U.S. patent application number 17/485240 was filed with the patent office on 2022-03-31 for method for detecting lens cleanliness using spectral differential flat field correction.
The applicant listed for this patent is MLOptic Corp. Invention is credited to Peihong Bai, Jiang He, Teresa Zhang, Wei Zhou.
Application Number | 20220099595 17/485240 |
Document ID | / |
Family ID | 1000005901505 |
Filed Date | 2022-03-31 |
![](/patent/app/20220099595/US20220099595A1-20220331-D00000.png)
![](/patent/app/20220099595/US20220099595A1-20220331-D00001.png)
![](/patent/app/20220099595/US20220099595A1-20220331-D00002.png)
![](/patent/app/20220099595/US20220099595A1-20220331-D00003.png)
United States Patent
Application |
20220099595 |
Kind Code |
A1 |
He; Jiang ; et al. |
March 31, 2022 |
METHOD FOR DETECTING LENS CLEANLINESS USING SPECTRAL DIFFERENTIAL
FLAT FIELD CORRECTION
Abstract
A method for detecting lens cleanliness of a lens disposed in a
flat-field optical path, the flat-field optical path including a
light source, the lens, a camera, the light source is a narrow-band
multispectral uniform surface light source, the camera's
light-sensitive surface is disposed perpendicular to an optical
axis of the lens and in the light position of the lens, the method
including collecting the bright-field image data and dark-field
image data in a plurality of spectra through the lens; for each
pixel, performing a spectral differential flat-field correction
operation to yield a plurality of spectral differentials; and
displaying the spectral differentials in the form of a plurality of
images to show a uniformity of each of the plurality of images,
wherein a non-uniform area on each of the plurality of images is
determined to have been caused by an impurity of the lens.
Inventors: |
He; Jiang; (Hangzhou City,
CN) ; Zhang; Teresa; (Albany, NY) ; Zhou;
Wei; (Sammamish, WA) ; Bai; Peihong; (Nanjing
City, CN) |
|
Applicant: |
Name |
City |
State |
Country |
Type |
MLOptic Corp |
Redmond |
WA |
US |
|
|
Family ID: |
1000005901505 |
Appl. No.: |
17/485240 |
Filed: |
September 24, 2021 |
Current U.S.
Class: |
1/1 |
Current CPC
Class: |
G01N 2021/9583 20130101;
G01N 21/958 20130101; G01M 11/0257 20130101; G01N 2021/8825
20130101; G06T 2207/30108 20130101; G06T 5/50 20130101; H04N 5/2256
20130101 |
International
Class: |
G01N 21/958 20060101
G01N021/958 |
Foreign Application Data
Date |
Code |
Application Number |
Sep 27, 2020 |
CN |
2020110300154 |
Claims
1. A method for detecting lens cleanliness of a lens disposed in a
flat-field optical path, the flat-field optical path comprising a
light source, the lens, a camera, the light source is a narrow-band
multispectral uniform surface light source, the camera's
light-sensitive surface is disposed perpendicular to an optical
axis of the lens and in the light position of the lens, said method
comprising: (a) collecting the bright-field image data and
dark-field image data in a plurality of spectra through the lens;
(b) for each pixel, performing a spectral differential flat-field
correction operation to yield a plurality of spectral
differentials, wherein said plurality of spectral differentials
comprise AiAj spectral differential=(bright-field image data for an
Ai spectrum--dark-field image data for said Ai
spectrum)/(bright-field image data for an Aj spectrum--dark-field
image data for said Aj spectrum), Ai and Aj are two different
spectra, i is an index ranging from 1 to N, j is an index ranging
from 1 to N and N is the number of spectra; and (c) displaying said
spectral differentials in the form of a plurality of images to show
uniformity of each said image, wherein a non-uniform area on each
said image is determined to have been caused by an impurity of the
lens.
2. The method of claim 1, wherein said plurality of spectra
comprise R, G and B.
3. The method of claim 1, wherein said narrow-band multispectral
uniform surface light source comprises an RGB trichromatic narrow
band uniform surface light source, wherein said plurality of images
comprises six images comprising: (a) RG spectral differential=(R
spectral bright-field image data-R spectral dark-field image
data)/(G spectral bright-field image data-G spectral dark-field
image data); (b) GR spectral differential=(G spectral bright-field
image data-G spectral dark-field image data)/(R spectral
bright-field image data-R spectral dark-field image data); (c) RB
spectral differential=(R spectral bright-field image data-R
spectral dark-field image data)/(B spectral bright-field image
data-B spectral dark-field image data); (d) BR spectral
differential=(B spectral bright-field image data-B spectral
dark-field image data)/(R spectral bright-field image data-R
spectral dark-field image data); (e) BG spectral differential=(B
spectral bright-field image data-B spectral dark-field image
data)/(G spectral bright-field image data-G spectral dark-field
image data); and (f) GB spectral differential=(G spectral
bright-field image data-G spectral dark-field image data)/(B
spectral bright-field image data-B spectral dark-field image
data).
4. The method of claim 1, wherein said collecting step comprises
measuring the dark-field image data under each of said plurality of
spectra individually and obtaining the bright-field image data and
dark-field image data for each pixel under each of said plurality
of spectra.
Description
PRIORITY CLAIM AND RELATED APPLICATIONS
[0001] This non-provisional application claims the benefit of
priority from Chinese Pat. App. No. 2020110300154 filed on Sep. 27,
2020. Said application is incorporated by reference in its
entirety.
BACKGROUND OF THE INVENTION
1. The Field of the Invention
[0002] The present invention relates to a lens inspection method.
More specifically, the present invention is directed to a method
for detecting lens cleanliness using spectral differential flat
field correction.
2. Background Art
[0003] Cleanliness is an important indicator of an imaging system
and cleanliness is directly related to stray light, ghosting,
uniformity and other key imaging factors. In a lens manufacturing
process, the generation of surface defects is often unavoidable. In
general, surface defects are local physical or chemical properties
of product surface uneven areas, such as inclusions, damage,
stains, etc., all having adverse impacts on the cleanliness of the
product. Therefore, a lens manufacturer attaches great importance
to lens cleanliness inspection, through timely discovery of surface
defects of the lens, effective control of product quality, further
analysis and solution of problems in the production process,
thereby eliminating or reducing the generation of defective
products.
[0004] Finished lens are predominantly visually inspected. Such
method of inspection yields a low sampling rate and accuracy, is
real-time poor, inefficient and labor-intensive. These shortcomings
are further exacerbated by inspectors' work experience and skills
while machine vision-based inspection methods can largely overcome
the shortcomings.
[0005] Machine vision can be utilized in non-contact and
non-destructive automatic inspections, therefore making it an
effective means to achieve equipment automation, intelligence and
precision control, with advantages in safety, reliability, wide
spectral response ranges, reduction of long working hours in harsh
environments and high productivity. Machine vision includes an
image analysis module, a data management module and a human-machine
interface module. An image acquisition module can include a
charge-coupled device (CCD) camera, an optical lens, a light source
and its clamping device, etc. Its function is to complete the
acquisition of images of a product surface. Under the illumination
of a light source, a surface of a product is imaged on the camera
sensor through an optical lens and the light signal obtained of the
surface of the product is converted into an electrical signal,
which is then converted into a digital signal that can be processed
by a computer. Currently, industrial cameras are mainly based on
CCD or complementary metal oxide semiconductor (CMOS) chip
technology. CCD is currently the most commonly used image sensor
for machine vision. A light source directly affects image quality
and its role is to overcome ambient light interference, to ensure
image stability and result in images with the highest possible
contrast. Currently used light sources are halogen lamps,
fluorescent lamps and light-emitting diode (LED). An LED light
source is beneficial as it comes in a small form factor, is low in
power consumption, is fast in response time, is a good
light-emitting monochrome, is highly reliable, is a uniform and
stable light, is easy to integrate and is applicable to a wide
range of applications.
[0006] Illumination systems composed of light sources can be
divided into bright-field and dark-field illumination, structured
light illumination and stroboscopic illumination according to their
illumination methods. Since the bright-field signal itself carries
information about the relative illuminance of the large-angle field
of view, it can have a non-negligible effect on imaging. However,
the manner in which the influence of a low relative illuminance of
a large-angle field of view is suppressed, the manner in which the
observable range of the impurity to be detected is amplified and
the manner in which the detection efficiency of impurity is
effectively improved, are the emphases of current researches in the
field of lens inspection.
SUMMARY OF THE INVENTION
[0007] In accordance with the present invention, there is provided
a method for detecting lens cleanliness of a lens disposed in a
flat-field optical path, the flat-field optical path including a
light source, the lens, a camera, the light source is a narrow-band
multispectral uniform surface light source, the camera's
light-sensitive surface is disposed perpendicular to an optical
axis of the lens and in the light position of the lens, the method
including: [0008] (a) collecting the bright-field image data and
dark-field image data in a plurality of spectra through the lens;
[0009] (b) for each pixel, performing a spectral differential
flat-field correction operation to yield a plurality of spectral
differentials, wherein said plurality of spectral differentials
comprise AiAj spectral differential=(bright-field image data for an
Ai spectrum--dark-field image data for the Ai
spectrum)/(bright-field image data for an Aj spectrum--dark-field
image data for the Aj spectrum), Ai and Aj are two different
spectra, i is an index ranging from 1 to N, j is an index ranging
from 1 to N and N is the number of spectra; and [0010] (c)
displaying the spectral differentials in the form of a plurality of
images to show uniformity of each of the plurality of images,
wherein a non-uniform area on each of the plurality of images is
determined to have been caused by an impurity of the lens.
[0011] In one embodiment, the plurality of spectra include R, G and
B. In one embodiment, the narrow-band multispectral uniform surface
light source includes an RGB trichromatic narrow band uniform
surface light source, wherein the plurality of images includes six
images including: [0012] (a) RG spectral differential=(R spectral
bright-field image data-R spectral dark-field image data)/(G
spectral bright-field image data-G spectral dark-field image data);
[0013] (b) GR spectral differential=(G spectral bright-field image
data-G spectral dark-field image data)/(R spectral bright-field
image data-R spectral dark-field image data); [0014] (c) RB
spectral differential=(R spectral bright-field image data-R
spectral dark-field image data)/(B spectral bright-field image
data-B spectral dark-field image data); [0015] (d) BR spectral
differential=(B spectral bright-field image data-B spectral
dark-field image data)/(R spectral bright-field image data-R
spectral dark-field image data); [0016] (e) BG spectral
differential=(B spectral bright-field image data-B spectral
dark-field image data)/(G spectral bright-field image data-G
spectral dark-field image data); and [0017] (f) GB spectral
differential=(G spectral bright-field image data-G spectral
dark-field image data)/(B spectral bright-field image data-B
spectral dark-field image data).
[0018] In one embodiment, the collecting step includes measuring
the dark-field image data under each of the plurality of spectra
individually and obtaining the bright-field image data and
dark-field image data for each pixel under each of the plurality of
spectra.
[0019] An object of the present invention is to provide a method
for detecting lens cleanliness using spectral differential flat
field correction in order to effectively improve the efficiency of
impurity detection.
[0020] Whereas there may be many embodiments of the present
invention, each embodiment may meet one or more of the foregoing
recited objects in any combination. It is not intended that each
embodiment will necessarily meet each objective. Thus, having
broadly outlined the more important features of the present
invention in order that the detailed description thereof may be
better understood, and that the present contribution to the art may
be better appreciated, there are, of course, additional features of
the present invention that will be described herein and will form a
part of the subject matter of this specification.
BRIEF DESCRIPTION OF THE DRAWINGS
[0021] In order that the manner in which the above-recited and
other advantages and objects of the invention are obtained, a more
particular description of the invention briefly described above
will be rendered by reference to specific embodiments thereof which
are illustrated in the appended drawings. Understanding that these
drawings depict only typical embodiments of the invention and are
not therefore to be considered to be limiting of its scope, the
invention will be described and explained with additional
specificity and detail through the use of the accompanying drawings
in which:
[0022] FIG. 1 depicts a spectral differential flat-field correction
test system.
[0023] FIG. 2 depicts an image of the differential flat-field
correction results of the RGB trichromatic spectrum.
[0024] FIG. 3 depicts a combinatorial arrangement of spectral
differential flat-field correction.
PARTS LIST
[0025] 2--lens [0026] 4--camera [0027] 6--light source
DETAILED DESCRIPTION OF A PREFERRED EMBODIMENT
[0028] There is provided a method for detecting lens cleanliness
using spectral differential flat field correction. The method
includes building a flat-field test optical path, selecting a
plurality of spectra and measuring bright-field image data and
dark-field image data in different spectra and performing a
spectral differential flat-field correction operation on each
pixel. The spectrum of the flat-field test optical path is
configured to be adjustable. The flat-field test optical path
includes a light source, the lens under test and a monochrome
camera. The monochrome camera's light-sensitive surface is placed
perpendicular to the optical axis of the lens under test and panned
to the light position of the lens under test. Bright-field image
data is defined as the data obtained when the brightest value at
the center of the field of view is 80% to 90% of the saturation
value. Dark-field image data is defined as the data obtained when
there is no signal input. The camera exposure time is fixed and the
brightness of the light source is adjustable. The bright-field
image data and dark-field image data are collected for each pixel
under each spectrum. Subsequently, for each pixel, a spectral
differential flat-field correction operation is performed where
AiAj spectral differential=(bright-field image data for an Ai
spectrum--dark-field image data for the Ai spectrum)/(bright-field
image data for an Aj spectrum--dark-field image data for the Aj
spectrum), Ai and Aj are two different spectra, i is an index
ranging from 1 to N, j is an index ranging from 1 to N and N is the
number of spectra. The spectral differential flat-field correction
operation traverses spectrum A1, spectrum A2, spectrum A3, spectrum
AN-1, and spectrum AN. N.times.(N-1) combinations of spectral
differentials are obtained.
[0029] Preferably, the light source is selected as an RGB
trichromatic narrow band uniform surface light source and the
following flat-field correction operation is performed on the
bright-field image data and dark-field image data of each pixel in
the RGB trichromatic spectrum.
RG spectral differential=(R spectral bright-field image data-R
spectral dark-field image data)/(G spectral bright-field image
data-G spectral dark-field image data).
GR spectral differential=(G spectral bright-field image data-G
spectral dark-field image data)/(R spectral bright-field image
data-R spectral dark-field image data).
RB spectral differential=(R spectral bright-field image data-R
spectral dark-field image data)/(B spectral bright-field image
data-B spectral dark-field image data).
BR spectral differential=(B spectral bright-field image data-B
spectral dark-field image data)/(R spectral bright-field image
data-R spectral dark-field image data).
BG spectral differential=(B spectral bright-field image data-B
spectral dark-field image data)/(G spectral bright-field image
data-G spectral dark-field image data).
GB spectral differential=(G spectral bright-field image data-G
spectral dark-field image data)/(B spectral bright-field image
data-B spectral dark-field image data).
[0030] The six spectral differentials are displayed in the form of
six images for the determination of the lens cleanliness.
[0031] Compared to existing technology, the invention has the
following significant effects: As the plurality of spectra of the
narrow-band surface light source were measured independently and
the image data obtained of the plurality of spectra can suppress
the influence of large-angle field of view low relative
illumination. The signal-to-noise ratio of the large field of view
range is improved. Edge enhancement of impurity imaging is
obtained. The observable range of the impurity to be detected is
enlarged. After offsetting an impurity relative to the optical
center into symmetrical distributions with differential impurity
imaging, observable patterns of the impurity are then unified. As
the smallest resolvable size is the detector pixel size, the
spectral sensitivity of the impurity itself can be effectively
used.
[0032] The following is a detailed description of the technical
scheme of the present invention, taking the RGB trichromatic
spectrum as an example, together with the accompanying drawings.
There is provided a method for detecting the cleanliness of a lens
using spectral differential flat field correction, the method
including: [0033] (a) providing a flat-field test light path, the
light path including a homogeneous area light source with a narrow
band of the multicolor spectrum, a lens to be tested and a
monochrome camera that meets the resolution requirements. Camera
pixel size is the main factor that affects the accuracy of
detecting contaminant size. Camera resolution is the main factor
that affects the detection range. It is necessary to determine the
monochrome camera that meets the resolution requirement according
to the detection accuracy. In this example, the narrow-band uniform
surface light source uses an 8-inch multi-LED integrating sphere,
the multi-LED integrating sphere including RGB tricolor LEDs with
center wavelengths at 641 nm, 520 nm, 457 nm, and half-peak widths
of 20 nm, respectively. The RGB tricolor LEDs can be lit
independently. The lens is mounted onto a V-block tool. The lens is
disposed at a large field of view, e.g., 120*120 degrees view
angle. The camera is disposed in a manner where the camera's
light-sensitive surface is perpendicular to the optical axis of the
lens and the camera is translated to a position to detect light
through the lens. In this example, the camera uses a pixel size of
5.5 um and a pixel count of 8000*6000, placed at the rear focal
plane of the lens; [0034] (b) selecting a spectrum and measuring
the bright-field image data and dark-field image data required for
flat-field correction. The bright-field image data is the data
obtained when the brightest value at the center of the field of
view is 80% to 90% of the saturation value and the dark-field image
data is the data collected when there is no signal input. The
measuring step is performed with the exposure time fixed, the
brightness of the light source adjusted so that the camera output
falls within its range. The image data can then be collected.
Different spectral dark-field image data is collected separately to
improve data accuracy. In this example, the camera exposure time is
fixed at 50 ms, the integrating sphere is adjusted to output 641 nm
monochrome R light and the output is adjusted to 0 Nits, i.e., no
light output and the camera acquires images as dark-field image
data. The integrating sphere is adjusted to output 641 nm
monochrome R light at 50 Nits. The R light is adjusted such that
the camera center Region of Interest (ROI) of 1000*1000 pixels is
disposed at an average value of 80% of the maximum range. The
camera functions in a 12-bit mode, i.e., the average gray value of
about 3300, and the image is collected as bright-field image data;
[0035] (c) sequentially replacing the remaining spectra of the
light source, and repeating step (b) to collect the bright-field
image data and dark-field image data under each spectrum,
respectively. In general, dark-field image data for an optical
system is common. The present invention involves the switching of
light sources, and in order to eliminate potential negative
effects, a separate dark-field image data measurement is performed
for each light source switch in this example. In this example, the
measurement procedure for each of G, B light is similar to that for
R light. The integrating sphere is fixed at 50 ms exposure time.
The integrating sphere is adjusted to output 520 nm monochromatic G
light. Dark-field image data and bright-field image data are
collected. The integrating sphere is adjusted to output 457 nm
monochromatic B light. Again, dark-field image data and
bright-field image data are collected; and [0036] (d) performing
the flat-field correction operation on the bright-field image data
and dark-field image data of each spectrum.
[0037] Calculations for spectral differentials are as follows:
[0038] AiAj spectral differential=(bright-field image data for Ai
spectra--dark-field image data for Ai spectra)/(bright-field image
data for Aj spectra--dark-field image data for Aj spectra) where Ai
and Aj are two different spectra in spectra A1, spectra A2, spectra
A3, spectra AN-1 and spectra AN where i is an index ranging from 1
to N, j is an index ranging from 1 to N and N is the number of
spectra.
[0039] For spectra A1, spectra A2, spectra A3, spectra AN-1 and
spectra AN, there are up to N*(N-1) combinations of spectral
differentials as shown in FIG. 3. The results obtained from
different combinations of spectra vary depending on the physical
properties of the impurities, including but not limited to size,
three-dimensional shape, transmittance, refractive index, etc.
Therefore, traversing various combinations of spectra enriches the
detection information and improves the judgment efficiency. In this
example, six sets of data are obtained for the light-field and
dark-field of R, G, and B spectra, respectively, through steps (a)
through (c), i.e., each pixel of the camera has a corresponding six
data sets independent of other pixels and these data sets exist in
the form of a matrix. For each pixel P, a flat-field correction
operation is performed and the data processing for each pixel is
independent of each other, with no interaction between pixels. The
following six differential operations are obtained.
RG spectral differential=(R spectral bright-field image data-R
spectral dark-field image data)/(G spectral bright-field image
data-G spectral dark-field image data).
GR spectral differential=(G spectral bright-field image data-G
spectral dark-field image data)/(R spectral bright-field image
data-R spectral dark-field image data).
RB spectral differential=(R spectral bright-field image data-R
spectral dark-field image data)/(B spectral bright-field image
data-B spectral dark-field image data).
BR spectral differential=(B spectral bright-field image data-B
spectral dark-field image data)/(R spectral bright-field image
data-R spectral dark-field image data).
BG spectral differential=(B spectral bright-field image data-B
spectral dark-field image data)/(G spectral bright-field image
data-G spectral dark-field image data).
GB spectral differential=(G spectral bright-field image data-G
spectral dark-field image data)/(B spectral bright-field image
data-B spectral dark-field image data).
[0040] The resulting matrix of spectral differentials is displayed
directly in the form of an image for the determination of lens
cleanliness. As can be seen from the differentials, the
differential flat-field correction results are less affected by the
intensity distribution of the image itself, which can suppress the
effect of low relative illumination of the large-angle field of
view. The same impurity appears in two locations. At a first
location, a first data point appears smaller than the surrounding
pixels and the image is presented as a dark spot. At a second
location, a second data point appears larger than the surrounding
pixels and the image is presented as a bright spot. The impurity
information is extracted and enlarged in the positive and negative
directions. There is a uniform pattern of distribution where two
positions of the same impurity overlap one another and the edges of
the overlap show a clear contrast with an edge-enhancing
effect.
* * * * *