Vibration displacement measurement technology for cylindrical structures using camera images

Đăng ngày 7/3/2019 11:36:43 PM | Thể loại: | Lần tải: 0 | Lần xem: 38 | Page: 12 | FileSize: 3.05 M | File type: PDF
Vibration displacement measurement technology for cylindrical structures using camera images. In this paper, a method is proposed to measure vibration displacements remotely using a camera without having to approach the structure. Furthermore, an estimation method for the measurement resolution and measurement error is proposed for the vibration displacement of a cylindrical structure measured using the proposed measurement method. The proposed methods are described, along with experimental results that verify their accuracy.
Available online at www.sciencedirect.com
ScienceDirect
Original Article
VIBRATION DISPLACEMENT MEASUREMENT TECHNOLOGY
FOR CYLINDRICAL STRUCTURES USING CAMERA IMAGES
KI-SUNG SON a, HYEONG-SEOP JEON a, JIN-HO PARK b, and JONG WON PARK c,*
a SEA-AN Engineering Co. RM910, Byuksan Digital Valley II, 481-10, Gasan-Dong, Kumchun-Gun, Seoul, 153-783, South Korea
b Korea Atomic Energy Research Institute, 1045 Daedeok-daero, Yuseong-gu, Daejeon 305-353, South Korea
c Chungnam National University, 99 Daehak-ro, Yuseong-gu, Daejeon 305-764, South Korea
a r t i c l e
i n f o
a b s t r a c t
Article history:
Acceleration sensors are usually used to measure the vibration of a structure. Although
Received 22 October 2014
this is the most accurate method, it cannot be used remotely because these are contact-
Received in revised form
type
sensors.
This
makes
measurement
difficult
in
areas
that
cannot
be
easily
7 January 2015
approached by surveyors, such as structures located in high or dangerous areas. Therefore,
Accepted 14 January 2015
a method that can measure the structural vibration without installing sensors is required
Available online 27 March 2015
for the vibration measurement of structures located in these areas. Many conventional
studies have been carried out on nonecontact-type vibration measurement methods using
Keywords:
cameras. However, they have been applied to structures with relatively large vibration
Camera
displacements such as buildings or bridges, and since most of them use targets, people still
Cylindrical structure
have to approach the structure to install the targets. Therefore, a new method is required
Image processing
to supplement the weaknesses of the conventional methods. In this paper, a method is
Measurement
proposed to measure vibration displacements remotely using a camera without having to
Pipe
approach the structure. Furthermore, an estimation method for the measurement reso-
Vibration displacement
lution and measurement error is proposed for the vibration displacement of a cylindrical
structure measured using the proposed measurement method. The proposed methods are
described, along with experimental results that verify their accuracy.
Copyright © 2015, Published by Elsevier Korea LLC on behalf of Korean Nuclear Society.
1.
Introduction
much time and expense are required when measuring a
structure that is difficultto approach or located in a dangerous
Contact-type acceleration sensors are usually used for vibra-
area such as a high-temperature/high-pressure structure and
tion measurement of structures. However, because contact-
high radiation area in nuclear power plants. Therefore, a
type sensors need to be directly attached to the structure,
remote measurement method is required for measuring the
approaching the structure is necessary, and consequently,
vibration of structures in areas that are difficult to approach.
* Corresponding author.
This is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (http://
creativecommons.org/licenses/by-nc/3.0) which permits unrestricted non-commercial use, distribution, and reproduction in any
medium, provided the original work is properly cited.
1738-5733/Copyright © 2015, Published by Elsevier Korea LLC on behalf of Korean Nuclear Society.
489
Accordingly, along with the rapid advancement of camera
coordinate values instead of approximated integer-type co-
hardware, many studies have been carried out on vibration
ordinate values. Therefore, instead of conventional image
measurement using cameras in recent years [1e6]. Although
processing techniques, an improved real-number-type edge
the
measurement
resolutions
and
sampling
speeds
of
detection method is required.
methods using cameras are lower than those of sensor mea-
surement methods, they have the great advantage that
remote
and
simultaneous
multipoint
measurements
are
2.1.
Basic theory
possible. Therefore, many studies have been conducted on
relatively large structures that have low frequencies and large
In an image, the boundary between the subject and back-
vibration displacements such as bridges and buildings. How-
ground is not perfectly distinguished, as in reality. In most
ever, most of the conventional methods use targets, and in
cases, the brightness value of the edge area, which is the
general, the measurement resolution is constrained by the
boundary between the subject and background in the image,
camera
resolution.
Furthermore,
these
methods
have
a
does not change radically, but changes gradually. This is
disadvantage, where measurement is impossible, when the
because of image resolution problems and blurring.
structure is difficult to approach by surveyors because these
Afteracquiringcontinuousimagesofthevibratingstructure,
methods require the installation of targets such as contact
Fig. 1 shows the influence of the vibration on the brightness
sensors on the object being measured. Recently, a vibration
changes and brightness variation rates in the edge area. As
measurement method using the edges of a structure was
shown in Fig. 1A, a vibration measurement point (edge area)
proposed, but the error range of the measurement signals and
with a size of 9  9 was selected from the whole image S (m  n
the measurement resolutions could not be accurately deter-
size). Here, the edge is in the horizontal direction, whereas the
mined, and the measurement environment had an influence
subjectvibratesintheverticaldirection.InFig.1B,ninepixelsin
on the image noise [6].
theverticaldirectionareselectedarbitrarily,andthebrightness
In this paper, a remote vibration displacement measure-
changesinthetimedomainareshowninFig.1C.Thebrightness
ment method is proposed that uses an improved edge detec-
value changes of each pixel from pixel P0 to P8, which corre-
tion method and a camera. The proposed method uses the
spond to the edge area, were due to the vibration ofthe subject,
second derivative of an image and has a measurement reso-
and the changes appeared clearly at the boundary between the
lutionequalto thepixelresolution,orhigher.Thismethodhas
background and subject. Fig. 1D shows the brightness values of
the advantage that multipoint measurement is possible. In
P0~P8fromthe38th and70th frameimages,whichcorrespondto
addition, we propose a method for estimating the measure-
the vibration P-P. The vibration of the structure affected the
ment resolution and error range when measuring the vibra-
images,anditcanbeseenthatthebrightnessvaluesintheedge
tion displacement of a cylindrical structure.
area changed as a result of the vibration.
Fig. 2 shows the first and second derivative results in the
edge area using a regular image processing technique. Most of
2.
Vibration displacement measurement
the regular image processing techniques for determining the
method using cameras
edge in an image use the first or second derivative.
The first derivative of an image determines the gradient
The single pixel value of a grayscale image is determined by
between pixels by obtaining the differences in neighboring
the brightness value (0~255), which depends on the amount of
pixel values, as shown in Eq. (1). The conventional image
light received by the image sensor. This value can be defined
processing techniques for detecting the edge using the first
as the brightness value of a two-dimensional area transmitted
derivative include the Sobel, Prewitt, and Roberts methods [7].
from the photographed area of the subject to the image sensor
through the optical lens. Measuring the vibration displace-
GðxÞ ¼ fðx  1;yÞ fðx þ1;yÞ
(1)
ment of a subject begins with the movement detection of a
specific point on the subject. The measurement point can be a
GðyÞ ¼ fðx;y 1Þ fðx;y þ1Þ
mark specified by a user or an edge of the subject. Conven-
wherex is a vertical axis coordinate value of the image, y is the
tional image processing methods for edge detection include
horizontal axis coordinate value, f is an input image, and g is a
Laplacian,
Sobel,
and
Canny
edge
detectors
[7].
These
first derivative for f. In general, the edge detection method
methods determine the edge pixel as the pixel having a strong
using the first derivative determines the integer-type pixel
edge component in the image using the first or second de-
coordinate having the maximum derivative value as the edge.
rivatives.
Conventional
vibration
measurement
methods
A method has been proposed whereby these values are
using cameras detect the edge and monitor its variation rate
expressed in the form of a Gaussian distribution, and by
using one of these image processing methods. However, the
regarding it as a probability distribution and obtaining the
edge coordinates determined through the image processing
centerofmassvalue,therealnumber-typeedgeis determined
technique introduced earlier are approximated integer-type
[6]. However, it is difficult to accurately determine the edge
pixel coordinate values. Therefore, if a regular image pro-
area used in the calculation, and measurement error occurs if
cessing technique is used to measure the vibration displace-
the calculation area is incorrectly determined.
ment, a large error will be inherent. To obtain the coordinates
The conventional methods for detecting edges using the
of the actual edges of the vibration-displacement measure-
second derivative, which reacts more sensitively to the edge,
ment subject, they should be expressed as real number-type
include the Laplacian and Laplacian of Gaussian [7,8].