FORM 2
THE PATENTS ACT, 1970
(39 of 1970)
&
THE PATENTS RULES, 2003
COMPLETE SPECIFICATION
(See section 10, rule 13)
“DEVICE AND METHOD FOR INSPECTING BATTERY
ELECTRODE”
LG ENERGY SOLUTION, LTD. of Tower1 108, Yeoui-daero
Yeongdeungpo-gu Seoul 07335, KR;
The following specification particularly describes the invention and the
manner in which it is to be performed.
- 2 -
[DESCRIPTION]
[Invention Title]
DEVICE AND METHOD FOR INSPECTING BATTERY ELECTRODE
This application claims priority to and the benefit
of Korean Patent Application No. 10-2022-0109765 filed in5
the Korean Intellectual Property Office on August 31, 2022
and Korean Patent Application No.10-2023-0103916 filed in
the Korean Intellectual Property Office on August 9, 2023,
the entire contents of which are incorporated herein by
reference.10
[Technical Field]
The present invention relates to an device and method
for inspecting electrodes of batteries, and more
particularly, to an device and method for inspecting
electrodes of batteries, which determines whether the15
electrode is defective due to surface defects from
electrode images using a pre-trained learning model.
[Background Art]
As the price of energy sources rises due to depletion
of fossil fuels and interest in environmental pollution20
increases, a demand for secondary batteries as an eco-
friendly alternative energy source is rapidly increasing.
Lithium secondary batteries among secondary batteries
are being applied to various industrial fields such as from
mobile application devices to automobiles, robots, and25
- 3 -
energy storage devices, as a response to today's
environmental regulations and high oil price issues.
These lithium batteries are generally classified into
cylindrical, prismatic, or pouch types depending on the
shape of the exterior material in which the electrode5
assembly is accommodated.
Among these, cylindrical batteries may be provided in
a cell-to-pack (CTP) structure consisting of a plurality of
battery cells. In other words, a cylindrical battery is
provided in a structure which is assembled by winding10
electrodes including a separator between an anode and a
cathode and inserting it inside a battery can.
In the electrode preparation process of a cylindrical
battery, a battery electrode inspection device is be used
to detect defects on electrode surfaces caused by foreign15
substances, scratches, etc.
In general, a battery electrode inspection device
analyzes electrode images obtained from cameras installed
in a process equipment to determine whether there exists
any defect in the electrode surface. Accordingly, in a20
battery electrode inspection device, result data may be
affected by field situation or environment, such as shaking
of equipment, changes in process conditions, etc.
Accordingly, a conventional battery electrode
inspection device used to apply program logic designed25
- 4 -
based on defect images obtained from a camera installed in
the field during initial application of the process.
Accordingly, the conventional battery electrode
inspection device has the disadvantage of requiring a long
time for initial application of the process.5
In addition, the conventional battery electrode
inspection device had the disadvantage that the program
logic needs to be continuously redesigned by an engineer
when any change occurs in the field situation or
environment.10
[Detailed Description of the Invention]
[Technical Problem]
To obviate one or more problems of the related art,
embodiments of the present disclosure provide a system for
inspecting electrodes of batteries.15
To obviate one or more problems of the related art,
embodiments of the present disclosure also provide an
device for inspecting electrodes of batteries.
To obviate one or more problems of the related art,
embodiments of the present disclosure also provide a method20
for inspecting electrodes of batteries.
[Technical Solution]
In order to achieve the objective of the present
disclosure, an device for inspecting battery electrodes
based on artificial intelligence model may include at25
- 5 -
least one processor and a memory configured to store at
least one instruction executed by the at least one
processor, and the at least one instruction may include an
instruction to obtain an image of an inspection object in
which an area suspected of being defective is extracted5
from an electrode image in which at least one electrode
surface is captured; and an instruction to output result
data on whether the battery electrode is defective by
inputting the image of the inspection object into a pre-
trained learning model.10
Here, the instruction to output result data may
include an instruction to determine any one learning model
among a plurality of learning models according to an amount
of training data.
The instruction to determine any one learning model15
may include an instruction to, upon the amount of training
data being less than a predetermined reference value,
output the result data using a pre-trained machine learning
based first learning model.
The instruction to determine any one learning model20
may include an instruction to, upon the amount of training
data being greater than or equal to the predetermined
reference value, output the result data using a pre-trained
deep learning based second learning model.
The instruction to output the result data using the25
- 6 -
first learning model may include an instruction to extract
at least one image feature value from the image of the
inspection object and an instruction to obtain the result
data by inputting the at least one image feature value into
the first learning model.5
The instruction to extract at least one image feature
value from the image of the inspection object may include
an instruction to extract the at least one image feature
value from the image of the inspection object using a rule-
based algorithm.10
The at least one image feature value may include data
regarding one or more of pixel height, pixel width, pixel
maximum value (Peak White), pixel minimum value (Peak Dark),
aspect ratio, and roundness which are extracted from the
image of the inspection object.15
The first learning model may be a random forest-based
learning model.
The instruction to output the result data using the
pre-trained deep learning based second learning model may
include an instruction to obtain the result data by20
inputting the at least one image of inspection object into
the second learning model.
The second learning model may be a convolutional
neural network (CNN) based learning model.
The at least one instruction may further include an25
- 7 -
instruction to re-train the pre-trained learning model by
using the result data as learning data.
According to another embodiment of the present
disclosure, a method for inspecting battery electrodes,5
using an device for inspecting battery electrodes based on
artificial intelligence model, may include obtaining an
image of an inspection object in which an area suspected
of being defective is extracted from an electrode image in
which at least one electrode surface is captured; and10
outputing result data on whether the battery electrode is
defective by inputting the image of the inspection object
into a pre-trained learning model.
Here, the outputing result data on whether the
battery electrode is defective includes determining any one15
learning model among a plurality of learning models
according to an amount of at least one training data.
The determining any one learning model among the
plurality of learning models may include, upon the amount
of training data being less than a predetermined reference20
value, outputting the result data using a pre-trained
machine learning based first learning model.
The determining any one learning model among the
plurality of learning models may include, upon the amount
of training data being greater than or equal to the25
- 8 -
predetermined reference value, outputting the result data
using a pre-trained deep learning based second learning
model.
The outputting the result data using the first
learning model may include extracting at least one image5
feature value from the image of the inspection object; and
obtaining the result data by inputting the at least one
image feature value into the first learning model.
The extracting at least one image feature value from
the image of the inspection object may include extracting10
the at least one image feature value from the image of the
inspection object using a rule-based algorithm.
The at least one image feature value may include data
regarding one or more of pixel height, pixel width, pixel
maximum value (Peak White), pixel minimum value (Peak Dark),15
aspect ratio, and roundness which are extracted from the
image of the inspection object.
The first learning model may be a random forest-based
learning model.
The outputting the result data using the pre-trained20
deep learning based second learning model may include
obtaining the result data by inputting the at least one
image of inspection object into the second learning model.
The second learning model may be a convolutional
neural network (CNN) based learning model.25
- 9 -
The method may further include re-training the pre-
trained learning model by using the result data as learning
data.
According to another embodiment of the present
disclosure, a system for inspecting battery electrodes5
based on artificial intelligence model may include a
camera configured to produce an electrode image by
capturing a surface of at least one electrode; and a
battery electrode inspecting device configured to obtain
an image of an inspection object in which an area10
suspected of being defective is extracted from the
electrode image, input the image of the inspection object
into a pre-trained learning model, and output result data
on whether the battery electrode is defective.
[Advantageous Effects]15
The battery electrode inspection device and method,
and the system including the battery electrode inspection
device according to embodiments of the present invention
may inspect whether an inspection object image is defective
by applying different types of learning models according to20
amount of learning data and re-train the learning model
based on learning data which is updated with result data,
thereby obtaining improved reliability of inspection.
[Brief Description of the Drawings]
FIG. 1 is a conceptual diagram of a process of25
- 10 -
applicating detection logic in a general battery electrode
inspection device.
FIG. 2 is a conceptual diagram of operations of a
general battery electrode inspection device when an error
occurs.5
FIG. 3 is a block diagram of a battery electrode
inspection system according to embodiments of the present
invention.
FIG. 4 is a block diagram of a battery electrode
inspection device according to embodiments of the present10
invention.
FIG. 5 is an operational flowchart of a battery
electrode inspection method using the battery electrode
inspection device according to embodiments of the present
invention.15
FIG. 6 is an image for explaining the step of
extracting inspection object images in the battery
electrode inspection method according to embodiments of the
present invention.
FIG. 7 is an operational flowchart for explaining the20
step of determining whether an electrode is faulty in the
battery electrode inspection method according to
embodiments of the present invention.
FIG. 8 is a conceptual diagram for explaining a
battery electrode inspection method using a first learning25
- 11 -
model of the battery electrode inspection method according
to an embodiment of the present invention.
FIG. 9 is a conceptual diagram illustrating a battery
electrode inspection method using a second learning model
of the battery electrode inspection method according to5
another embodiment of the present invention.
FIG. 10 is a conceptual diagram for explaining the
re-learning step in the battery electrode inspection method
according to another embodiment of the present invention.
[Description of reference numbers]10
1000: camera 5000: battery electrode inspection
device
100: memory 200: processor
300: transceiver device 400: input interface device
500: output interface device600: storage device15
700: bus
[Best Modes for Practicing the Disclosure]
The present invention may be modified in various
forms and have various embodiments, and specific
embodiments thereof are shown by way of example in the20
drawings and will be described in detail below. It should
be understood, however, that there is no intent to limit
the present invention to the specific embodiments, but on
the contrary, the present invention is to cover all
modifications, equivalents, and alternatives falling within25
- 12 -
the spirit and technical scope of the present invention.
Like reference numerals refer to like elements throughout
the description of the figures.
It will be understood that, although the terms such
as first, second, A, B, and the like may be used herein to5
describe various elements, these elements should not be
limited by these terms. These terms are only used to
distinguish one element from another. For example, a first
element could be termed a second element, and, similarly, a
second element could be termed a first element, without10
departing from the scope of the present invention. As used
herein, the term "and/or" includes combinations of a
plurality of associated listed items or any of the
plurality of associated listed items.
It will be understood that when an element is15
referred to as being "coupled" or "connected" to another
element, it can be directly coupled or connected to the
other element or an intervening element may be present. In
contrast, when an element is referred to as being "directly
coupled" or "directly connected" to another element, there20
is no intervening element present.
Terms used in the present application are used only
to describe specific embodiments, and are not intended to
limit the present invention. A singular form includes a
plural form if there is no clearly opposite meaning in the25
- 13 -
context. In the present application, it should be
understood that the term “include” or “have” indicates that
a feature, a number, a step, an operation, a component, a
part or the combination thereof described in the
specification is present, but does not exclude a5
possibility of presence or addition of one or more other
features, numbers, steps, operations, components, parts or
combinations thereof, in advance.
Unless otherwise defined, all terms used herein,
including technical and scientific terms, have the same10
meanings as commonly understood by one skilled in the art
to which the present invention belongs. It will be further
understood that terms, such as those defined in commonly
used dictionaries, should be interpreted as having meanings
that are consistent with their meanings in the context of15
the relevant art and will not be interpreted in an
idealized or overly formal sense unless expressly so
defined herein.
Hereinafter, exemplary embodiments of the present
invention will be described in detail with reference to the20
accompanying drawings.
FIG. 1 is a conceptual diagram of a process of
applicating detection logic in a general battery electrode
inspection device.25
- 14 -
Referring to FIG. 1, a battery electrode inspection
device can detect defects on the surface of electrodes and
determine whether the electrodes are faulty.
A general battery electrode inspection device uses
pre-designed detection logic to determine a defect type and5
fault of at least one electrode image obtained from a
camera installed in the field. Here, the detection logic
may be logic previously designed by an engineer using image
features of at least one electrode image for which a defect
type has been confirmed.10
FIG. 2 is a conceptual diagram of operations of a
general battery electrode inspection device when an error
occurs.
Referring to FIG. 2, the detection logic of a general15
battery electrode inspection device is designed based on a
defect image of the electrode obtained from a camera
installed in the field during initial application of the
process. Accordingly, the general battery electrode
inspection device has the disadvantage that when at least20
one of field environmental factors or process conditions
changes, a detection error can occur due to distortion or
deformation of electrode images, and thus, precision of
defect detection can decreases. Accordingly, the general
electrode detection device has the disadvantage that the25
- 15 -
detection logic should be continuously modified by an
engineer.
In addition, when modifying the detection logic to
prevent this problem, the detection logic needs to be
modified by acquiring defect images generated from cameras5
installed in the field to take the process factors
described above into account, which has the disadvantage of
taking a long time.
In addition, a general battery electrode inspection
device has a disadvantage in that when the detection logic10
is modified, it is difficult to determine classification
accuracy of at least one defect image for which inspection
has been completed before modification. Here, the defect
image may be an image of an electrode that is determined to
be faulty due to occurrence of a defect as a result of the15
inspection of the inspection logic.
Accordingly, the present invention will describe an
electrode detection device and method that extracts an
image of an inspection object suspected of having a surface
defect of an electrode from an electrode image using a pre-20
trained learning model and determines whether the electrode
is faulty based on this.
FIG. 3 is a block diagram of a battery electrode
inspection system according to embodiments of the present25
- 16 -
invention.
Referring to FIG. 3, the battery electrode inspection
system may detect electrodes with surface defects during an
electrode preparation process and determine whether the
electrode is faulty.5
In more detail, the battery electrode inspection
system may include a camera 1000 and a battery electrode
inspection device 5000.
The camera 1000 may be installed at a site where an
electrode preparation process is performed. For example,10
the camera 1000 may be fixedly installed in an equipment in
which at least one electrode is sequentially moved.
Accordingly, the camera 1000 may acquire at least one
electrode image by individually photographing the at least
one electrode.15
The battery electrode inspection device 5000 may be
connected with at least one camera 1000 to obtain at least
one electrode image from the camera 1000. However, the
scope of the present invention is not limited to this, but
the battery electrode inspection device 5000 may acquire at20
least one electrode image through various ways other than a
camera.
Thereafter, the battery electrode inspection device
5000 may determine whether the battery electrode is faulty
based on at least one electrode image using a pre-trained25
- 17 -
learning model.
The battery electrode inspection device 5000 will be
described in more detail in the follwing.
FIG. 4 is a block diagram of a battery electrode5
inspection device according to embodiments of the present
invention.
Referring to FIG. 4, as described above, the battery
electrode inspection device 5000 according to embodiments
of the present invention may extract at least one image of10
an inspection object area which is suspected of being
defective from at least one electrode image and determine
whether the electrode is defective based on the extracted
image of inspection using a pre-trained learning model.
Here, the defects may include foreign substances, scratches,15
cracks, etc.
Describing the battery electrode inspection device in
detail by configuration, the battery electrode inspection
device 5000 may include a memory 100, a processor 200, a
transceiver device 300, an input interface device 400, an20
output interface device 500, and a storage device 600.
According to embodiments, respective components 100,
200, 300, 400, 500, and 600 included in the battery
electrode inspection device are connected by a bus 700 and
can communicate with each other.25
- 18 -
Among the components 100, 200, 300, 400, 500, and 600,
the memory 100 and the storage device 600 may include at
least one of a volatile storage medium and a non-volatile
storage medium. For example, the memory 100 and the
storage device 600 may include at least one of read only5
memory (ROM) and random access memory (RAM).
Among these, the memory 100 may include at least one
instruction executed by the processor 200.
According to embodiments, the at least one
instruction may include an instruction to obtain an image10
of an inspection object in which an area suspected of being
defective is extracted from an electrode image in which at
least one electrode surface is captured; and an instruction
to output result data on whether the battery electrode is
defective by inputting the image of the inspection object15
into a pre-trained learning model.
Here, the instruction to output result data may
include an instruction to determine any one learning model
among a plurality of learning models according to an amount
of at least one training data.20
The instruction to determine any one learning model
may include an instruction to, upon the amount of training
data being less than a predetermined reference value,
output the result data using a pre-trained machine learning
based first learning model.25
- 19 -
The instruction to determine any one learning model
may include an instruction to, upon the amount of training
data being greater than or equal to the predetermined
reference value, output the result data using a pre-trained
deep learning based second learning model.5
The instruction to output the result data using the
first learning model may include an instruction to extract
at least one image feature value from the image of the
inspection object and an instruction to obtain the result
data by inputting the at least one image feature value into10
the first learning model.
The instruction to extract at least one image feature
value from the image of the inspection object may include
an instruction to extract the at least one image feature
value from the image of the inspection object using a rule-15
based algorithm.
The at least one image feature value may include data
regarding one or more of pixel height, pixel width, pixel
maximum value (Peak White), pixel minimum value (Peak Dark),
aspect ratio, and roundness which are extracted from the20
image of the inspection object.
The first learning model may be a random forest-based
learning model.
The instruction to output the result data using the
pre-trained deep learning based second learning model may25
- 20 -
include an instruction to obtain the result data by
inputting the at least one image of inspection object into
the second learning model.
The second learning model may be a convolutional
neural network (CNN) based learning model5
The at least one instruction may further include an
instruction to re-train the pre-trained learning model by
using the result data as learning data.
Meanwhile, the processor 200 may refer to a central
processing unit (CPU), a graphics processing unit (GPU), or10
a dedicated processor on which methods according to
embodiments of the present invention are performed.
As described above, the processor 200 may execute at
least one program command stored in the memory 100.
15
FIG. 5 is an operational flowchart of a battery
electrode inspection method using the battery electrode
inspection device according to embodiments of the present
invention.
Referring to FIG. 5, the battery electrode inspection20
device 5000 may obtain at least one electrode image (S1000).
Thereafter, the battery electrode inspection device
5000 may extract at least one inspection object image from
at least one electrode image (S3000). Here, the electrode
image may be an image taken of an electrode surface by at25
- 21 -
least one camera 1000 mounted on a electrode preparation
process equipment.
In addition, the inspection object image may be an
image in which at least one area in which a defect occurs
in at least one electrode image is individually extracted.5
Here, the defect may include, for example, scratches,
foreign matter, deterioration, etc., but is not limited to
these.
Thereafter, the battery electrode inspection device
5000 may determine whether the electrode is faulty based on10
the inspection object image using a pre-trained learning
model (S5000).
Thereafter, the battery electrode inspection device
5000 may re-train the learning model based on at least one
updated learning data (S7000).15
FIG. 6 is an image for explaining the step of
extracting inspection object images in the battery
electrode inspection method according to embodiments of the
present invention.20
Referring to FIG. 6, the inspection object image may
be obtained using a rule-based algorithm. In other words,
the battery electrode inspection device 5000 may output
inspection object images by inputting electrode images and
a predefined condition command into a rule-based algorithm.25
- 22 -
Here, the rule based algorithm may calculate an
average brightness value of the electrode image according
to a predefined condition command, select at least one
pixel that differs by more than a certain range based on
the calculated average brightness, and extract a certain5
area including the corresponding pixels, thereby obtaining
and outputting inspection object images. For example, the
inspection object image may be a binarized image and may be
provided in a size of 128 Pixel x 128 Pixel.
10
FIG. 7 is an operational flowchart for explaining the
step of determining whether an electrode is faulty in the
battery electrode inspection method according to
embodiments of the present invention.
Referring to FIG. 7, the battery electrode inspection15
device 5000 according to embodiments of the present
invention may check an amount of learning data which is
classified by a data set in the memory 100 (S5100). Here,
the at least one learning data may include at least one
inspection result data pre-inspected by a battery electrode20
inspection device 5000. In other words, the learning data
may include fault data or normal data. For example, fault
data may be classified and stored according to type of
defects within the data set.
Accordingly, the battery electrode inspection device25
- 23 -
5000 may select a type of pre-trained learning model.
According to an embodiment, the battery electrode
inspection device 5000 may select a pre-trained first
learning model when the amount of learning data is less
than a predetermined reference value (S5300).5
Afterwards, the battery electrode inspection device
5000 may extract image feature values from the inspection
object images (S5310).
Thereafter, the battery electrode inspection device
5000 may determine whether the battery electrode is faulty10
using the pre-trained first learning model (S5350).
Meanwhile, according to another embodiment, the
battery electrode inspection device 5000 may select a pre-
trained second learning model when the amount of learning
data is greater than or equal to the predetermined15
reference value (S5500).
Thereafter, the battery electrode inspection device
5000 may determine whether the battery electrode is faulty
using the selected second learning model (S5700).
20
FIG. 8 is a conceptual diagram for explaining a
battery electrode inspection method using a first learning
model of the battery electrode inspection method according
to an embodiment of the present invention.
Referring to FIG. 8, as described above, the battery25
- 24 -
electrode inspection device 5000 may select the pre-trained
first learning model when the amount of learning data is
less than the predetermined reference value.
Thereafter, the battery electrode inspection device
5000 may extract image feature values from the inspection5
object images.
Here, the image feature value may be at least one
pixel information extracted from the inspection object
images. For example, the image feature value may include
one or more of pixel height, pixel width, pixel maximum10
value (Peak White), pixel minimum value (Peak Dark), aspect
ratio, and roundness.
According to an embodiment, the battery electrode
inspection device 5000 may obtain image feature values from
at least one inspection object image using a rule-based15
algorithm when the amount of learning data is less than a
the predetermined reference value.
Thereafter, the battery electrode inspection device
5000 may determine whether the battery electrode is faulty
using the pre-trained first learning model.20
To be more specific, the battery electrode inspection
device 5000 may receive image feature values extracted from
an image to be inspected as input data to a pre-trained
first learning model. Thereafter, the battery electrode
inspection device 5000 may output result data on whether25
- 25 -
the battery electrode is faulty based on the inspection
object image. Here, the battery electrode may be an
electrode corresponding to the inspection object image.
According to an embodiment, the first pre-trained
learning model may be a machine learning-based algorithm.5
For example, the first learning model may be a bagging-type
random forest that is pre-trained based on at least one
learning data which is pre-classified in the data set of
the memory 100.
10
FIG. 9 is a conceptual diagram illustrating a battery
electrode inspection method using a second learning model
of the battery electrode inspection method according to
another embodiment of the present invention.
Referring to FIG. 9, the battery electrode inspection15
device 5000, as described above, may determine whether the
battery electrode is defective using the pre-trained second
learning model when the amount of learning data is equal to
or more than the predetermined reference value.
In more detail, the battery electrode inspection20
device 5000 may receive an inspection object image as input
data to the pre-trained second learning model. Accordingly,
the battery electrode inspection device 5000 may output
data as a result of determining whether the battery
electrode is faulty based on the inspection object image.25
- 26 -
According to an embodiment, the pre-trained second
learning model may be a pre-trained learning model based on
at least one learning data which is pre-classified in the
data set in the memory 100.
To be more specific, the pre-trained second learning5
model may be an artificial intelligence (AI)-based
classification model. In other words, the second learning
model may be a deep-learning based classification model
which is pre-trained according to pre-stored learning data
based on the result data of the battery electrode10
inspection device 5000.
According to an embodiment, the pre-trained second
learning model may be a learning model based on an
artificial neural network of convolutional neural network
(CNN). For example, the battery electrode inspection15
device 5000 may pre-train the second learning model so that
the second learning model extracts image feature data of
the inspection object image using the convolutional layer,
reduces the feature data to a lower dimension through the
pooling layer, classifies it according to the defect type20
through the fully connected layer, and determines whether
the battery electrode is faulty.
FIG. 10 is a conceptual diagram for explaining the
re-learning step in the battery electrode inspection method25
- 27 -
according to another embodiment of the present invention.
Referring to FIG. 10, the battery electrode
inspection device 5000 may store the result data obtained
from the embodiments of the present invention, that is, the
result data of determining whether the electrode is5
defective, in the memory 100 and classify it into a data
set.
Thereafter, the battery electrode inspection device
5000 may re-train the learning model based on at least one
updated learning data. In other words, the battery10
electrode inspection device 5000 may re-train the learning
model by using previously learned result data as learning
data. Accordingly, the amount of learning data increases,
and thus, the inspection reliability of the battery
electrode inspection device 5000 can be improved.15
The battery electrode inspection device and method
according to embodiments of the present invention have been
described above.
The battery electrode inspection device and method20
according to embodiments of the present invention may
extract an inspection object image including an area
suspected of being defective based on an electrode image
obtained from a camera, determine and apply different types
of learning models for determining whether an electrode25
- 28 -
corresponding to the inspection object image is defective
according to the amount of learning data, thereby
shortening application time due to initial process
application and changes in the process environment.
In addition, the battery electrode inspection device5
and method can obtain highly reliable result data with
improved precision by re-training the learning model based
on updated learning data.
The operations of the method according to the10
embodiments of the present invention may be implemented as
a computer-readable program or code on a computer-readable
recording medium. The computer-readable recording medium
includes all types of recording devices in which data
readable by a computer system is stored. In addition, the15
computer-readable recording medium may be distributed in a
network-connected computer system to store and execute
computer-readable programs or codes in a distributed manner.
In addition, the computer-readable recording medium
may include hardware devices specially configured to store20
and execute program instructions, such as ROM, RAM, flash
memory, etc. The program instructions may include not only
machine language code created by a compiler, but also high-
level language code that can be executed by a computer
using an interpreter.25
- 29 -
Although some aspects of the invention have been
described in the context of the apparatus, it may also
represent a description according to a corresponding method,
wherein a block or apparatus corresponds to a method step
or feature of a method step. Similarly, aspects described5
in the context of a method may also represent a feature of
a corresponding block or item or a corresponding apparatus.
Some or all of the method steps may be performed by (or
using) a hardware device, such as, for example, a
microprocessor, a programmable computer, or an electronic10
circuit. In some embodiments, one or more of the most
important method steps may be performed by such an
apparatus.
In the forgoing, the present invention has been
described with reference to the exemplary embodiment of the15
present invention, but those skilled in the art may
appreciate that the present invention may be variously
corrected and changed within the range without departing
from the spirit and the area of the present invention
described in the appending claims.20
- 30 -
[CLAIMS]
[Claim 1]
A battery electrode inspection device based on
artificial intelligence model, the battery electrode
inspection device comprising:5
at least one processor; and
a memory configured to store at least one instruction
executed by the at least one processor,
wherein the at least one instruction includes:
an instruction to obtain an image of an inspection10
object in which an area suspected of being defective is
extracted from an electrode image in which at least one
electrode surface is captured; and
an instruction to output result data on whether the
battery electrode is defective by inputting the image of15
the inspection object into a pre-trained learning model.
[Claim 2]
The battery electrode inspection device of claim 1,
wherein the instruction to output the result data includes20
an instruction to determine any one learning model among a
plurality of learning models according to an amount of at
least one training data.
[Claim 3]25
- 31 -
The battery electrode inspection device of claim 2,
wherein the instruction to determine any one learning model
includes:
an instruction to, upon the amount of training data
being less than a predetermined reference value, output the5
result data using a pre-trained machine learning based
first learning model.
[Claim 4]
The battery electrode inspection device of claim 3,10
wherein the instruction to determine any one learning model
includes:
an instruction to, upon the amount of training data
being greater than or equal to the predetermined reference
value, output the result data using a pre-trained deep15
learning based second learning model.
[Claim 5]
The battery electrode inspection device of claim 3,
wherein the instruction to output the result data using the20
first learning model includes:
an instruction to extract at least one image feature
value from the image of the inspection object; and
an instruction to obtain the result data by inputting
the at least one image feature value into the first25
- 32 -
learning model.
[Claim 6]
The battery electrode inspection device of claim 5,
wherein the instruction to extract at least one image5
feature value from the image of the inspection object
includes an instruction to extract the at least one image
feature value from the image of the inspection object using
a rule-based algorithm.
10
[Claim 7]
The battery electrode inspection device of claim 5,
wherein the at least one image feature value includes data
regarding one or more of pixel height, pixel width, pixel
maximum value Peak White, pixel minimum value Peak Dark,15
aspect ratio, and roundness which are extracted from the
image of the inspection object.
[Claim 8]
The battery electrode inspection device of claim 3,20
wherein the first learning model is a random forest-based
learning model.
[Claim 9]
The battery electrode inspection device of claim 4,25
- 33 -
wherein the instruction to output the result data using the
pre-trained deep learning based second learning model
includes an instruction to obtain the result data by
inputting the at least one image of inspection object into
the second learning model.5
[Claim 10]
The battery electrode inspection device of claim 4,
wherein the second learning model is a convolutional neural
network CNN based learning model.10
[Claim 11]
The battery electrode inspection device of claim 1,
wherein the at least one instruction further includes an
instruction to re-train the pre-trained learning model by15
using the result data as learning data.
[Claim 12]
A battery electrode inspection method using a battery
electrode inspection device based on artificial20
intelligence model, the battery electrode inspection
method comprising:
obtaining an image of an inspection object in which
an area suspected of being defective is extracted from an
electrode image in which at least one electrode surface is25
- 34 -
captured; and
outputing result data on whether the battery
electrode is defective by inputting the image of the
inspection object into a pre-trained learning model.
5
[Claim 13]
The battery electrode inspection method of claim 12,
wherein the outputing the result data on whether the
battery electrode is defective includes determining any one
learning model among a plurality of learning models10
according to an amount of at least one training data.
[Claim 14]
The battery electrode inspection method of claim 13,
wherein the determining any one learning model among the15
plurality of learning models includes:
upon the amount of training data being less than a
predetermined reference value, outputting the result data
using a pre-trained machine learning based first learning
model.20
[Claim 15]
The battery electrode inspection method of claim 14,
wherein the determining any one learning model among the
plurality of learning models includes:25
- 35 -
upon the amount of training data being greater than
or equal to the predetermined reference value, outputting
the result data using a pre-trained deep learning based
second learning model.
5
[Claim 16]
The battery electrode inspection method of claim 14,
wherein the outputting the result data using the first
learning model includes:
extracting at least one image feature value from the10
image of the inspection object; and
obtaining the result data by inputting the at least
one image feature value into the first learning model.
[Claim 17]15
The battery electrode inspection method of claim 16,
wherein the extracting at least one image feature value
from the image of the inspection object includes extracting
the at least one image feature value from the image of the
inspection object using a rule-based algorithm.20
[Claim 18]
The battery electrode inspection method of claim 16,
wherein the at least one image feature value includes data
regarding one or more of pixel height, pixel width, pixel25
- 36 -
maximum value Peak White, pixel minimum value Peak Dark,
aspect ratio, and roundness which are extracted from the
image of the inspection object.
[Claim 19]5
The battery electrode inspection method of claim 14,
wherein the first learning model is a random forest-based
learning model.
[Claim 20]10
The battery electrode inspection method of claim 15,
wherein the outputting the result data using the pre-
trained deep learning based second learning model includes
obtaining the result data by inputting the at least one
image of inspection object into the second learning model.15
[Claim 21]
The battery electrode inspection method of claim 15,
wherein the second learning model is a convolutional neural
network CNN based learning model.20
[Claim 22]
The battery electrode inspection method of claim 12,
further comprising re-training the pre-trained learning
model by using the result data as learning data.25
- 37 -
[Claim 23]
A battery electrode inspection system based on
artificial intelligence model, the battery electrode
inspection system comprising:5
a camera configured to produce an electrode image by
capturing a surface of at least one electrode; and
a battery electrode inspecting device configured to
obtain an image of an inspection object in which an area
suspected of being defective is extracted from the10
electrode image, input the image of the inspection object
into a pre-trained learning model, and output result data
on whether the battery electrode is defective.