Battery appearance inspection device and battery sorting system

Abstract

A battery appearance inspection device according to one aspect of the present invention comprises: a lighting unit configured to change lighting conditions, including at least one of the direction, angle, or intensity of light, at intervals of time while shining the light onto a battery; an image generation unit configured to capture a target portion of the battery on the basis of the light emitted by the lighting unit, and thereby generate a plurality of images of the target portion captured under different lighting conditions; an image processing unit configured to generate at least one composite image having depth information about the target portion by synthesizing at least two images among the plurality of images; a first inspection unit configured to determine whether the battery is defective on the basis of at least one image among the plurality of images; and a second inspection unit configured to determine whether the battery is non-defective on the basis of the at least one composite image.

Description

Battery appearance inspection device and battery sorting system

[0001] This application claims priority from Korean Patent Application No. 10-2024-0082154, filed June 24, 2024, the entire disclosure of which is incorporated herein by reference.

[0002] The present invention relates to a battery appearance inspection device and a battery sorting system, and more particularly, to a battery appearance inspection device that inspects the appearance of a battery based on an image, and a battery sorting system including the battery appearance inspection device.

[0003] Generally, secondary batteries refer to batteries that can be repeatedly charged and discharged, such as lithium-ion batteries, lithium polymer batteries, nickel-cadmium batteries, nickel-metal hydride batteries, and nickel-zinc batteries. The most basic secondary battery, a battery cell, can provide an output voltage of approximately 2.5 V to 4.2 V.

[0004] Recently, these secondary batteries have been widely used in small wireless devices such as cell phones, tablet PCs, and laptops, as well as in devices requiring high output voltage and large charging capacity, such as electric vehicles and Energy Storage Systems (ESS). This has led to the mass production of various types of batteries. Consequently, interest in and demand for technologies capable of quickly and accurately inspecting the appearance of assembled batteries is growing.

[0005] However, in mass-production systems for batteries, the method of having workers visually inspect the exterior of assembled batteries presents a significant challenge: it requires significant time and money. Furthermore, existing vision inspection methods, designed to address these issues, can result in false negatives (defective batteries being judged as good) or false positives (good batteries being judged as defective) due to changes in the direction, angle, or intensity of light irradiating the battery under inspection, image recognition errors, and the emergence of new types of defects.

[0006] The technical problem to be solved by the present invention is to provide a battery appearance inspection device that reduces the time and cost required for battery appearance inspection while improving the accuracy and reliability of the inspection results, and a battery sorting system including the battery appearance inspection device.

[0007] According to one aspect of the present invention, a battery appearance inspection device includes: an illumination unit configured to change an illumination condition including at least one of a direction, an angle, and an intensity of the irradiated light at intervals of time, as an illumination unit for irradiating light onto a battery; an image generation unit configured to capture a target portion of the battery based on the light irradiated by the illumination unit and generate a plurality of images of the target portion, each of the plurality of images captured under different illumination conditions; an image processing unit configured to synthesize at least two images among the plurality of images to generate at least one composite image having depth information of the target portion; a first inspection unit configured to determine whether the battery is defective based on at least one image among the plurality of images; and a second inspection unit configured to determine whether the battery is good based on the at least one composite image.

[0008] In one embodiment, the lighting unit may include a plurality of light sources arranged at different locations and configured to irradiate light toward the battery at different times.

[0009] In one embodiment, the image generating unit may include a plurality of cameras configured to photograph different target portions of the battery, each at a different location.

[0010] In one embodiment, the plurality of cameras may include a first camera that photographs a first target portion located at an upper portion of the battery; a second camera that photographs a second target portion located at a lower portion of the battery; and at least one third camera that photographs a third target portion located at a side of the battery.

[0011] In one embodiment, the image processing unit may be configured to generate a first image matrix by arranging at least two images among the plurality of images, and the first inspection unit may be configured to determine whether the battery is defective based on the first image matrix.

[0012] In one embodiment, the image processing unit may be configured to generate a second image matrix by arranging at least one image among the plurality of images and the at least one composite image, and the second inspection unit may be configured to determine whether the battery is good based on the second image matrix.

[0013] In one embodiment, the first inspection unit is configured to determine whether the battery is defective before the second inspection unit, and the second inspection unit may include a first determination module configured to determine whether the battery is good based on the at least one composite image if the first inspection unit determines that the battery is not defective.

[0014] In one embodiment, the first determination module may be configured to determine whether the at least one synthetic image belongs to any one of the types of the good battery images based on a first deep learning model learned using good battery images, and determine that the battery does not correspond to a good product if the at least one synthetic image corresponds to OOD (Out of Distribution) data that does not belong to any of the types.

[0015] In one embodiment, the second inspection unit may further include a second determination module configured to determine whether the battery is defective based on the at least one composite image, if the first inspection unit determines that the battery is defective.

[0016] In one embodiment, the second judgment module may be configured to determine whether the battery is defective based on a second deep learning model learned using defective battery images.

[0017] In one embodiment, the first judgment module may be configured to determine whether the battery is a good product based on the at least one composite image, if the judgment result by the second judgment module indicates that the battery is not a defective product.

[0018] In one embodiment, the method may further include a learning image generation unit configured to process an image showing a defective part of the battery among the plurality of images and the at least one composite image, thereby generating a learning image used for training a deep learning model for battery appearance inspection.

[0019] In one embodiment, the learning image generation unit may be configured to generate the learning image by setting a partial area including the defective part in the entire area of ​​the image in which the defective part appears, and cropping the remaining area excluding the partial area from the entire area.

[0020] A battery sorting system according to another aspect of the present invention includes: the battery appearance inspection device; and a transport device configured to transport the battery to one of a plurality of pre-designated locations based on an inspection result of the battery appearance inspection device.

[0021] In one embodiment, the transport device may include a first transport unit configured to transport the battery to a first location when the battery is determined to be a good product by at least the second inspection unit among the first inspection unit and the second inspection unit of the battery appearance inspection device; a second transport unit configured to transport the battery to a second location when the battery is determined to be not a defective product by the first inspection unit but not a good product by the second inspection unit; and a third transport unit configured to transport the battery to a third location when the battery is determined to be a defective product by the first inspection unit and again determined to be a defective product by the second inspection unit.

[0022] The battery appearance inspection device according to the present invention performs multiple inspections of the appearance of a battery based on a plurality of images generated by photographing a target portion of a battery under different lighting conditions and a composite image generated by synthesizing the plurality of images, thereby reducing the time and cost required for battery appearance inspection while improving the accuracy and reliability of inspection results.

[0023] In addition, the battery appearance inspection device according to the present invention includes a first inspection unit configured to determine whether the battery is a defective product based on at least one image among the plurality of images, and a second inspection unit configured to determine whether the battery is a good product based on the composite image having depth information of the target portion, thereby reducing the false negative rate and false positive rate occurring in vision inspection.

[0024] In addition, if the first inspection unit determines that the battery is not a defective product, the second inspection unit determines whether the synthetic image belongs to any one of the types of the good battery images based on the first deep learning model learned using the good battery images, and if the synthetic image corresponds to OOD (Out of Distribution) data that does not belong to any of the types, determines the battery as not a good product, thereby minimizing the failure rate of the battery appearance inspection and detecting a battery with a new type of defect.

[0025] In addition, when the battery is determined to be defective by the first inspection unit, the second inspection unit determines whether the battery is defective based on a second deep learning model learned for each type of defect, thereby minimizing the over-inspection rate of the battery's appearance inspection and preventing the loss of good batteries.

[0026] In addition, the battery appearance inspection device according to the present invention generates an image matrix by arranging the plurality of images, or the plurality of images and the composite image, in a matrix format, and inspects the battery using this image matrix, thereby further improving the accuracy and reliability of the inspection result while preventing an increase in inspection time.

[0027] In addition, the battery appearance inspection device according to the present invention processes an image showing a defective part of a battery among the plurality of images and the composite image, thereby generating a learning image used for learning a deep learning model for battery appearance inspection, thereby facilitating learning and upgrading of the deep learning model.

[0028] Furthermore, those skilled in the art will readily understand from the following description that various embodiments of the present invention can solve various technical problems not mentioned above.

[0029] FIG. 1 is a block diagram showing a battery appearance inspection device according to one embodiment of the present invention.

[0030] Figure 2 is a perspective view showing an example of a battery that can be tested according to the present invention.

[0031] Figure 3 is a cross-sectional view of the battery illustrated in Figure 2.

[0032] FIG. 4 is a drawing showing an image generation method of a battery appearance inspection device according to one embodiment of the present invention.

[0033] FIG. 5 is a drawing showing a plurality of cameras of a battery appearance inspection device according to one embodiment of the present invention.

[0034] Figure 6 is a drawing showing an example of a first image matrix transmitted to the first inspection unit.

[0035] Figure 7 is a drawing showing an example of a synthetic image generation method applicable to the present invention.

[0036] Figure 8 is a drawing showing an example of a second image matrix transmitted to a second inspection unit.

[0037] FIG. 9 is a block diagram showing a second determination module of a first inspection unit and a second inspection unit of a battery appearance inspection device according to one embodiment of the present invention.

[0038] FIG. 10 is a drawing showing a learning image generation method of a battery appearance inspection device according to one embodiment of the present invention.

[0039] Figure 11 is a flowchart illustrating a battery appearance inspection method according to one embodiment of the present invention.

[0040] FIG. 12 is a block diagram showing a battery sorting system according to one embodiment of the present invention.

[0041] Hereinafter, embodiments of the present invention will be described in detail with reference to the attached drawings to clarify solutions corresponding to the technical challenges of the present invention. However, when describing the present invention, descriptions of related known technologies may be omitted if they obscure the gist of the present invention. Furthermore, the terms used in this specification are defined in consideration of their functions in the present invention, and these may vary depending on the intentions or practices of designers, manufacturers, etc. Therefore, the definitions of terms described below should be based on the contents throughout this specification.

[0042] FIG. 1 is a block diagram showing a battery appearance inspection device (100) according to one embodiment of the present invention.

[0043] As illustrated in FIG. 1, a battery appearance inspection device (100) according to one embodiment of the present invention includes a lighting unit (110), an image generation unit (120), and a control unit (130). Depending on the embodiment, the battery appearance inspection device (100) may further include one or two or more of a communication unit (140), an input unit (150), a storage unit (160), and an output unit (170).

[0044] The above lighting unit (110) may be configured to change lighting conditions including at least one of the direction, angle, and intensity of light irradiated onto the battery at intervals while irradiating light onto the battery to be inspected.

[0045] To this end, the lighting unit (110) may include a plurality of light sources (112) each arranged at different locations and configured to irradiate light toward the battery at different times. For example, the lighting unit (110) may include m light sources (112) (m is an integer greater than or equal to 2).

[0046] In one embodiment, the plurality of light sources (112) may all have the same shape and structure. In another embodiment, the plurality of light sources (112) may have two or more types of shapes or structures. For example, the plurality of light sources (112) may include one or two or more types of spherical light sources, rod-shaped light sources, ring-shaped light sources, and dome-shaped light sources.

[0047] The image generation unit (120) is configured to capture a target portion of the battery based on light irradiated by the lighting unit (110) and generate multiple images of the target portion. In this case, the multiple images may be captured under different lighting conditions. To this end, the image generation unit (120) may include a camera that captures the target portion of the battery from a predetermined location.

[0048] In one embodiment, the image generation unit (120) may be configured to divide the battery into a plurality of target portions and generate a plurality of images captured under different lighting conditions for each target portion. To this end, the image generation unit (120) may include a plurality of cameras (120) that capture different target portions of the battery from different locations.

[0049] For example, the image generation unit (120) may include a first camera that photographs a first target portion located at the top of the battery, a second camera that photographs a second target portion located at the bottom of the battery, and at least one third camera that photographs a third target portion located at the side of the battery.

[0050] The control unit (130) controls the image generation unit (120) and is configured to inspect the appearance of the battery using images generated by the image generation unit (120). To this end, the control unit (130) may include a general-purpose processor or an application-specific integrated circuit (ASIC) for executing inspection logic, and may optionally further include hardware such as a memory, a register, etc., depending on the embodiment. Meanwhile, the control unit (130) may be configured as a combination of hardware and software. That is, the control logic of the control unit (130) is configured as a computer program and stored in the control unit's (130) own memory or the storage unit (160) described below, and the stored computer program may be executed through the hardware of the control unit (130).

[0051] Meanwhile, the control unit (130) includes an image processing unit (132), a first inspection unit (134), and a second inspection unit (136) as functional components for performing a battery appearance inspection. In one embodiment, the control unit (130) may further include a learning image generation unit (138).

[0052] The image processing unit (132) is configured to generate at least one composite image having depth information of the target portion by synthesizing at least two images among a plurality of images generated by photographing the target portion of the battery. For example, the image processing unit (132) may generate the at least one composite image using a photometric stereo algorithm.

[0053] The first inspection unit (134) is configured to determine whether the battery is defective based on at least one image among the plurality of images. In this case, each of the plurality of images may be a general RGB image. In addition, the first inspection unit (134) may be configured to determine whether the battery is defective using a rule-based inspection algorithm, and to classify the defective location and type of defect in a battery determined to be defective.

[0054] In one embodiment, the image processing unit (132) may be configured to generate a first image matrix by arranging at least two images among the plurality of images in a matrix format. In addition, the image processing unit (132) may provide the generated first image matrix to the first inspection unit (134).

[0055] In this case, the first inspection unit (134) may be configured to determine whether the battery is defective based on the first image matrix. For example, the first inspection unit (134) may determine whether the battery is defective based on each image included in the first image matrix, and may synthesize the determination results for each image to finally determine whether the battery is defective.

[0056] In this way, the present invention can prevent an increase in inspection time while improving the accuracy and reliability of inspection results by performing multiple inspections for defects in batteries using an image matrix.

[0057] The second inspection unit (136) is configured to determine whether the battery is a good product based on at least one composite image generated by the image processing unit (132).

[0058] In one embodiment, the first inspection unit (134) may be configured to determine whether the battery is defective before the second inspection unit (136).

[0059] In this case, the second inspection unit (136) may be configured to determine whether the battery is a good product or a defective product based on the determination result of the first inspection unit (134). To this end, the second inspection unit (136) may include a first determination module (136a) and a second determination module (136b).

[0060] The first determination module (136a) of the second inspection unit (136) may be configured to determine whether the battery is a good product based on at least one composite image, if the first inspection unit (134) determines that the battery is not a defective product.

[0061] To this end, the first judgment module (136a) may utilize a first deep learning model trained using images of good batteries. In this case, the first judgment module (136a) may determine, based on the first deep learning model, whether the at least one synthetic image belongs to any one of the types of good battery images, and, if the at least one synthetic image corresponds to OOD (Out of Distribution) data that does not belong to any of the types, determine the battery as at least not good. The first deep learning model may be one to which a CNN (Convolutional Neural Network) is applied.

[0062] It should be noted that even if the battery is determined to be non-good, it is not immediately confirmed as defective. As will be explained further below, batteries determined to be non-good may be classified separately from batteries determined to be good and batteries determined to be defective, or may be re-determined as defective by the second determination module (136b) described below.

[0063] Meanwhile, the second judgment module (136b) of the second inspection unit (136) may be configured to determine whether the battery is defective based on at least one composite image when the first inspection unit (134) determines that the battery is defective.

[0064] To this end, the second judgment module (136b) may utilize a second deep learning model trained using images of defective batteries. In this case, the second judgment module (136b) may utilize multiple second deep learning models each corresponding to a defect type, such as a dent, scratch, contamination, or rust. In this case, each second deep learning model may be trained using battery images showing a corresponding type of defect. In addition, each second deep learning model may be one to which a convolutional neural network (CNN) is applied.

[0065] In one embodiment, the first determination module (136a) may be configured to re-determine whether the battery is a good product based on the at least one composite image if the battery is not a defective product as determined by the second determination module (136b). In this case, batteries determined as not good products by the first determination module (136a) may be classified separately from batteries determined as good products and batteries determined as defective products.

[0066] In one embodiment, the image processing unit (132) may be configured to generate a second image matrix by arranging at least one image among the plurality of images generated by the image generating unit (120) and the at least one composite image in a matrix format. In addition, the image processing unit (132) may provide the generated second image matrix to the second inspection unit (136).

[0067] In this case, the second inspection unit (136) may be configured to determine whether the battery is good or bad based on the second image matrix. For example, the second inspection unit (136) may calculate a probability value for each image included in the second image matrix that the battery is good or bad, and may ultimately determine whether the battery is good or bad based on the probability values ​​calculated for each image.

[0068] In this way, the battery appearance inspection device according to the present invention generates an image matrix by arranging the plurality of images, or the plurality of images and the composite image, in a matrix format, and inspects the battery using this image matrix, thereby further improving the accuracy and reliability of the inspection result while preventing an increase in inspection time.

[0069] Meanwhile, as mentioned above, the battery exterior inspection device (100) may include a learning image generation unit (138). The learning image generation unit (138) may be configured to process an image in which a defective part of the battery appears among the plurality of images and the at least one composite image, thereby generating a learning image used for training a deep learning model for battery exterior inspection.

[0070] For example, the learning image generation unit (138) may be configured to generate the learning image by setting a partial area including the defective part in the entire area of ​​the image where the defective part of the battery appears, and cropping the remaining area excluding the partial area from the entire area. In this case, the partial area may be set such that the defective part of the battery is located at the center thereof.

[0071] The training images generated in this way can be used for training a new deep learning model or upgrading the second deep learning model through retraining.

[0072] The image processing unit (132), the first inspection unit (134), the second inspection unit (136), and the learning image generation unit (138) of the above-described control unit (130) may be implemented as a combination of a processor and a program executed by the processor. In this case, the control unit (130) may be implemented as a single processor or as two or more processors that are interconnected.

[0073] In one embodiment, the battery appearance inspection device (100) may further include a communication unit (140). The communication unit (140) may be configured to receive data transmitted from a remotely located server or communication terminal via a wired and / or wireless communication network and transmit the data to the control unit (130), or to transmit image data, inspection result data, etc. processed by the control unit (130) to the remotely located server or communication terminal. To this end, the communication unit (140) may include a communication modem that performs wired communication and / or wireless communication.

[0074] In one embodiment, the battery appearance inspection device (100) may further include an input unit (150). The input unit (150) may be configured to receive commands or data from a user or administrator. To this end, the input unit (150) may include an input device such as a keyboard, operation buttons, or a touch panel.

[0075] In one embodiment, the battery appearance inspection device (100) may further include a storage unit (160). The storage unit (160) may be configured to store and manage data necessary for the operation of the battery appearance inspection device (100). To this end, the storage unit (160) may include one or two or more of a ROM, a RAM, an EEPROM, a register, a flash memory, a CD-ROM, a magnetic tape, a hard disk, a floppy disk, and an optical data recording device.

[0076] In one embodiment, the battery appearance inspection device (100) may further include an output unit (170). The output unit (170) may be configured to output the inspection results performed by the control unit (130) as a visual signal, an auditory signal, or an audiovisual signal. To this end, the output unit (170) may include a visual output device such as a light-emitting diode, a monitor, a display panel, or a touch screen. In addition, the output unit (170) may further include a sound generating device such as a speaker.

[0077] Figure 2 is a perspective view showing an example of a battery (20) that can be inspected according to the present invention.

[0078] As illustrated in FIG. 2, the battery appearance inspection device (100) according to the present invention can inspect the appearance of a cylindrical battery (20). Depending on the embodiment, the battery appearance inspection device (100) according to the present invention can be configured to inspect not only other types of batteries, such as pouch-type batteries and square batteries, but also battery modules or battery packs including multiple batteries.

[0079] The above battery (20) has a cylindrical case (22) forming an outer body. This case (22) has an upper surface (22a) and a lower surface (22b), and a rivet (24a) that functions as a first electrode terminal may be provided at the center of the upper surface (22a).

[0080] According to an embodiment, the upper surface (22a) of the case (22) may be configured to perform the function of a second electrode terminal. In this case, an insulation gasket (26) may be provided between the upper surface (22a) of the case (22) and the rivet (24a).

[0081] Figure 3 is a cross-sectional view of the battery (20) illustrated in Figure 2.

[0082] As illustrated in FIG. 3, the case (22) of the battery (20) accommodates an electrode assembly (24) and an electrolyte material. The electrode assembly (24) has a laminated structure in which a positive electrode plate and a negative electrode plate are laminated with a separator interposed therebetween. For example, the electrode assembly (24) may have a jelly-roll shape. In this case, a hollow space (H1) may be formed in the center of the electrode assembly (24).

[0083] Additionally, the lower surface (22b) of the case (22) may be provided with a venting structure (22c) configured to discharge gas generated inside the battery (20).

[0084] An insulator (28) may be placed between the upper bare portion of the electrode assembly (24) electrically connected to the rivet (24a) and the upper surface (22a) of the case (22).

[0085] The internal structure of these batteries is designed to be space-intensive to reduce size and improve energy density. Therefore, any deformation or defects in the battery's exterior directly impact its performance and compromise its safety. Therefore, it is necessary to thoroughly inspect the battery's exterior for defects before activating or using the assembled battery.

[0086] FIG. 4 is a drawing showing an image generation method of a battery appearance inspection device according to one embodiment of the present invention.

[0087] As illustrated in FIG. 4, the lighting unit (110) of the battery exterior inspection device (100) according to the present invention is configured to irradiate light to the battery (20) to be inspected. In this case, the lighting unit (110) uses a plurality of light sources (112a to 112m) that are each arranged at different locations and configured to irradiate light toward the battery (20) at different times, thereby changing the lighting conditions including at least one of the direction, angle, and intensity of the light irradiated to the battery (20) at different times.

[0088] Then, the image generation unit (120) can capture a target portion (e.g., an upper portion) of the battery through a camera (122) based on the light irradiated by the lighting unit (110) and generate multiple images of the target portion, each image captured under different lighting conditions.

[0089] FIG. 5 is a drawing showing a plurality of cameras of a battery appearance inspection device according to one embodiment of the present invention.

[0090] As illustrated in FIG. 5, the image generation unit (120) of the battery appearance inspection device (100) according to the present invention can generate multiple images captured under different lighting conditions for each target portion by using multiple cameras (120) that capture different target portions of the battery appearance inspection device (100) at different locations, respectively.

[0091] For example, the image generating unit (120) may include a first camera (122a) that photographs a first target portion (e.g., a rivet portion) located at the top of the battery (20), a second camera (122b) that photographs a second target portion (e.g., a vent structure portion) located at the bottom of the battery (20), and at least one third camera (122c, 122d) that photographs a third target portion located at the side of the battery.

[0092] Figure 6 is a drawing showing an example of a first image matrix transmitted to the first inspection unit (134).

[0093] As illustrated in FIG. 6, the image processing unit (132) of the battery exterior inspection device (100) according to the present invention can generate a first image matrix (m1) by arranging at least two images among the plurality of images generated by the image generating unit (120) in a matrix format. The number of rows and columns of the first image matrix (m1) can be varied.

[0094] The image processing unit (132) can provide the first image matrix (m1) generated as described above to the first inspection unit (134). Then, the first inspection unit (134) can determine whether the battery is defective based on the first image matrix (m1).

[0095] For example, the first inspection unit (134) can determine whether the corresponding battery is defective based on each image included in the first image matrix (m1), and can synthesize the determination results for each image to finally determine whether the corresponding battery is defective.

[0096] In this way, the present invention can prevent an increase in inspection time while improving the accuracy and reliability of inspection results by performing multiple inspections for defects in batteries using an image matrix.

[0097] Figure 7 is a drawing showing an example of a synthetic image generation method applicable to the present invention.

[0098] As illustrated in FIG. 7, the image processing unit (132) is configured to generate at least one composite image having depth information of the target portion by synthesizing at least two images among a plurality of images generated by photographing a target portion of the battery. For example, the image processing unit (132) can generate a composite image (CI) having depth information of the target portion by synthesizing images (I1 to I4) captured under different lighting conditions through a photometric stereo algorithm.

[0099] Figure 8 is a drawing showing an example of a second image matrix transmitted to the second inspection unit (136).

[0100] As illustrated in FIG. 8, the image processing unit (132) can generate a second image matrix (m2) by arranging images (I1 to I5) captured under different lighting conditions and composite images (CI1 to CI3) synthesized in different combinations in a matrix format. The number of rows and columns of the second image matrix (m2) can be varied. In addition, the combination of images constituting the second image matrix (m2) can be determined according to the type of defect determined by the first inspection unit (134).

[0101] The image processing unit (132) can provide the second image matrix (m2) generated as described above to the second inspection unit (136). Then, the second inspection unit (136) can determine whether the battery is a good product or a defective product based on the second image matrix (m2).

[0102] For example, the second inspection unit (136) calculates a probability value for each image included in the second image matrix (m2) that the battery is a good product or a defective product, and can finally determine whether the battery is a good product or a defective product based on the probability values ​​calculated for each image.

[0103] In this way, the battery appearance inspection device according to the present invention can prevent an increase in inspection time while further improving the accuracy and reliability of inspection results by inspecting the battery using an image matrix in which a photographed image and a synthetic image are combined.

[0104] FIG. 9 is a block diagram showing a first inspection unit (134) and a second judgment module (136b) of a second inspection unit (136) of a battery appearance inspection device according to one embodiment of the present invention.

[0105] As illustrated in FIG. 9, the first inspection unit (134) can be configured to determine whether the battery is defective by using the first image matrix (m1) generated for each target portion of the battery, and to classify the location of the defect and the type of defect among the entire surface of the battery.

[0106] In one embodiment, the first inspection unit (134) may include a first classification module (134a) and a second classification module (134b). In this case, the first classification module (134a) may divide the entire surface of the battery into a plurality of target portions, and classify the location where a defect in the battery occurs into one of the plurality of target portions.

[0107] The second classification module (134b) can classify the type of defect that occurred in the battery into one of several predetermined types based on a rule-based inspection algorithm.

[0108] The second judgment module (136b) of the second inspection unit (136) can determine whether the battery is defective based on the second image matrix (m2) including the above-described composite image when the first inspection unit (134) determines that the battery is defective.

[0109] To this end, the second judgment module (136b) may include a plurality of deep learning models (d1 to dk) trained using defective battery images. In this case, the plurality of deep learning models (d1 to dk) may each be models trained in relation to one type of defect, such as a dent, scratch, contamination, or rust.

[0110] For example, if the type of defect determined by the first inspection unit (134) is a dent, the second determination module (136b) can determine whether the battery is defective by using a deep learning model trained with battery images showing dents among the plurality of deep learning models (d1 to dk).

[0111] FIG. 10 is a drawing showing a learning image generation method of a battery appearance inspection device according to one embodiment of the present invention.

[0112] As illustrated in FIG. 10, the learning image generation unit (138) of the battery exterior inspection device (100) according to one embodiment of the present invention processes an image (I1) showing a defective part (X1) of a battery among the images generated by the image generation unit (120) or the synthesized images synthesized by the image processing unit (132), thereby generating a learning image (I1') used for training a deep learning model for battery exterior inspection.

[0113] For example, the learning image generation unit (138) may set a partial area (A1) including the defective area (X1) of the battery in the entire area of ​​the image (I1) in which the defective area (X1) of the battery appears, and crop the remaining area excluding the partial area (A1) from the entire area of ​​the image (I1) to generate the learning image (I1'). In this case, the partial area (A1) may be set so that the defective area (X1) of the battery is located at the center thereof. In addition, the learning image (I1') may be an image that enlarges the partial area (A1).

[0114] The training images generated in this way can be used for training a new deep learning model or upgrading the second deep learning model through retraining.

[0115] Fig. 11 is a flowchart illustrating a battery appearance inspection method according to one embodiment of the present invention. Referring to Fig. 11, the operation of the battery appearance inspection device (100) described above will be described in time series.

[0116] As illustrated in FIG. 11, while the lighting unit (110) of the battery appearance inspection device (100) irradiates light onto the battery while changing the lighting conditions including at least one of the direction, angle, and intensity of light at intervals, the image generation unit (120) of the battery appearance inspection device (100) photographs a target portion of the battery and generates a plurality of images, each of which is photographed under different lighting conditions, as a plurality of images of the target portion (S10).

[0117] Next, the image processing unit (132) of the battery appearance inspection device (100) synthesizes at least two images among the plurality of images to generate at least one synthesized image having depth information of the corresponding target portion (S20).

[0118] Next, the first inspection unit (134) of the battery appearance inspection device (100) determines whether the battery is defective based on at least one image among the plurality of images (S30, S32).

[0119] For example, the image processing unit (132) generates a first image matrix by arranging at least two images among the plurality of images in a matrix format and provides the first inspection unit (134), and the first inspection unit (134) can determine whether the battery is defective based on the provided first image matrix.

[0120] If the battery is determined not to be a defective product as a result of the first determination by the first inspection unit (134), the second inspection unit (136) of the battery appearance inspection device (100) can determine whether the battery is a good product based on at least one composite image synthesized by the image processing unit (132) (S40, S42).

[0121] For example, the image processing unit (132) generates a second image matrix by arranging at least one image among the plurality of images generated by the image generating unit (120) and the at least one composite image in a matrix format and provides the second image matrix to the second inspection unit (136), and the first determination module (136a) of the second inspection unit (136) can determine whether the battery is good or not based on the provided second image matrix. In this case, the first determination module (136a) can determine whether the battery is good or not using a first deep learning model trained using good battery images.

[0122] If the battery is determined to be a good product as a result of the second determination by the first determination module (136a) of the second inspection unit (136), the battery can be transferred to a first location where subsequent processes for good batteries are performed (S60). On the other hand, if the battery is determined to be a bad product, the battery can be transferred to a second location where batteries subject to re-inspection are gathered (S70).

[0123] Meanwhile, if the battery is determined to be defective as a result of the first determination by the first inspection unit (134), the second inspection unit (136) of the battery appearance inspection device (100) can determine whether the battery is defective based on at least one composite image synthesized by the image processing unit (132) (S50, S52).

[0124] For example, the second judgment module (136b) of the second inspection unit (136) can determine whether the battery is defective based on the second image matrix provided from the image processing unit (132). In this case, the second judgment module (136b) can determine whether the battery is defective using a second deep learning model trained using defective battery images.

[0125] If the battery is determined to be defective as a result of the second determination by the second determination module (136b) of the second inspection unit (136), the battery can be transferred to a third location where defective batteries are collected (S80). On the other hand, if the battery is determined not to be defective, the first determination module (136a) can determine again whether the battery is a good product.

[0126] The above battery appearance inspection device (100) can repeat the above-described steps (S10 to S80) if the next inspection target battery exists (S90).

[0127] Fig. 12 is a block diagram showing a battery sorting system (10) according to one embodiment of the present invention.

[0128] As illustrated in FIG. 12, a battery sorting system (10) according to one embodiment of the present invention includes the battery appearance inspection device (100) described above and a transport device (200).

[0129] The above transport device (200) is configured to transport the battery to one of a plurality of pre-designated locations based on the inspection results of the battery appearance inspection device (100).

[0130] For example, the transport device (200) may include a first transport unit (210), a second transport unit (220), and a third transport unit (230).

[0131] The first transport unit (210) may be configured to transport the battery to a first location where a subsequent process for a good battery is performed when the battery is determined to be good by at least the second inspection unit (136) among the first inspection unit (134) and the second inspection unit (136) of the battery appearance inspection device (100).

[0132] The second transport unit (220) may be configured to transport the battery to a second location where batteries to be re-inspected are gathered, if the battery is determined not to be a defective product by the first inspection unit, but is also determined not to be a good product by the second inspection unit.

[0133] The third transport unit (230) may be configured to transport the battery to a third location where defective batteries are collected, when the battery is determined to be defective by the first inspection unit and again determined to be defective by the second inspection unit.

[0134] Each of the first transfer unit (210), the second transfer unit (220) and the third transfer unit (230) may include one or two or more of a conveyor, a shuttle and a robot arm.

[0135] As described above, the battery appearance inspection device according to the present invention performs multiple inspections of the appearance of the battery based on a plurality of images generated by photographing a target portion of the battery under different lighting conditions and a composite image generated by synthesizing the plurality of images, thereby reducing the time and cost required for battery appearance inspection while improving the accuracy and reliability of the inspection results.

[0136] In addition, the battery appearance inspection device according to the present invention includes a first inspection unit configured to determine whether the battery is a defective product based on at least one image among the plurality of images, and a second inspection unit configured to determine whether the battery is a good product based on the composite image having depth information of the target portion, thereby reducing the false negative rate and false positive rate occurring in vision inspection.

[0137] In addition, if the first inspection unit determines that the battery is not a defective product, the second inspection unit determines whether the synthetic image belongs to any one of the types of the good battery images based on the first deep learning model learned using the good battery images, and if the synthetic image corresponds to OOD (Out of Distribution) data that does not belong to any of the types, determines the battery as not a good product, thereby minimizing the failure rate of the battery appearance inspection and detecting a battery with a new type of defect.

[0138] In addition, when the battery is determined to be defective by the first inspection unit, the second inspection unit determines whether the battery is defective based on a second deep learning model learned for each type of defect, thereby minimizing the over-inspection rate of the battery's appearance inspection and preventing the loss of good batteries.

[0139] In addition, the battery appearance inspection device according to the present invention generates an image matrix by arranging the plurality of images, or the plurality of images and the composite image, in a matrix format, and inspects the battery using this image matrix, thereby further improving the accuracy and reliability of the inspection result while preventing an increase in inspection time.

[0140] In addition, the battery appearance inspection device according to the present invention processes an image showing a defective part of a battery among the plurality of images and the composite image, thereby generating a learning image used for learning a deep learning model for battery appearance inspection, thereby facilitating learning and upgrading of the deep learning model.

[0141] Furthermore, it goes without saying that embodiments according to the present invention can solve various technical problems other than those mentioned in the present specification, not only in the relevant technical field but also in related technical fields.

[0142] The present invention has been described with reference to specific embodiments. However, those skilled in the art will clearly understand that various modifications can be implemented within the technical scope of the present invention. Therefore, the embodiments disclosed above should be considered illustrative rather than limiting. In other words, the true scope of the present invention is set forth in the claims, and all differences within the scope equivalent thereto should be construed as being encompassed by the present invention.

[0143] [Explanation of symbols]

[0144] 10: Battery sorting system

[0145] 100: Battery appearance inspection device

[0146] 110: Lighting Department

[0147] 120: Image generation unit

[0148] 130: Control unit

[0149] 132: Image Processing Unit

[0150] 134: First Inspection Division

[0151] 136: Second Inspection Division

[0152] 136a: First Discrimination Module

[0153] 136b: Second Discrimination Module

[0154] 138: Learning Image Generation Unit

[0155] 140: Communications Department

[0156] 150: Input section

[0157] 160: Storage

[0158] 170: Output section

[0159] 200: Transport device

Claims

1. A lighting unit configured to change lighting conditions including at least one of the direction, angle, and intensity of the irradiated light at intervals while irradiating light onto a battery; An image generating unit configured to capture a target portion of the battery based on light irradiated by the lighting unit and generate a plurality of images of the target portion, each of the plurality of images captured under different lighting conditions; An image processing unit configured to synthesize at least two images among the plurality of images to generate at least one composite image having depth information of the target portion; A first inspection unit configured to determine whether the battery is defective based on at least one image among the plurality of images; and A battery appearance inspection device including a second inspection unit configured to determine whether the battery is of good quality based on at least one composite image.

2. In paragraph 1, The above lighting unit, A battery appearance inspection device characterized by comprising a plurality of light sources arranged at different locations and configured to irradiate light toward the battery at different times.

3. In paragraph 1, The above image generation unit, A battery appearance inspection device characterized by including a plurality of cameras configured to photograph different target portions of the battery at different locations.

4. In paragraph 3, The above multiple cameras, A first camera for photographing a first target portion located at the top of the battery; A second camera for photographing a second target portion located at the bottom of the battery; and A battery appearance inspection device characterized by including at least one third camera for photographing a third target portion located on the side of the battery.

5. In paragraph 1, The image processing unit is configured to generate a first image matrix by arranging at least two images among the plurality of images, A battery appearance inspection device characterized in that the first inspection unit is configured to determine whether the battery is defective based on the first image matrix.

6. In paragraph 1, The image processing unit is configured to generate a second image matrix by arranging at least one image among the plurality of images and the at least one composite image, A battery appearance inspection device characterized in that the second inspection unit is configured to determine whether the battery is a good product based on the second image matrix.

7. In paragraph 1, The first inspection unit is configured to determine whether the battery is defective before the second inspection unit, The second inspection department above, A battery appearance inspection device characterized in that it includes a first judgment module configured to determine whether the battery is a good product based on the at least one composite image when the battery is determined by the first inspection unit not to be a defective product.

8. In paragraph 7, A battery appearance inspection device characterized in that the first judgment module is configured to determine whether the at least one synthetic image belongs to any one of the types of the good battery images based on a first deep learning model learned using good battery images, and to determine that the battery does not correspond to a good product if the at least one synthetic image corresponds to OOD (Out of Distribution) data that does not belong to any of the types.

9. In paragraph 7, The second inspection department above, A battery appearance inspection device characterized in that it further includes a second judgment module configured to determine whether the battery is defective based on the at least one composite image when the battery is determined to be defective by the first inspection unit.

10. In paragraph 9, A battery appearance inspection device characterized in that the second judgment module is configured to determine whether the battery is defective based on a second deep learning model learned using defective battery images.

11. In paragraph 9, A battery appearance inspection device characterized in that the first judgment module is configured to determine whether the battery is a good product based on the at least one composite image when the battery is not a defective product as a result of the judgment by the second judgment module.

12. In paragraph 1, A battery appearance inspection device further comprising a learning image generation unit configured to process an image showing a defective part of the battery among the plurality of images and at least one composite image, thereby generating a learning image used for training a deep learning model for battery appearance inspection.

13. In paragraph 12, A battery appearance inspection device characterized in that the learning image generation unit is configured to generate the learning image by setting a portion of the entire area of ​​the image in which the defective portion appears, including the defective portion, and cropping the remaining area excluding the portion of the entire area.

14. The battery appearance inspection device according to any one of the first to thirteenth clauses; and A battery sorting system including a transport device configured to transport the battery to one of a plurality of pre-designated locations based on the inspection results of the battery appearance inspection device.

15. In paragraph 14, The above transport device, A first transport unit configured to transport the battery to a first location when the battery is determined to be a good product by at least the second inspection unit among the first inspection unit and the second inspection unit of the battery appearance inspection device; A second transport unit configured to transport the battery to a second location when the battery is determined to be not a defective product by the first inspection unit, but is also determined to be not a good product by the second inspection unit; and A battery sorting system characterized by including a third transport unit configured to transport the battery to a third location when the battery is determined to be defective by the first inspection unit and again determined to be defective by the second inspection unit.

Images