Image analysis – Applications – Manufacturing or product inspection
Reexamination Certificate
2001-12-05
2004-07-20
Niebling, John F. (Department: 2812)
Image analysis
Applications
Manufacturing or product inspection
Reexamination Certificate
active
06766047
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
The present invention relates to an improved defects inspection method of a three-dimensional shape.
2. Description of the Related Art
As a method of detecting surface defects of a three-dimensional work for instance, Japanese Unexamined Patent Application Publication No. 10-10053 discloses a method of detecting a defect based on a change or a differential at the boundaries of patterns by irradiating striped-pattern beams on the work.
However, this method detects the presence or absence of defects based simply on changes in the brightness on the work. Thus, it is difficult for the method to detect defects on the surface of work that can form shadows due to complex projections or recessed shapes or that have different shades due to oil contamination or marking made by a marker. This method often produces false alarms by detecting normal marks caused by oil contamination or markings made by a marker as defects.
In order to solve such a problem, defect inspection methods have been invented which use the three-dimensional shape of an object being inspected in accordance with know trigonometric measurement principles.
The principle of a defect inspection method using a three-dimensional shape based on trigonometric measurement principle is explained briefly with reference to FIG.
6
(
a
).
A light projector
1
, for example, a laser beam, emits a beam which is directed by a mirror
3
to linearly irradiate a work
2
with a pattern having a width in the thickness direction (X-axis direction) of FIG.
6
(
a
). A mirror driving means (not shown in the figure) rocks the mirror
3
at a predetermined pitch in a direction which moves the locus of radiation orthogonal to the linear irradiation, in other words, in a direction which moves the radiation in the right-to-left direction (Y-axis direction) of FIG.
6
(
a
). A rotating position detecting means (not shown in the figure) detects the rocking position of the mirror
3
. A camera
4
takes a picture of the work
2
at each rocking position of the mirror
3
by maintaining a predetermined relative position with respect to the mirror
3
. An image processing unit
5
finds the projection/recess shapes of the work
2
by storing a value (mentioned as a distance code hereinafter) indicating the rocking angle of the mirror
3
for a matrix of laser beam detecting positions in a camera coordinate system of the camera
4
for which the linear direction of the irradiation and the moving locus of radiation are two orthogonal axes.
A line L
1
(&agr;) shown in FIG.
6
(
a
) is the optical path of the laser beam that is reflected by the mirror
3
to irradiate the work
2
. A line L
2
is a straight line connecting the point on the work
2
irradiated by the laser beam to the camera
4
.
The position of the mirror
3
relative to the camera
4
, in other words, the length of the straight line connecting the mirror
3
and the camera
4
, remains constant. An angle between the straight line connecting the mirror
3
and the camera
4
and the line L
1
is easily found given knowledge of the rocking angle of the mirror
3
.
An angle between the straight line connecting the mirror
3
and the camera
4
, and the line L
2
is calculated as an angle between the line L
2
and the optical axis of the camera
4
on the basis of the detected laser beam position on a camera coordinate system of the camera
4
.
Accordingly, although it is possible to find distance between the camera
4
and the work
2
in accordance with the trigonometrical measurement principle, the computation is complex and requires a long time for computation. Thus, instead of calculating the distance between the camera
4
and the work
2
, the rocking angles of the mirror
3
which are related to the distances, are stored as a corresponding matrix of a laser beam detecting position on a camera coordinate system of the camera
4
. Therefore, the shape of the work
2
is stored and the need for a complex computation is omitted.
For instance, if it is assumed that a matrix (X, Y) on a camera coordinate system for which the linear direction of the irradiation and the moving locus direction of the irradiation are two orthogonal axes consists of n-row by m-column matrix elements (i
1
, j
1
) to (i
n
, j
m
). Then, if the rocking angle of the mirror
3
is &agr; when the irradiated beam is detected at (X
(x=1 to n)
, j) on the camera coordinate system as shown in FIG.
6
(
a
), the &agr; value is stored in all of the n matrix elements in the X-axis direction of the matrix (X
(x=1 to n)
, j) on the camera coordinate system. It is found that the height of the work
2
(shown only with a solid line) is the same at a section where radiation hits, since the &agr; value is constant.
If there is a protrusion
2
f
(as shown with a chain line in FIG.
6
(
a
)) at the location (i, j) on the surface of the work
2
, the irradiation does not hit the top face of the protrusion
2
f
when the rocking angle of the mirror
3
is &agr;. Irradiation hits the top face of the protrusion
2
f
when, for example, the rocking angle of the mirror
3
is &agr;′. Thus, the &agr;′ value is stored only at the position of the protrusion
2
f
, which is (i, j), among the n matrix elements in the X-axis direction of the matrix (X
(x=1 to n)
, j) on the camera coordinate system Accordingly, the existence of the protrusion
2
f
on the work
2
is confirmed based on a difference between the rocking angles &agr; and &agr;′.
The processing unit
5
repeatedly executes the above-mentioned processing; that is the rocking angles of the mirror
3
, in other words, distance code values are stored at each rocking position of the mirror
3
for a matrix of a laser beam detecting positions in the camera coordinate system of the camera
4
while the mirror
3
is rocked to shift the radiation in the Y-axis direction.
FIG.
7
(
a
) is one example of an image when a distance code value on each matrix is shaded and visualized by making a distance code value stored at each matrix element correspond to a gray-scale density. For instance, the distance code value is 0 (black on a gray scale) when the irradiation is at the right end of the work
2
, and the distance code value is 255 (white on a gray scale) when the irradiation is at the left end of the work
2
. Such an image is referred to as a distance code image hereinafter.
When a surface defect of the work
2
is detected based on such a principle, the projection/recess shape of the perfect work
2
is first found by the image processing unit
5
. Then. the distance code image of the perfect work
2
is obtained as shown in, for instance, FIG.
7
(
a
). The distance code image is mentioned as a distance code image as a reference.
Subsequently, the projection/recess shape of the inspected work
2
is similarly found as described above by the image processing unit
5
, thus obtaining a distance code image of the inspected work
2
. FIG.
7
(
b
) is an example of a distance code image obtained from the inspected work
2
having a defect
6
at the center.
Then, the presence/absence of a defect on the inspected work
2
is determined by a comparison at each matrix element (i
x (x=l to n)
, j
y (y=l to m)
) between distance codes stored in a reference distance code image obtained from the perfect work
2
, and distance codes stored in a distance code image obtained from the inspected work
2
.
Specifically, a reference distance code image as shown in FIG.
7
(
a
) is first compared with a distance code image as an inspection object shown in FIG.
7
(
b
) to find a difference of distance codes at each matrix element (i
x (x=l to n)
, j
y (y=l to m)
). This difference is stored for each matrix element (i
x (x=l to n)
, j
y (y=l to m)
). As the scale of a difference of distance codes is visualized as a corresponding gray-scale density, as described above, an image shown in, for example, FIG.
7
(
c
) is obtained. Such an image is referred to as a differenti
Akabane Taro
Muraki Yutaka
Murota Naoya
Takeshi Masami
Darby & Darby
Niebling John F.
Stevenson Andre′ C.
Suzuki Motor Corporation
LandOfFree
Defect inspection method for three-dimensional object does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Defect inspection method for three-dimensional object, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Defect inspection method for three-dimensional object will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3247985