve the image contrast. Applying several filters, main regions
of the image are emphasized remaining those ones which present a linear shape. They are identified as potential
towers. Next, a rectangular region is defined in the upper area of the rotor. For this region, the number of white
pixels is computed for every frame of the video. Maximum values correspond with blade sweeping that allow us
to compute the angular frequency
Introduction
A wind turbine is a machine that converts the kinetic
energy in wind into mechanical energy. If the mechanical
energy is used directly by machinery, the machine
is usually called a windmill. If the mechanical energy
is converted to electricity, the machine is called a wind
generator, or more commonly a wind turbine.
Fixed-speed wind turbines produce voltage
flicker during operations. Some reserchers have tried
to predict flicker emission from wind turbines at a certain
site previously to installation [7]. Others have analyzed
the Doppler spectral content of the wind turbine
clutter (WTC) signal and have characterized it,
in order to develop specific mitigation schemes [2].
Image processing has been used in many scientific
fields, such as in medicine or biology, where researchers
represent different types of cells by its texture
properties [4], or distinguish between alive or
dead cells by analysing their images [6]. In the renewable
energy field, it has been also used to develop
a field calibration technique for aligning a wind direction
sensor to the true north [3], and to monitor the
damage of wind turbine blades by measuring its bending
[1].
In this work, we have used image processing
techniques to calculate automatically the angular frequency
of an unknown number of wind turbines from
a video. The proposed method allows us to analise flicker emission from a wind turbine in such a way
that it is unnecessary to obtain the data from the wind
farms. This avoids to get permissions and other bureaucracy
that is sometimes a tough task.
In section 2 we present the process followed to
identify the towers and the blade sweeping. Obtained
results are showed in section 3. Section 4 gathers the
achieved conclusions.
2 Methods
2.1 Image adquisition
Having a short video with some wind turbines working,
what we need to do first is to separate it into
frames. Matlab has a function that allows us to obtain
the frames of a video, so we use it, and we save those
frames in a directory so that we can use them later as
independent images. We suppose that the videocamera
has not been moving while taking the video. If
not, it would not be possible to apply this method. In
case that the video is recorded from different points of
view, we just need to edit it and save each perspective
as an independent record.
2.2 Preprocessing
Once we have obtain the frames, we choose one to
locate the towers of the wind turbines. In order to do that, we have to process the image so that we can
distinguish between the wind turbines and the background.
The preprocessing consists on turning the
color image into greyscale format, and then, mapping
the intensity values of the image to a new range by
modifying its histogram. To deal with that, the values
in the original image are saturated at low and high
intensities.
Segmentation
Segmentation consists on dividing an image into its
different parts. For doing it, we firstly convert the
greyscale image to a binary image, using the Otsu’s
method [5]. So, a threshold that minimizes the intraclass
variance of the black and white pixels is chosen.
The output binary image has values of 1 (white) for
all pixels in the original image whose luminance is
greater than such threshold and 0 (black) for all other
pixels.