DESCRIPTION
TECHNICAL FIELD
[0001] The present invention relates to a measuring device, a
measuring method, and a measuring program for measuring the
diameter of a cylindrical object based on a parallax image
obtained by making exposures from two viewpoints.
BACKGROUND ART
[0002] Taking a viewpoint image of an observation object from
each of two viewpoints a fixed distance apart and indicating a
specific point situated on the observation object in each of the
viewpoint images allow calculation of a distance from an imaging
position to the observation object based on the principle of
triangulation, with the use of these viewpoint images as a
parallax image.
[0003] Likewise, it is also possible to measure the diameter of
a cylindrical object by taking two viewpoint images of the
cylindrical object from two viewpoints and using the viewpoint
images as a parallax image. According to a technique described
in patent document 1, a plane orthogonal to a central axis of the
cylindrical object is virtually determined from the parallax
image. A pair of tangents to a circle, which corresponds to a
line of intersection of this plane and a cylindrical surface of
the object, is drawn in the plane to obtain the coordinates of
each contact point and the coordinates of an intersection point
of the pair of tangents. In addition, considering the distance
between the two viewpoints, the distance from the imaging position
to the cylindrical object, and the like, the diameter of the
cylindrical object is calculated.
3
[0004] Also, as is known by patent document 2, an active method
can be used by which viewpoint images are taken from two
viewpoints, while light beams are applied to three points situated
on the cylindrical object. This method can use light spots of
the light beams projected to the cylindrical object together with
the parallax image, and has the advantage of permitting the
measurement without any inconvenience even if the cylindrical
object is inclined in its depth direction.
PRIOR ART DOCUMENTS
Patent Documents
[0005] Patent Document 1: Japanese Patent Laid-Open Publication
No. 7-139918
Patent Document 2: Japanese patent Laid-Open Publication
No. 2010-243273
SUMMARY OF THE INVENTION
Problems to be Solved by the Invention
[0006] However, according to the technique described in the
patent document 1, an error easily creeps in, in the course of
drawing the tangents to each circumferential surface, which is
imaged as an outline of the cylindrical object in each viewpoint
image, and mathematizing the tangents. Furthermore, in a case
where the cylindrical object is imaged in a state of being inclined
in a depth direction of the image, an error tends to creep in also
in the course of setting the plane orthogonal to the central axis
of the cylinder. Also, the technique of the patent document 2,
which has to apply the light beams to a plurality of portions of
the observation object, has difficulties such as increase in size
of a device and need for much expense in time and effort of initial
setting and adjustment.
[0007] An object of the present invention is to provide a
4
measuring device and a measuring method that can measure the
diameter of a cylindrical object with high precision based on an
easily taken parallax image, even if the position of the
cylindrical object whose diameter is to be measured is not
specified.
Means for Solving the Problems
[0008] A diameter measuring device of a cylindrical object
according to the present invention includes a parallax image
storage, a parallax image display, an outline detector, a
measurement point designator, a corresponding point deriver, a
measurement point updater, and a diameter calculator. The
parallax image storage stores a first viewpoint image and a second
viewpoint image obtained by imaging the cylindrical object from
a first viewpoint and a second viewpoint, respectively, as a
parallax image. The parallax image display displays the first
viewpoint image and the second viewpoint image. The outline
detector detects in the first viewpoint image a first outline and
a second outline that are parallel to a central axis of the object.
The measurement point designator designates a first measurement
point and a second measurement point on the detected first and
second outlines, respectively. The corresponding point deriver
derives in the second viewpoint image a first corresponding point
and a second corresponding point that correspond to the first and
second measurement points. The measurement point updater scans
the second measurement point on the second outline while fixing
the first measurement point, and updates the second corresponding
point whenever the second measurement point is updated in
accordance with a scan position. The diameter calculator
calculates a diameter of the object using the first corresponding
point and the second corresponding point whenever the second
corresponding point is updated by the measurement point updater,
5
and determines a minimum value of the calculated diameters as the
diameter of the object. In view of increasing measurement
precision, out of the first and second measurement points, one
having a shorter object distance is preferably set as the first
measurement point.
[0009] In a case where in a course of scanning the second
measurement point, the second measurement point is across a first
measurement frame set in advance in the first viewpoint image or
the second corresponding point is across a second measurement
frame set in advance in the second viewpoint image, the
measurement point updater preferably fixes the second measurement
point on the second outline of the first parallax image such that
the second measurement point and the second corresponding point
are confined within the first measurement frame and the second
measurement frame, respectively, and updates the first
measurement point while scanning the first measurement point on
the first outline, and sequentially updates the first
corresponding point in the second parallax image.
[0010] When the measurement point designator designates the
first and second measurement points on the first and second
outlines of the first viewpoint image, the corresponding point
deriver preferably derives the first and second corresponding
points corresponding to the first and second measurement points,
respectively, in the second viewpoint image by a stereo matching
process, and retrieves a first corresponding outline and a second
corresponding outline of the second viewpoint image that
correspond to the first and second outlines in a straight line
detection area having a certain width that is set in each of
extending directions of a line segment connecting the first and
second corresponding points, and updates positions of the first
and second corresponding points to positions on the detected first
and second corresponding outlines.
6
[0011] When the measurement point designator designates the
first and second measurement points on the first and second
outlines of the first viewpoint image, the corresponding point
deriver may derive the first and second corresponding points
corresponding to the first and second measurement points,
respectively, in the second viewpoint image by a stereo matching
process, and then change orientations of straight line detection
areas having a certain width that are set adjacently to the first
corresponding point and the second corresponding point of the
second viewpoint image based on a line segment connecting the
first measurement point and the second measurement point or an
angle  of a line segment connecting the first corresponding point
and the second corresponding point relative to a parallax
direction, and then retrieve in the straight line detection area
a first corresponding outline and a second corresponding outline
of the second viewpoint image that correspond to the first and
second outlines, and update positions of the first and second
corresponding points to positions on the detected first and second
corresponding outlines.
[0012] In a case where the above angle  is 45 or more, the
straight line detection area is preferably set under an upper
corresponding point and over a lower corresponding point out of
the first and second corresponding points, and in a case where
the angle  is less than 45, the straight line detection area
is set on a right side of a left corresponding point and on a left
side of a right corresponding point out of the first and second
corresponding points.
[0013] A diameter measuring method of a cylindrical object
according to the present invention includes a parallax image
storage step, a parallax image display step, a parallax image
display step, an outline detection step, a measurement point
designation step, a corresponding point derivation step, a
7
measurement point update step, and a diameter calculation step.
In the parallax image storage step, a first viewpoint image and
a second viewpoint image obtained by imaging the cylindrical
object from a first viewpoint and a second viewpoint are stored
to a memory as image data. In the parallax image display step,
the first viewpoint image and the second viewpoint image are
displayed, preferably side by side under an equal magnification.
In outline detection step, a first outline and a second outline
of the object that are parallel to a central axis of the object
are detected in the first viewpoint image. In the measurement
point designation step, a first measurement point and a second
measurement point are designated on the first and second outlines,
respectively. In the corresponding point derivation step, a
first corresponding point and a second corresponding point that
correspond to the first and second measurement points are derived
in the second viewpoint image. In the measurement point update
step, the second measurement point is scanned on the second
outline while the first measurement point is fixed, and the second
corresponding point is updated whenever the second measurement
point is updated in accordance with a scan position. In the
diameter calculation step, a diameter of the object is calculated
using the first corresponding point and the second corresponding
point whenever the second corresponding point is updated, and
determines a minimum value of calculated diameters as the diameter
of the object. Out of the first and second measurement points,
one having a shorter object distance may be set as the first
measurement point.
[0014] In a case where in a course of scanning the second
measurement point, the second measurement point is across a first
measurement frame set in advance in the first viewpoint image or
the second corresponding point is across a second measurement
frame set in advance in the second viewpoint image, the
8
measurement point update step preferably fixes the second
measurement point on the second outline of the first parallax
image such that the second measurement point and the second
corresponding point are confined within the first measurement
frame and the second measurement frame, respectively, and updates
the first measurement point while scanning the first measurement
point on the first outline, and sequentially updates the first
corresponding point in the second parallax image.
[0015] When the first and second measurement points are
designated on the first and second outlines of the first viewpoint
image in the measurement point designation step, the
corresponding point derivation step may derive the first and
second corresponding points corresponding to the first and second
measurement points, respectively, in the second viewpoint image
by a stereo matching process, and retrieve in the second viewpoint
image a first corresponding outline and a second corresponding
outline that correspond to the first and second outlines within
a straight line detection area having a certain width that is set
in each of extending directions of a line segment connecting the
first and second corresponding points, and update positions of
the first and second corresponding points to positions on the
detected first and second corresponding outlines.
[0016] When the first and second measurement points are
designated on the first and second outlines of the first viewpoint
image in the measurement point designation step, the
corresponding point derivation step effectively derives the first
and second corresponding points corresponding to the first and
second measurement points, respectively, in the second viewpoint
image by a stereo matching process, and then changes orientations
of straight line detection areas having a certain width that are
set adjacently to the first corresponding point and the second
corresponding point of the second viewpoint image based on a line
9
segment connecting the first measurement point and the second
measurement point or an angle  of a line segment connecting the
first corresponding point and the second corresponding point
relative to a parallax direction, and then retrieves in the
straight line detection area a first corresponding outline and
a second corresponding outline of the second viewpoint image that
correspond to the first and second outlines, and updates positions
of the first and second corresponding points to positions on the
detected first and second corresponding outlines. In a case where
the angle  is 45 or more, the straight line detection area may
be set under an upper corresponding point and over a lower
corresponding point out of the first and second corresponding
points, and in a case where the angle  is less than 45, the
straight line detection area may be set on a right side of a left
corresponding point and on a left side of a right corresponding
point out of the first and second corresponding points.
[0017] A diameter measuring program of a cylindrical object
according to the present invention includes a parallax image
storage step, a parallax image display step, an outline detection
step, a measurement point designation step, a corresponding point
derivation step, a measurement point update step, and a diameter
calculation step. In the parallax image storage step, a first
viewpoint image and a second viewpoint image obtained by imaging
the cylindrical object from a first viewpoint and a second
viewpoint are stored to a memory or the like. In the parallax
image display step, the first viewpoint image and the second
viewpoint image are displayed on a parallax image display. In
the outline detection step, a first outline and a second outline
of the object that are parallel to a central axis of the object
are detected in the first viewpoint image. In the measurement
point designation step, a first measurement point and a second
measurement point are designated on the first and second outlines,
10
respectively. In the corresponding point derivation step, a
first corresponding point and a second corresponding point that
correspond to the first and second measurement points are derived
in the second viewpoint image. In the measurement point update
step, the second measurement point is scanned on the second
outline while the first measurement point is fixed, and the second
corresponding point is updated whenever the second measurement
point is updated in accordance with a scan position. In the
diameter calculation step, a diameter of the object is calculated
using the first corresponding point and the second corresponding
point whenever the second corresponding point is updated, and a
minimum value of the calculated diameters is determined as the
diameter of the object.
Effect of the Invention
[0018] According to the present invention, a new means that
measures the diameter by scanning the measurement point allows
measurement of the diameter of the approximate cylinder with high
precision, even if there is no constant relation maintained
between the position of the approximately cylindrical object
whose diameter is to be measured and the parallax direction, with
the use of a simple hand-hold imaging device such as a 3D-capable
digital camera for imaging the parallax image.
BRIEF DESCRIPTION OF DRAWINGS
[0019] Fig. 1 is a schematic view of a measuring device using
the present invention;
Fig. 2 is a block diagram showing the electrical structure
of the device according to the present invention;
Fig. 3 is a flowchart showing an example of a basic
measurement process according to the present invention;
Fig. 4 is an explanatory view showing an example of a
11
parallax image of a cylindrical object;
Fig. 5 is an explanatory view showing a state of scanning
a measurement point;
Fig. 6 is an explanatory view in deriving the coordinates
of a central axis of the cylindrical object from a pair of
measurement points;
Fig. 7 is an explanatory view in deriving the diameter of
the cylindrical object;
Fig. 8 is a graph showing the correlation between the
coordinates of the measurement point and the calculated diameter;
Fig. 9 is an explanatory view showing an example of a
parallax image in which the position of the cylindrical object
is inclined;
Fig. 10A is an explanatory view in normally scanning one
of the measurement points;
Fig. 10B is an explanatory view in a case where the scan
of the measurement point is regulated with a measurement frame;
Fig. 10C is an explanatory view in a case where the
measurement point to be scanned is changed;
Fig. 11 is a flowchart showing an example of a process in
changing the measurement point to be scanned;
Fig. 12 is an explanatory view showing a state of updating
a corresponding point that is mistakenly set in a right viewpoint
image;
Fig. 13 is a flowchart showing an example of a process of
preventing the wrong setting of the corresponding point in the
right viewpoint image;
Fig. 14 is a flowchart showing an example of a process of
preventing the wrong setting of the corresponding point due to
a background pattern;
Fig. 15A is an explanatory view in deriving the
corresponding point without being influenced by a vertical line
12
in a background;
Fig. 15B is an explanatory view in deriving the
corresponding point without being influenced by a horizontal line
in the background; and
Fig. 15C is an explanatory view in deriving the
corresponding point without being influenced by an oblique line
in the background.
DESCRIPTION OF INVENTION
[0020] In Fig. 1, a twin-lens stereo camera 1 having parallax
in a horizontal direction takes an image of a cylindrical object
4. A left viewpoint image and a right viewpoint image, which have
parallax therebetween, are taken independently through each lens
1a, 1b of the stereo camera 1, and recorded to a memory of the
stereo camera 1 as image data for composing a parallax image. The
stereo camera 1 is automatically in focus when making an exposure.
Object distance information that is obtained in focusing is stored
with the image data.
[0021] To measure the diameter of the imaged cylindrical object
4, the image data of the parallax image is transferred from the
digital camera 1 to a personal computer 2 through a communication
cable 3. The personal computer 2 has an electrical structure as
shown in functional blocks of Fig. 2, and is used with an
application program that calculates the diameter of a cylindrical
object based on the parallax image taken by the stereo camera 1.
Note that, the cylindrical object is not limited to a circular
cylinder shape of strict meaning, but may be an elliptical
cylinder whose cross section is substantially regarded as a
circle, a circular truncated cone whose top surface and bottom
surface have slightly different diameters, or the like.
[0022] In Fig. 2, each of the left viewpoint image and the right
viewpoint image obtained by the stereo camera 1 is inputted as
13
the image data to a data input port 6. The inputted image data
is stored to a data memory 17 together with tag data, which differs
on a viewpoint image basis. The tag data includes various imaging
conditions such as information on the distance between principal
points of the lenses 1a and 1b, the object distance information,
and imaging magnification information.
[0023] A viewpoint image coordinate reader 7 reads the
coordinates of a point designated in the viewpoint image, based
on the image data of the viewpoint image. As the unit of the
coordinates, a pixel number of an image sensor for imaging the
left and right viewpoint images is used. An outline determiner
8 evaluates an image pattern of the cylindrical object 4 formed
in the viewpoint image, and detects outlines of a peripheral
surface that manifest themselves as a pair of parallel lines. An
input operation unit 9 is used in setting measurement points on
the outlines of the viewpoint image in accordance with an input
operation of an operator.
[0024] When the measurement points are set in the pair of outlines
of the viewpoint image by the input operation of the operator,
a measurement point scanner 10 shifts (scans) one of the
measurement points along the outline. When a pair of measurement
points is designated, which measurement point to fix and which
measurement point to scan may be selectively determined by the
operator, but may be determined automatically in accordance with
a designation order, other factors, and the like described later
on. Scanning one of the measurement points varies the distance
between the measurement points. The positions of the pair of
measurement points are determined so as to minimum the distance
therebetween. A 3D coordinates arithmetic unit 12 performs a
coordinate analysis of the image of the cylindrical object 4 in
each viewpoint image, to obtain outline coordinates and central
axis coordinates of each viewpoint image, the parallax, and the
14
like. A central axis coordinates and diameter calculator 13
calculates the central axis coordinates and the diameter of the
cylindrical object 4 in 3D measurement space derived by the 3D
coordinates arithmetic unit 12, whenever scanning the measurement
point.
[0025] These functional blocks are connected to a system
controller 18, a program memory 18, and a working memory 19 through
a bus 11. The program memory 18 stores the application program
for calculating the diameter of the cylindrical object 4, which
includes a judgment program for deriving parallax information
based on the pair of viewpoint images and recognizing an image
of the cylindrical object 4 in each viewpoint image, a sequential
program for actuating each of the functional blocks described
above in a predetermined order, and the like. The system
controller 14 controls the entire performance of the application
program stored in the program memory 18 using the working memory
19.
[0026] On an image display 16, the parallax image composed of
the left and right viewpoint images is displayed. When the
measurement points are inputted, the image display 16 displays
the positions of the inputted measurement points on its screen.
Note that, the image display 16 also displays the coordinates of
the input positions of the measurement points, the position of
the diameter that is virtually set in the course of calculation
of the diameter, and other various types of data requiring
verification. Integrating a touch panel into the image display
16 allows direct input of the measurement points on a display
screen that is displaying the parallax image, using a finger or
a touch pen.
[0027] The fundamental operation of the above structure will be
described. To calculate the diameter of the cylindrical object
4, first and second viewpoint images viewed from different
15
viewpoints are required. Although the direction of parallax
between the first and second viewpoint images is not limited to
a horizontal direction, as in the case of the stereo camera 1 shown
in Fig. 1, description is presented with the use of a left viewpoint
image and a right viewpoint image having parallax in the
horizontal direction, as the first viewpoint image and the second
viewpoint image composing a parallax image.
[0028] Fig. 3 shows examples of the viewpoint images. Images
on a left side represent various types of left viewpoint images
21L, 22L, and 23L. Right viewpoint images 21R, 22R, and 23R
corresponding to the left viewpoint images 21L, 22L, and 23L are
represented on a right side. As shown in the images of the
cylindrical object 4, the shape and the outline of the cylindrical
object 4 vary among the left and right viewpoint images 21L to
23L and 21R to 23R, in accordance with a vertical shift of a view
point or imaging of the cylindrical object 4 in an inclined manner.
The diameter of the cylindrical object 4 can be calculated from
any of these viewpoint image pairs by a processing procedure of
Fig. 4.
[0029] According to Fig. 4, in a data input process S1, the image
data of the left viewpoint image and the right viewpoint image
of the cylindrical object 4 taken by the stereo camera 1 is inputted
to the data memory 17. Based on the image data stored to the data
memory 17, the left viewpoint image 21L and the right viewpoint
image 21R of the cylindrical object 4 are displayed side by side
at an equal magnification on the image display 16, as shown in
Fig. 2.
[0030] In the next outline detection process S2, a pair of
outlines formed by the peripheral surface of the cylindrical
object 4 is extracted from the left viewpoint image 21 and the
right viewpoint image 21R. Taking Fig. 5 as an example, a first
outline 24a and a second outline 24b that are parallel to each
16
other are detected in the left viewpoint image 21L, and left and
right outlines 25a and 25b are detected in the right viewpoint
image 21R. In this example, it is judged by processing the image
data that the cylindrical object 4 is vertically erected without
having a large depression or elevation angle in the image, so the
pair of outlines parallel to the central axis is detected. Note
that, in the case of the left and right viewpoint images 22L and
23R and the viewpoint images 23L and 23R as shown in Fig. 3, it
is possible to judge the position of the cylindrical object 4 in
the image by the image processing, and detect the pair of outlines
formed by the peripheral surface with respect to the central axis.
[0031] A measurement point input process S3 is a process of
designating a first measurement point 26 and a second measurement
point 27 on the first and second outlines 24a and 24b, which are
obtained by the outline detection process S2, of one of the
viewpoint images, for example, the left viewpoint image 21L. This
process requires operation from the input operation unit 9, e.g.
a cursor operation using a mouse or a touch operation to the touch
panel integrated in the image display 16. Note that, in a case
where a position in the vicinity of each of the two outlines
detected by the outline detection process S2 is designated as the
measurement point, the measurement point is automatically
corrected to a position on the outline and inputted. In a case
where a position far away from each of the two outlines detected
by the outline detection process S2 is designated as the
measurement point, an error message is displayed to suggest
re-input of the measurement point.
[0032] According to a corresponding point deriving process S4,
when the measurement points 26 and 27 are designated in the left
viewpoint image 21L, corresponding points 28 and 29 are
automatically set on the outlines 25a and 25b of the right
viewpoint image 21R by a stereo matching process. In the
17
following parallax deriving process S5, the parallax between the
left and right viewpoint images 21L and 21R with respect to the
central axis of the cylindrical object 4 is calculated from the
coordinates to the central axis of the left viewpoint image 21L
based on the measurement points 26 and 27 and the coordinates of
the central axis of the right viewpoint image 21R derived in a
like manner.
[0033] To calculate the parallax, it is necessary to calculate
the coordinates of the central axis in each viewpoint image. A
concrete method for the calculation is as follows. As shown in
Fig. 6, in a case where a point A and a point B situated on a surface
31 of projection (corresponding to an imaging surface of an image
sensor) of the left viewpoint image 21L are a pair of measurement
points situated on the outlines that the peripheral surface of
the cylindrical object 4 forms on its side, the coordinates C of
the central axis passing through the center of a cylinder can be
obtained.
[0034] In Fig. 6, OD, DA, and DB can be obtained from a focal
length of the lens 1a, a pixel size, a pixel number, and the like,
and then OA and OB can be obtained using the Pythagorean theorem.
Moreover, OA:OB=AC:BC derives the coordinates of C using OA, OB,
and AB. The same steps are applied to the right viewpoint image
21R, and therefore it is possible to obtain the parallax between
the left and right viewpoint images with respect to the position
of the central axis of the cylindrical object 4, by using the
obtained coordinates of the central axes in the two viewpoint
images.
[0035] In a diameter deriving process S6, the coordinates of the
central axis and the diameter of the approximate cylinder, being
an object to be measured, can be calculated from the pair of
measurement points 26 and 27 inputted in the measurement point
input process S3, the pair of corresponding points 28 and 29
18
derived in the corresponding point deriving process S4, and the
parallax between the two viewpoint images with respect to the
positions of the central axes obtained in the parallax deriving
process S5.
[0036] Fig. 7 shows a three-dimensional system in one of the
viewpoint images of the cylinder. In Fig. 7, the angle FHE = the
angle FGH by a theorem about a tangent of a circle, and the angle
EFH = the angle HFG = 90, so the triangle EFH and the triangle
HFG are geometrically similar. Accordingly, EF:FH=HF:FG, and
the following mathematical expression (1) is derived. Note that,
x and L2 are three-dimensional lengths. P is the number of pixels
between the two points on the outlines, and  is a coefficient
for calculating a length from the number P of the pixels at the
three-dimensional length x. The value of  can be calculated
based on the object distance and an imaging magnification
contained in the tag data, which is attached to each parallax image
data stored to the data memory 17.
[0037] [Mathematical Expression 1]
: L x
2
Px
2
x : Px 2 



(1)
x: a three-dimensional length
L2: a three-dimensional length
P: the number of pixels between two points on the outlines
: a coefficient for calculating a length from the number
P of the pixels at the three-dimensional length x
[0038] Organizing the mathematical expression (1) derives a
mathematical expression (2). Thus, it is known that the
three-dimensional length x is a physical quantity that can be
calculated from the three-dimensional length L2, which is
obtained from the coordinates of the central axes in the two
viewpoint images and the parallax derived in the parallax deriving
process S5.
19
[0039] [Mathematical Expression 2]
P 4
x 4L2 2
2
 
 (2)
[0040] The distance D of the cylindrical object 4, being an object
whose diameter is to be measured, can be expressed by x, L2, P,
and  using the Pythagorean theorem, and a mathematic expression
(3) is derived.
[0041] [Mathematical Expression 3]
 
2
2
2 2
Px x L 2 D 



 
   (3)
D: the diameter of a cylinder
[0042] Using the mathematic expressions (2) and (3) can calculate
the diameter D of the cylindrical object 4, being the object whose
diameter is to be measured.
[0043] By the way, the calculation process of the diameter D is
carried out on the precondition that the first and second
measurement points 26 and 27 are inputted in the left viewpoint
image 21L. Although the first and second measurement points 26
and 27 are set on the outlines 24a and 24b, respectively, the
vertical positions of the first and second measurement points 26
and 27 are different. Taking Fig. 5 as an example, it is desirable
in principle that the measurement points 26 and 27 are designated
so as to minimize a mutual distance between the measurement points
26 and 27. However, precise designation is difficult in actual
fact, and calculation with the measurement points 26 and 27 having
the different vertical positions brings about a large error in
the diameter D.
[0044] Considering above, a measurement point update process
performed in steps S7 to S12 is provided. In this measurement
point update process, while the measurement point 26, being one
of the first and second measurement points 26 and 27 designated
in the left viewpoint image 21L is fixed, the other measurement
20
point 27 is scanned along the outline 24b and updated to each scan
position. At this time, the corresponding point 29 in the right
viewpoint image 21R that corresponds to the measurement point 27
in the left viewpoint image 21L is sequentially updated too by
the stereo matching process in synchronization with the scan of
the measurement point 27. Whenever the measurement point 27 and
the corresponding point 29 are updated, the diameter calculation
process is repeated in a like manner, and the calculation results
of the coordinates of the central axes of the cylinder and the
diameter are stored sequentially. Note that, while the
measurement point 27 is fixed, the other measurement point 26 may
be scanned along the outline 24a. In this case, the corresponding
point 28 is scanned in the right viewpoint image 21R in a
synchronized manner.
[0045] In the step 7, in the course of sequentially calculating
the diameter D, storing a smaller value has higher priority than
storing a larger value. In the steps S8 to S10, if a latter
diameter D calculated afterward is enough larger than a former
diameter D calculated before, the direction of the scan is judged
to be opposite and hence is changed to an appropriate
predetermined direction. Also, in the step S8, in a case where
a newly calculated diameter D is larger than a former calculation
value and its increment is smaller than a minute amount , it
is judged that a minimum value of the diameter D is calculated
just moments before the calculation of the new diameter D. The
steps S11 and S12 carry out a process for sequentially updating
the minimum value of the diameter D.
[0046] While the measurement point 27 is scanned in the
appropriated direction, if it is detected in the step S8 that a
new value of the diameter D is increased from a former value by
a degree less than the minute amount , the step S13 judges that
the former value is the minimum value of the diameter D. Note
21
that, Fig. 8 is a plot of the diameter D that is calculated while
the measurement point 27 is scanned on the outline 24b.
[0047] The correlation of Fig. 8 represents how a cross section
that is used for measurement of the diameter is taken in the
cylindrical object 4. The cross section becomes a perfect circle
if the central axis of the cylinder is orthogonal to the cross
section, and a correct diameter is calculated in the diameter
deriving process S6. On the other hand, the cross section becomes
an ellipse, if central axis is not orthogonal to the cross section.
The more the cross section is inclined with respect to the central
axis, the longer major axis the ellipse has.
[0048] The major axis of the ellipse is calculated as the diameter
D of the cylindrical object 4, in the diameter deriving process
S6. Thus, out of the diameters D obtained by scanning the
measurement point 27 in a measurement point updating process, the
minimum value is most suitably adopted as the diameter D of the
cylindrical object 4. Note that, at the time of completing a
sequence of the calculation process, the diameter D calculated
in each scan position of the measurement point may be displayed
on the image display 16 in a list form with being associated with
the coordinates of the scanned measurement point.
[0049] By the way, in a case where the cylindrical object 4 is
not directly opposed as shown in the left and right viewpoint
images 32L and 32R of Fig. 9, out of measurement points 42 and
43 designated on outlines 33 and 34 of the left viewpoint image
32L, the measurement point 43 having a short object distance is
fixed and the measurement point 42 having a long object distance
is preferably scanned. In calculating three-dimensional
information from the parallax image, a short object distance has
a precision advantage, so as shown in the drawing, the farther
measurement point 42 is preferably scanned to the direction of
shortening the object distance to perform the measurement. Also
22
in the right viewpoint image 32, the position of the corresponding
position 44 is sequentially updated in synchronization with the
scan of the measurement point 42.
[0050] As for the scan of the measurement point, the following
improvement is effective. In a left viewpoint image 35L shown
in Fig. 10A, provided that a measurement point 47a being on a
scanned side is shifted toward a measurement point 47b, while a
measurement point 48a is on a fixed side, if a measurement frame
50 is set in advance in the display screen, the scan of the
measurement point is banned outside the measurement frame 50. The
measurement frame 50 is determined in contemplation of image
distortion that tends to occur in a periphery of the screen, in
consideration of a property of a general imaging optical system
including the stereo camera 1. The size of the measurement frame
50 is preferably variable in accordance with measurement
precision.
[0051] The measurement point 47a is scanned toward the scan
position P1 at which the diameter D becomes its minimum, in a normal
operation. In the course of the scan, as shown in Fig. 10B, when
the measurement point 47a arrives at a border of the measurement
frame 50, performing a measurement frame detection process S14
included in a flowchart of Fig. 11 causes a branch to a scanned-side
measurement point switching process S15. In the scanned-side
measurement point switching step S15, the measurement point 47a
is fixed on the border of the measurement frame 50 at that point
in time, and the minimum value of the diameter D calculated before
then is cleared. Then, the other measurement point 48a is
switched to a measurement point on a scanned side. The
measurement point 48a is shifted from an initially designated
position along its border, and is finally scanned to a scan
position P2 at which the minimum value of the diameter D is
obtained. Note that, the right viewpoint image is omitted in the
23
drawing, but the switching of the corresponding points between
the fixed side and the scanned side is performed, in
synchronization with the switching of the measurement points 47a
and 48a as shown in Figs. 10A to 10C.
[0052] As to the switching process of the measurement points
between the fixed side and the scanned side, the following method
can obtain the same effect. In Fig. 10A, the measurement point
47a to be scanned is shifted across the measurement frame 50. The
diameter D becomes its minimum at that point in time when the
measurement point 47a reaches the scan position P1, and the scan
position of the measurement point 47a at the time is checked. In
a case where the scan position is outside the measurement frame
50, the measurement point 47a is drawn back to the border of the
measurement frame 50, and fixed there as the measurement point
47a on the fixed side. Then, the other measurement point 48a is
switched to the scanned side, and scanned along its outline to
the scan position P2.
[0053] Next, an operation that is effective at preventing the
wrong setting of the corresponding point will be described. The
first and second measurement points are designated in the right
viewpoint image, and the corresponding points are automatically
set based on the first and second measurement points by the stereo
matching process in the right viewpoint image. At this time, the
wrong setting of the corresponding points may occur depending on
the set position of the measurement points. For example, Fig.
12 shows a state in which first and second measurement points 51a
and 51b are designated on outlines of a peripheral surface of a
fire extinguisher, being imaged in a left viewpoint image 36L as
an object whose diameter is to be measured.
[0054] On the other hand, in a right viewpoint image 36R of this
fire distinguisher, corresponding points 52a and 52b are derived
by the stereo matching process in accordance with the positions
24
of the measurement points 51a and 51b. Magnifying the measurement
point 52a and the corresponding point 52b corresponding thereto,
while the measurement point 51a is set appropriately on the
outline, the corresponding point 52b is set on a vertical line
of a label stuck on the front of the fire extinguisher, which is
misidentified as a corresponding outline 53a. Calculating the
diameter D in this state brings a value smaller than an actual
value, and causes deterioration in measurement precision. To
prevent this, a corresponding outline detection process as shown
in Fig. 13 is effectively used.
[0055] In the corresponding outline detection process, a
detection area of a certain width is set on the outside of each
corresponding point so as to protrude to the outside of an image
of the cylindrical object, with respect to the corresponding
points 52a and 52b set in accordance with the measurement points
51a and 51b, and a straight line is detected in this detection
area. As an example, Fig. 12 shows a state in which a detection
area 55 is set on the outside of the corresponding point 52b. A
straight line is detected in this detection area 55, and it is
judged whether or not the detected straight line is a
corresponding outline that corresponds to the outline on which
the measurement point 51b is set. In a case where the detected
straight line is confirmed to be the corresponding outline 53b,
the wrongly set corresponding point 52b is automatically updated
to a correct corresponding point 52c shifted on the corresponding
outline 53a.
[0056] Note that, if the detection area 55 is too wide, another
straight line (vertical line in this embodiment) in a background
may be mistakenly detected, so it is preferable to keep the width
of the detection area in a horizontal direction within the
appropriate confines, e.g. of the order of 50 pixels in this
embodiment. As for the vertical direction, setting the detection
25
area to have a length corresponding to the outline detected in
the left viewpoint image 36L is effective at judging whether the
detected straight line is just a pattern line or a corresponding
outline.
[0057] Oppositely to the above embodiment, a straight line in
a background may be mistakenly detected as a corresponding
outline, when the corresponding points are derived after the
designation of the measurement points. To prevent this, a
corresponding outline detection process of Fig. 14 is effective.
In this corresponding outline detection process, when a pair of
measurement points is set, an angle  formed between a line segment
connecting the measurement points and a parallax direction is
checked in step S20. The parallax direction is a horizontal
direction in the above embodiment, and coincides with a horizontal
array direction of the pixels in the imaging surface. Note that,
the angle  does not have a direction, and a smaller angle (an
acute angle) is used as the angle , out of angles that are formed
between the line segment connecting the pair of measurement points
and the horizontal direction.
[0058] Step S21 judges whether or not the angle  is 45 or more.
Operation branches to step S22 or step S23, according to whether
or not the angle  is 45 or more. The step S22 corresponds to
the case of designating measurement points 57a and 57b in a left
viewpoint image 38L as shown in Fig. 15A. In this case, as to
corresponding points 58a and 58b derived in a right viewpoint
image 38R, the detection areas 55 of a certain width are set under
the corresponding point 58a positioned in an upper portion of the
screen, and over the corresponding point 58b positioned in a lower
portion of the screen.
[0059] Since the detection areas 55 are set like this, no
detection area is set outside an image of a cylindrical object,
in other words, in the background. Therefore, even if a vertical
26
parallel line segment in the background is wrongly detected in
the right viewpoint image 38R by a malfunction of the stereo
matching process, reconfirming a straight line detected in the
detection area 55 detects a correct corresponding outline and the
corresponding point 58a, 58b is shifted to the corresponding
outline.
[0060] Likewise, operation of the step S23, which is performed
in a case where the angle  is less than 45, corresponds to the
case of designating measurement points 57a and 57b in the left
viewpoint image 38L as shown in Fig. 15B. In this case, as to
corresponding points 58a and 58b derived in the right viewpoint
image 38R, the detection areas 55 of the certain width are set
on the right side of the corresponding point 58a positioned in
a left portion of the screen, and on the left side of the
corresponding point 58b positioned in a right portion of the
screen. Setting the detection areas 55 in this manner facilitates
detecting the correct corresponding outlines without being
influenced by the background, and is effective at preventing the
wrong setting of the corresponding points 58a and 58b.
[0061] It is conceivable in most cases that the angle  of the
line segment connecting the pair of measurement points 57a and
57b is nearly 90 or 0 as shown in Fig. 15A or 15B, but the above
processes are effectively usable in a case shown by Fig. 15C, for
example. In the case of Fig. 15C, a line segment connecting the
measurement points 57a and 57b has an angle of the order of 30.
In this case, the detection areas 55 are set on the lower right
side of the corresponding point 58a positioned in a left portion,
and on the upper left side of the corresponding point 58b
positioned in a right portion. The detection area 55 is
preferably set in an inclined manner in accordance with the angle
 such that the detection area 55 does not lie off the outline
of the cylindrical object 4. The inclination angle can be
27
determined in accordance with the angle  of the line segment
connecting the measurement points 57a and 57b.
[0062] The present invention is not limited to the embodiments
described above. For example, the number of imaging viewpoints
is not two, but may be increased to three or more with maintaining
the parallax direction, in order to increase the measurement
precision. Instead of using one camera having two imaging
viewpoints, two or more monocular cameras may be used. Otherwise,
if the object whose diameter is to be measured is put still, a
plurality of exposures may be made with shifting one camera in
the parallax direction to obtain a plurality of viewpoint images.
The above embodiments are mainly described about a measuring
device, but the present invention is effectively usable as a
measuring method or a measuring program.
DESCRIPTION OF THE REFERENCE NUMERALS
[0063] 1 stereo camera
2 personal computer
4 cylindrical object
7 viewpoint image coordinate reader
8 outline determiner
9 input operation unit
10 measurement point scanner
12 parallax deriver
13 central axis coordinates and diameter deriver
14 system controller
16 image display
17 data memory
18 program memory
31 surface of projection
50 measurement frame
28

We Claim:-
1. A diameter measuring device of a cylindrical object
characterized in comprising:
a parallax image storage for storing a first viewpoint image
and a second viewpoint image obtained by imaging said cylindrical
object from a first viewpoint and a second viewpoint,
respectively;
a parallax image display for displaying said first
viewpoint image and said second viewpoint image;
an outline detector for detecting in said first viewpoint
image a first outline and a second outline of said object that
are parallel to a central axis of said object;
a measurement point designator for designating a first
measurement point and a second measurement point on said first
and second outlines, respectively;
a corresponding point deriver for deriving in said second
viewpoint image a first corresponding point and a second
corresponding point that correspond to said first and second
measurement points;
a measurement point updater for scanning said second
measurement point on said second outline while fixing said first
measurement point, and updating said second corresponding point
whenever said second measurement point is updated in accordance
with a scan position; and
a diameter calculator for calculating a diameter of said
object using said first corresponding point and said second
corresponding point whenever said second corresponding point is
updated by said measurement point updater, and determining a
minimum value of said calculated diameters as said diameter of
said object.
2. The diameter measuring device of the cylindrical object
29
according to claim 1, characterized in that out of said first and
second measurement points, one having a shorter object distance
is set as said first measurement point.
3. The diameter measuring device of the cylindrical object
according to claim 1 or 2, characterized in that in a case where
in a course of scanning said second measurement point, said second
measurement point is across a first measurement frame set in
advance in said first viewpoint image or said second corresponding
point is across a second measurement frame set in advance in said
second viewpoint image, said measurement point updater fixes said
second measurement point on said second outline of said first
parallax image such that said second measurement point and said
second corresponding point are confined within said first
measurement frame and said second measurement frame,
respectively, and updates said first measurement point while
scanning said first measurement point on said first outline, and
sequentially updates said first corresponding point in said
second parallax image.
4. The diameter measuring device of the cylindrical object
according to one of claims 1 to 3, characterized in that when said
measurement point designator designates said first and second
measurement points on said first and second outlines of said first
viewpoint image, said corresponding point deriver derives said
first and second corresponding points corresponding to said first
and second measurement points, respectively, in said second
viewpoint image by a stereo matching process, and retrieves a
first corresponding outline and a second corresponding outline
of said second viewpoint image that correspond to said first and
second outlines in a straight line detection area having a certain
width that is set in each of extending directions of a line segment
30
connecting said first and second corresponding points, and
updates positions of said first and second corresponding points
to positions on said detected first and second corresponding
outlines.
5. The diameter measuring device of the cylindrical object
according to one of claims 1 to 3, characterized in that when said
measurement point designator designates said first and second
measurement points on said first and second outlines of said first
viewpoint image, said corresponding point deriver derives said
first and second corresponding points corresponding to said first
and second measurement points, respectively, in said second
viewpoint image by a stereo matching process, and then changes
orientations of straight line detection areas having a certain
width that are set adjacently to said first corresponding point
and said second corresponding point of said second viewpoint image
based on a line segment connecting said first measurement point
and said second measurement point or an angle  of a line segment
connecting said first corresponding point and said second
corresponding point relative to a parallax direction, and then
retrieves in said straight line detection area a first
corresponding outline and a second corresponding outline of said
second viewpoint image that correspond to said first and second
outlines, and updates positions of said first and second
corresponding points to positions on said detected first and
second corresponding outlines.
6. The diameter measuring device of the cylindrical object
according to claim 5, characterized in that in a case where said
angle  is 45 or more, said straight line detection area is set
under an upper corresponding point and over a lower corresponding
point out of said first and second corresponding points, and in
31
a case where said angle  is less than 45, said straight line
detection area is set on a right side of a left corresponding point
and on a left side of a right corresponding point out of said first
and second corresponding points.
7. A diameter measuring method of a cylindrical object
characterized in comprising:
a parallax image storage step for storing a first viewpoint
image and a second viewpoint image obtained by imaging said
cylindrical object from a first viewpoint and a second viewpoint;
a parallax image display step for displaying said first
viewpoint image and said second viewpoint image;
an outline detection step for detecting in said first
viewpoint image a first outline and a second outline of said object
that are parallel to a central axis of said object;
a measurement point designation step for designating a
first measurement point and a second measurement point on said
first and second outlines, respectively;
a corresponding point derivation step for deriving in said
second viewpoint image a first corresponding point and a second
corresponding point that correspond to said first and second
measurement points;
a measurement point update step for scanning said second
measurement point on said second outline while fixing said first
measurement point, and updating said second corresponding point
whenever said second measurement point is updated in accordance
with a scan position; and
a diameter calculation step for calculating a diameter of
said object using said first corresponding point and said second
corresponding point whenever said second corresponding point is
updated, and determining a minimum value of said calculated
diameters as said diameter of said object.
32
8. The diameter measuring method of the cylindrical object
according to claim 7, characterized in that out of said first and
second measurement points, one having a shorter object distance
is set as said first measurement point.
9. The diameter measuring method of the cylindrical object
according to claim 8, characterized in that in a case where in
a course of scanning said second measurement point, said second
measurement point is across a first measurement frame set in
advance in said first viewpoint image or said second corresponding
point is across a second measurement frame set in advance in said
second viewpoint image, said measurement point update step fixes
said second measurement point on said second outline of said first
parallax image such that said second measurement point and said
second corresponding point are confined within said first
measurement frame and said second measurement frame,
respectively, and updates said first measurement point while
scanning said first measurement point on said first outline, and
sequentially updates said first corresponding point in said
second parallax image.
10. The diameter measuring method of the cylindrical object
according to one of claims 7 to 9, characterized in that when said
first and second measurement points are designated on said first
and second outlines of said first viewpoint image in said
measurement point designation step, said corresponding point
derivation step derives said first and second corresponding
points corresponding to said first and second measurement points,
respectively, in said second viewpoint image by a stereo matching
process, and retrieves in said second viewpoint image a first
corresponding outline and a second corresponding outline that
33
correspond to said first and second outlines within a straight
line detection area having a certain width that is set in each
of extending directions of a line segment connecting said first
and second corresponding points, and updates positions of said
first and second corresponding points to positions on said
detected first and second corresponding outlines.
11. The diameter measuring method of the cylindrical object
according to one of claims 7 to 9, characterized in that when said
first and second measurement points are designated on said first
and second outlines of said first viewpoint image in said
measurement point designation step, said corresponding point
derivation step derives said first and second corresponding
points corresponding to said first and second measurement points,
respectively, in said second viewpoint image by a stereo matching
process, and then changes orientations of straight line detection
areas having a certain width that are set adjacently to said first
corresponding point and said second corresponding point of said
second viewpoint image based on a line segment connecting said
first measurement point and said second measurement point or an
angle  of a line segment connecting said first corresponding
point and said second corresponding point relative to a parallax
direction, and then retrieves in said straight line detection area
a first corresponding outline and a second corresponding outline
of said second viewpoint image that correspond to said first and
second outlines, and updates positions of said first and second
corresponding points to positions on said detected first and
second corresponding outlines.
12. The diameter measuring method of the cylindrical object
according to claim 11, characterized in that in a case where said
angle  is 45 or more, said straight line detection area is set
34
under an upper corresponding point and over a lower corresponding
point out of said first and second corresponding points, and in
a case where said angle  is less than 45, said straight line
detection area is set on a right side of a left corresponding point
and on a left side of a right corresponding point out of said first
and second corresponding points.
13. A diameter measuring program of a cylindrical object
for making a computer perform:
a parallax image storage step for storing a first viewpoint
image and a second viewpoint image obtained by imaging said
cylindrical object from a first viewpoint and a second viewpoint;
a parallax image display step for displaying said first
viewpoint image and said second viewpoint image;
an outline detection step for detecting in said first
viewpoint image a first outline and a second outline of said object
that are parallel to a central axis of said object;
a measurement point designation step for designating a
first measurement point and a second measurement point on said
first and second outlines, respectively;
a corresponding point derivation step for deriving in said
second viewpoint image a first corresponding point and a second
corresponding point that correspond to said first and second
measurement points;
a measurement point update step for scanning said second
measurement point on said second outline while fixing said first
measurement point, and updating said second corresponding point
whenever said second measurement point is updated in accordance
with a scan position; and
a diameter calculation step for calculating a diameter of
said object using said first corresponding point and said second
corresponding point whenever said second corresponding point is
35
updated, and determining a minimum value of calculated diameters
as said diameter of said object.
Dated this the 14th day of April, 2014