Apparatus, method and system for detecting defect of battery

Abstract

A battery defect detection device according to one embodiment disclosed in this document includes a communication module, a processor, and a memory storing a first artificial intelligence model, a second artificial intelligence model, and instructions. When the instructions are executed by the processor, the battery defect detection device may be configured to acquire an image of a target product using the communication module, input the image of the target product into the first artificial intelligence model to classify the target product, and if the target product is classified as normal, input the image of the target product into the second artificial intelligence model to determine whether the image of the target product corresponds to first data among the training data of the first artificial intelligence model for classifying the target product as normal.

Description

Technology Field

[0001] The embodiments disclosed in this document relate to a battery defect detection device, method, and system. Background Technology

[0002] Recently, as the demand for portable electronic products such as laptops, video cameras, and mobile phones has increased rapidly, and the development of electric vehicles, energy storage batteries, robots, and satellites has accelerated, research on high-performance batteries capable of repeated charging and discharging is actively underway.

[0003] Currently commercialized batteries include nickel-cadmium, nickel-hydrogen, nickel-zinc, and lithium batteries. Among these, lithium batteries are gaining attention for their advantages, such as the ability to freely charge and discharge with almost no memory effect compared to nickel-based batteries, a very low self-discharge rate, and high energy density. The problem to be solved

[0004] During the manufacturing and assembly process of batteries, defects in the appearance of the battery pouch (e.g., corner dents, scratches, wrinkles, electrolyte contamination, etc.) may occur due to deformation, crumpling, or damage to the sealing portion of the battery casing. Since defects in the pouch's appearance can affect the reliability and / or safety of the battery, it is important to detect defects in the battery (e.g., pouch) before the battery is distributed.

[0005] Visual inspection systems commonly used to detect external defects in batteries (e.g., pouches) collect and learn images of both good and defective product types, and inspect batteries that have undergone manufacturing and assembly processes to determine whether external defects exist. However, if a new type of defect occurs that is not among the previously learned types, the battery may be judged as a good product. Furthermore, since the occurrence rate of new defect types is inherently low during the manufacturing and assembly processes of pouches, it is difficult to collect suitable defective product images for the visual inspection system's training.

[0006] Accordingly, the present invention aims to prevent the leakage of defects by compensating for the uncertainty in detecting novel defect types. Other objects and advantages of the present invention can be understood from the following description and will become more clearly known through the embodiments of the present invention. Furthermore, it will be readily apparent that the objects and advantages of the present invention can be realized by the means and combinations thereof set forth in the claims. means of solving the problem

[0007] A battery defect detection device according to one embodiment disclosed in this document includes a communication module, a processor, and a memory storing a first artificial intelligence model, a second artificial intelligence model, and instructions. When the instructions are executed by the processor, the battery defect detection device may be configured to acquire an image of a target product using the communication module, input the image of the target product into the first artificial intelligence model to classify the target product, and if the target product is classified as normal, input the image of the target product into the second artificial intelligence model to determine whether the image of the target product corresponds to first data among the training data of the first artificial intelligence model for classifying the target product as normal.

[0008] According to one embodiment, when the instructions are executed by the processor, the battery defect detection device may be configured to reclassify the target product as defective if the image does not correspond to the first data.

[0009] According to one embodiment, when the instructions are executed by the processor, the battery defect detection device may be configured to determine the image as second data for classifying the target product as defective if the image does not correspond to the first data.

[0010] According to one embodiment, the instructions may be configured such that when executed by the processor, the battery defect detection device trains the first artificial intelligence model based on the second data.

[0011] According to one embodiment, the image of the target product can be acquired in multiple channels by an image acquisition device.

[0012] According to one embodiment, when the instructions are executed by the processor, the battery defect detection device may be configured to obtain a similarity between the image and the first data through the second artificial intelligence model and reclassify the target product as defective or normal based on the similarity.

[0013] According to one embodiment, when the instructions are executed by the processor, the battery defect detection device may be configured to determine the image as second data for classifying the target product as defective if the similarity of the image is less than a preset reference value.

[0014] According to one embodiment, the instructions may be configured such that when executed by the processor, the battery defect detection device trains the first artificial intelligence model based on the second data.

[0015] A battery defect detection method according to one embodiment disclosed in this document may include the operation of acquiring an image of a target product, the operation of inputting the image of the target product into a first artificial intelligence model to classify the target product, and, when the target product is classified as normal, the operation of inputting the image of the target product into a second artificial intelligence model to determine whether the image of the target product corresponds to first data among the training data of the first artificial intelligence model for classifying the target product as normal.

[0016] According to one embodiment, the battery defect detection method may further include an operation of reclassifying the target product as defective when the image does not correspond to the first data.

[0017] According to one embodiment, the battery defect detection method may further include the operation of determining the image as second data for classifying the target product as defective when it does not correspond to the first data.

[0018] According to one embodiment, the battery defect detection method may further include the operation of training the first artificial intelligence model based on the second data.

[0019] According to one embodiment, the image of the target product can be acquired in multiple channels.

[0020] According to one embodiment, the battery defect detection method may further include the operation of obtaining a similarity between the image and the first data through the second artificial intelligence model, and the operation of reclassifying the target product as defective or normal based on the similarity.

[0021] According to one embodiment, the battery defect detection method may further include an operation of determining the image as second data for classifying the target product as defective when the similarity of the image is less than a preset threshold value.

[0022] According to one embodiment, the battery defect detection method may further include the operation of training the first artificial intelligence model based on the second data.

[0023] A battery defect detection system according to one embodiment disclosed in this document may include: a battery defect detection device that acquires an image of a target product and classifies the target product using a first artificial intelligence model that is trained using the image of the target product and previously stored training data; and a novel defect detection device that, when the target product is classified as normal, inputs the image of the target product into the second artificial intelligence model to determine whether the image of the target product corresponds to the first data among the training data of the first artificial intelligence model for classifying the target product as normal. Effects of the invention

[0024] The battery defect detection device according to the disclosure of this document can detect a new type of defective product not included in previously stored training data, thereby preventing the leakage of defective products.

[0025] The battery defect detection device according to the disclosure of this document can improve the precision of defect detection by learning images of a new type of defective product.

[0026] The battery defect detection system disclosed in this document allows for the combined use of a new defect detection device with an existing defect detection device, thereby increasing the convenience of introducing a new defect detection device.

[0027] The effects of the battery defect detection device, method, and system disclosed in this document are not limited to those mentioned above, and other unmentioned effects will be clearly understood by those skilled in the art in accordance with the disclosure of this document. Brief explanation of the drawing

[0028] FIG. 1 is a drawing showing a battery defect detection device according to one embodiment disclosed in this document. FIG. 2 illustrates an artificial intelligence model according to one embodiment disclosed in this document. FIG. 3a is a drawing illustrating first data for classifying a target product as normal among learning data according to one embodiment disclosed in this document. FIG. 3b is a drawing illustrating second data for classifying a target product as defective among the training data according to one embodiment disclosed in this document. FIG. 4 illustrates an artificial intelligence model according to one embodiment disclosed in this document. FIG. 5 illustrates an artificial intelligence model according to one embodiment disclosed in this document and a process for detecting good and / or defective products among target products using the artificial intelligence model. Figure 6 is a diagram illustrating images classified as normal by an artificial intelligence model. FIG. 7 is a drawing showing a battery defect detection method according to one embodiment disclosed in this document. FIG. 8 is a drawing showing a battery defect detection method according to one embodiment disclosed in this document. FIG. 9 is a drawing showing a battery defect detection method according to one embodiment disclosed in this document. FIG. 10 is a drawing showing a battery defect detection method according to one embodiment disclosed in this document. FIG. 11 is a drawing showing a battery defect detection system according to one embodiment disclosed in this document. In relation to the description of the drawings, the same or similar reference numerals may be used for identical or similar components. Specific details for implementing the invention

[0029] Hereinafter, various embodiments of the present invention are described with reference to the accompanying drawings. However, this is not intended to limit the present invention to specific embodiments and should be understood to include various modifications, equivalents, and / or alternatives of the embodiments of the present invention.

[0030] The various embodiments of this document and the terms used therein are not intended to limit the technical features described in this document to specific embodiments, and should be understood to include various modifications, equivalents, or substitutions of said embodiments. In connection with the description of the drawings, similar reference numerals may be used for similar or related components. The singular form of a noun corresponding to an item may include one or more of said items unless the relevant context clearly indicates otherwise.

[0031] In this document, each of the phrases such as “A or B,” “at least one of A and B,” “at least one of A or B,” “A, B or C,” “at least one of A, B and C,” and “at least one of A, B, or C” may include any one of the items listed together in the corresponding phrase, or all possible combinations thereof. Terms such as “first,” “second,” “first,” “second,” “A,” “B,” “(a),” or “(b)” may be used simply to distinguish a component from another component and, unless specifically stated otherwise, do not limit the components in any other aspect (e.g., importance or order).

[0032] In this document, where it is mentioned that any (e.g., 1) component is “connected,” “coupled,” or “joined” to another (e.g., 2) component, with or without the terms “functionally” or “communicationly,” or where it is mentioned as “coupled” or “connected,” it means that said component may be connected to said other component directly (e.g., by wire), wirelessly, or through a third component.

[0033] According to one embodiment, the method according to the various embodiments disclosed herein may be provided as included in a computer program product. The computer program product may be traded between a seller and a buyer as a product. The computer program product may be distributed in the form of a device-readable storage medium (e.g., compact disc read-only memory (CD-ROM)), or distributed online (e.g., download or upload) through an application store or directly between two user devices. In the case of online distribution, at least a portion of the computer program product may be temporarily stored or temporarily created on a device-readable storage medium, such as the memory of a manufacturer's server, an application store's server, or a relay server.

[0034] According to various embodiments, each component (e.g., module or program) of the components described above may include a singular or multiple entities, and some of the multiple entities may be separated and placed in other components. According to various embodiments, one or more of the components or operations of the aforementioned components may be omitted, or one or more other components or operations may be added. Generally or additionally, multiple components (e.g., module or program) may be integrated into a single component. In this case, the integrated component may perform one or more functions of each of the multiple components in the same or similar manner as those performed by the corresponding component among the multiple components prior to integration. According to various embodiments, operations performed by the module, program, or other components may be executed sequentially, in parallel, iteratively, or heuristically, or one or more of the operations may be executed in a different order, omitted, or one or more other operations may be added.

[0035] FIG. 1 is a drawing showing a battery defect detection device according to one embodiment disclosed in this document.

[0036] Referring to FIG. 1, a battery defect detection device (101) according to one embodiment disclosed in this document may be connected to an image acquisition device (103) and a user terminal (105) via wired and / or wireless connections.

[0037] In one embodiment, the connection between the battery defect detection device (101) and the image acquisition device (103) may be a communication connection via a wired and / or wireless network. In one embodiment, the wired network may be based on LAN (local area network) communication or power line communication. In one embodiment, the wireless network may be based on a short-range communication network (e.g., Bluetooth, WiFi (wireless fidelity) or IrDA (infrared data association)), or a long-range communication network (cellular network, 4G network, 5G network).

[0038] In another embodiment, the connection between the battery fault diagnosis device (101) and the image acquisition device (103) may be a connection via a device-to-device communication method (e.g., bus, GPIO (general purpose input and output), SPI (serial peripheral interface), or MIPI (mobile industry processor interface)).

[0039] In one embodiment, the connection between the battery fault diagnosis device (101) and the user terminal (105) may be a communication connection via a wired and / or wireless network.

[0040] In one embodiment, the image acquisition device (103) can acquire an image of a target (e.g., battery unit (115)). According to the embodiment, the target (e.g., battery unit (115)) may include a secondary battery. According to the embodiment, the secondary battery may include a pouch-type secondary battery, a prismatic secondary battery, and / or a cylindrical secondary battery.

[0041] The image acquisition device (103) can acquire an image of the target object by photographing the target object. According to one embodiment, the image acquisition device (103) can be implemented as a camera. According to one embodiment, the image acquisition device (103) can acquire images of the target object in multiple channels.

[0042] In one embodiment, the user terminal (105) may be a mobile device (e.g., mobile phone, laptop computer, smartphone, smart pad) or a PC (personal computer). In one embodiment, the user terminal (105) may be a terminal used by the manager of the battery defect detection device (101).

[0043] In one embodiment, the battery failure detection device (101) may include a communication circuit (120), a memory (140), and a processor (150). According to an embodiment, the battery failure detection device (101) illustrated in FIG. 1 may further include at least one component (e.g., a display, an input device, or an output device) in addition to the components illustrated in FIG. 1.

[0044] In one embodiment, the communication circuit (120) establishes a wired communication channel and / or a wireless communication channel between the battery defect detection device (101) and the image acquisition device (103) and / or the user terminal (105), and can transmit and receive data with the image acquisition device (103) and / or the user terminal (105) through the established communication channel.

[0045] In one embodiment, the memory (140) may include volatile memory and / or non-volatile memory.

[0046] In one embodiment, the memory (140) may store data used by at least one component (e.g., processor (150)) of the battery defect detection device (101). For example, the data may include a program (130) (or an instruction related thereto), input data, or output data. In one embodiment, the instruction may cause the battery defect detection device (101) to perform operations defined by the instruction when executed by the processor (150).

[0047] In one embodiment, the program (130) may include one or more software components (e.g., an artificial intelligence model learning unit (141), an image acquisition unit (143), a diagnosis unit (145), and one or more artificial intelligence models (161, 165)).

[0048] In one embodiment, the processor (150) may include a central processing unit, an application processor, a graphics processing unit, a neural processing unit (NPU), an image signal processor, a sensor hub processor, or a communication processor.

[0049] In one embodiment, the processor (150) can execute software (e.g., an artificial intelligence model learning unit (141), an image acquisition unit (143), a diagnosis unit (145), an anomaly processing unit (147), a preprocessing unit (149), and one or more artificial intelligence models (161, 165)) to control at least one other component (e.g., a hardware or software component) of the battery defect detection device (101) connected to the processor (150) and can perform various data processing or operations.

[0050] Hereinafter, with reference to FIGS. 2 to 5, a method is described in which a battery defect detection device (101) diagnoses an abnormality of a target product (e.g., battery unit (115)) through an artificial intelligence model learning unit (141), an image acquisition unit (143), a diagnosis unit (145), and one or more artificial intelligence models (161, 165).

[0051] 1st AI model training

[0052] FIG. 2 illustrates an artificial intelligence model (161) according to an embodiment disclosed in this document. FIG. 3a is a diagram illustrating first data for classifying a target product as normal among training data according to an embodiment disclosed in this document. FIG. 3b is a diagram illustrating second data for classifying a target product as defective among training data according to an embodiment disclosed in this document.

[0053] In one embodiment, the artificial intelligence model learning unit (141) can train the artificial intelligence model (161) based on images (201) and images (205) of a previously acquired target product. Here, the images (201) and images (205) may be previously acquired training data. Images (201) may represent first data for classifying the target product as normal among the training data. Images (205) may represent second data for classifying the target product as defective among the training data. In one embodiment, the images (201) and images (205) may include the entire image of the target product or images of each part of the target product. In one embodiment, when the images (201) and images (205) consist of the entire image, the images (201) may consist of images of the target product classified as normal, and the images (205) may consist of images of the target product classified as defective. In one embodiment, when the images (201) and images (205) are composed of partial images, the images (201) may be composed of images classified as normal among images obtained by dividing the entire image of the target product for each inspection item (e.g., dents, presses, printed text, welding), and the images (205) may be composed of images classified as defective among images obtained by dividing the entire image of the target product for each inspection item.

[0054] For example, referring to FIG. 3a, the first data (310) may include images classified as normal among images divided by inspection items. For example, images (311) may be images where the printed text is normal, and images (312) may be images where the degree of indentation is within the normal range. Likewise, referring to FIG. 3b, the second data (350) may include images classified as defective among images divided by inspection items. For example, images (351) may be images where the printed text is abnormal, and images (352) may be images where the degree of indentation or pressure is outside the normal range.

[0055] In one embodiment, the artificial intelligence model (161) may be a model capable of multinomial classification (e.g., a model based on a convolutional neural network (CNN)). In one embodiment, the artificial intelligence model (161) may be trained to classify whether a product is normal or defective based on images (201, 205) of the product. In one embodiment, the artificial intelligence model (161) may be trained to classify whether each of the inspection items of the product is normal or defective based on images (201, 205) of the product.

[0056] In one embodiment, the artificial intelligence model (161) may include an input layer (210), a hidden layer (220), and an output layer (230). Here, the input layer (210) may receive images (201) and images (205) of a previously acquired target product. The hidden layer (220) may have a structure in which a plurality of layers are connected sequentially. The output layer (230) may be a layer for outputting a multinomial classification result (209) for each of the images (201) and images (205). In one embodiment, the multinomial classification result (209) may include probability values ​​for classes (241, 243, 245, 247) according to inspection items for the images (201) and images (205). For example, class (241) may represent the probability that the printed text is normal, and class (243) may represent the probability that the printed text is defective. As another example, class (245) may represent the probability that the degree of dent or indentation is within the normal range, and class (247) may represent the probability that the degree of dent or indentation is outside the normal range. The sum of the probability values ​​for classes (241, 243, 245, 247) may be 1.

[0057] In one embodiment, the artificial intelligence model learning unit (141) can adjust the parameters of the hidden layer (220) so that the obtained polynomial classification result (209) and the prior classification result of each of the images (201) and images (205) match or are less than or equal to a reference difference value by inputting images (201) and images (205) into the artificial intelligence model (161).

[0058] Second AI model training

[0059] FIG. 4 illustrates an artificial intelligence model (165) according to one embodiment disclosed in this document.

[0060] In one embodiment, the artificial intelligence model learning unit (141) can train the artificial intelligence model (165) based on images (201) of a previously acquired target product. Here, the images used to train the artificial intelligence model (165) may be first data for classifying the target product as normal among the training data.

[0061] In one embodiment, the artificial intelligence model (165) may be a model capable of multinomial classification (e.g., a model based on a CNN). In one embodiment, the artificial intelligence model (165) may be trained to output a similarity with normal images for each of the inspection items based on first data (i.e., images (201)) for the target product. Here, similarity may refer to the maximum value of the probability value according to the multinomial classification result obtained by the artificial intelligence model (165).

[0062] In one embodiment, the artificial intelligence model (165) can be trained to distinguish whether an image is in-distribution (i.e., corresponding to a previously learned image of a good product) or out-of-distribution (i.e., corresponding to an image of a new type of defective product) by comparing a similarity based on first data (i.e., images (201)) for a target product with a reference value.

[0063] In one embodiment, the artificial intelligence model (165) may include an input layer (410), a hidden layer (420), and an output layer (430). Here, the input layer (410) may receive images (201) of a previously acquired target. The hidden layer (420) may have a structure in which a plurality of layers are connected sequentially. The output layer (430) may be a layer for outputting a multinomial classification result (409) for each of the images (201). In one embodiment, the multinomial classification result (409) may include probability values ​​for classes according to inspection items for the images (201). The classes set in the artificial intelligence model (165) may be fewer than the classes set in the artificial intelligence model (161). For example, only classes representing normal may exist in the artificial intelligence model (165). For example, the artificial intelligence model (165) may include a class that indicates the probability that the printed text is normal, or a class that indicates the probability that the printed text is normal. The sum of the probability values ​​for the classes of the artificial intelligence model (165) may be 1.

[0064] In one embodiment, the artificial intelligence model learning unit (141) can adjust the parameters of the hidden layer (420) so that the maximum value (441) of the probability value according to the obtained multinomial classification result (409) exceeds the reference value (449) by inputting images (201) into the artificial intelligence model (165).

[0065] In one embodiment, the artificial intelligence model learning unit (141) can test the artificial intelligence model (165) using images (205). For example, the artificial intelligence model learning unit (141) can check whether the maximum value (445) of the probability value according to the multinomial classification result (409) of the images (205) is less than or equal to a reference value (449). In one embodiment, if the maximum value (445) of the probability value according to the multinomial classification result (409) of the images (205) exceeds the reference value (449), the artificial intelligence model learning unit (141) can adjust the reference value (449) of the artificial intelligence model (165). For example, the artificial intelligence model learning unit (141) can adjust the reference value (449) based on a baseline algorithm or an ODIN (out of distribution detector for neural networks) algorithm. Here, the adjustment of the reference value (449) can be distinguished from the learning process for adjusting the parameters of the hidden layer (420).

[0066] Detection of defects in the target product

[0067] FIG. 5 illustrates a process for detecting good and / or defective products among target products using an artificial intelligence model (161) and an artificial intelligence model (165) according to an embodiment disclosed in this document. FIG. 6 is a drawing illustrating images classified as normal by an artificial intelligence model.

[0068] In one embodiment, the processor (120) can detect good products and / or defective products among the target products through the diagnostic unit (145). The processor (120) can classify the target products as normal or defective through the diagnostic unit (145) in order to detect good products and / or defective products. The processor (120) can classify the target products as normal or defective based on previously stored learning data through the diagnostic unit (145). In the following, unless otherwise noted, the operations of the processor (120) may be understood to be performed through the diagnostic unit (145).

[0069] In one embodiment, the processor (120) can classify a target product using an image (501) of the target product and previously stored training data. The processor (120) can compare the image (501) of the target product acquired from the image acquisition device (103) with the previously stored training data. Here, the image (501) of the target product acquired from the image acquisition device (103) can be acquired by the processor (120) through an image acquisition unit (143). In one embodiment, the image acquisition unit (143) may include a network driver for controlling the communication circuit (120). According to the embodiment, the processor (120) can compare the image (501) of the target product with previously stored training data to classify the target product. According to the embodiment, the processor (120) can classify the target product based on an artificial intelligence model (161) to which the previously stored training data is applied.

[0070] According to an embodiment, the training data may each include at least one data for classifying a target product as normal and at least one data for classifying a target product as defective. The first data may refer to the data among the training data for classifying a target product as normal. According to an embodiment, the first data may include an image of a good product. According to an embodiment, the processor (120) may learn criteria for classifying a target product as normal using the first data.

[0071] The processor (120) can classify a target product as normal or defective based on previously stored training data. A target product may refer to a target product classified as normal when the processor (120) classifies the target product based on previously stored training data. According to an embodiment, the processor (120) may classify a new type of defective product that has not been learned by the previously stored training data as normal. In this case, the target product may include not only good products but also new types of defective products. For example, the images (610) may be images of existing types of normal products, and the images (650) may be images of new types of defective products. In this case, when the images (610) are input into the artificial intelligence model (161), the target product for the images (610) may be classified as normal (i.e., good product). As another example, when images (650) of new types of defective products are input into the artificial intelligence model (161), the target product for the images (650) can also be classified as normal (i.e., good product).

[0072] The processor (120) can acquire target images (i.e., images (610, 650)) corresponding to the target product. The processor (120) can determine the characteristics of the target image using the first data and the target image. The processor (120) can detect a new type of defective product using the characteristics of the target image. According to an embodiment, the processor (120) can reclassify the target product as normal or defective using the characteristics of the target image. According to an embodiment, the processor (120) can classify the product based on an artificial intelligence model (165) to which the first data among the previously stored training data is applied. In one embodiment, the processor (120) can detect a new type of defective product by comparing the similarity obtained by inputting each of the target images (i.e., images (610, 650)) into the artificial intelligence model (165) with a reference value. Here, similarity may refer to the maximum value of the probability value according to the multinomial classification result obtained by the artificial intelligence model (165). For example, in the case of images (610, 650) where the similarity exceeds a reference value, the processor (120) may determine that the corresponding products are normal. In another example, in the case of images (650) where the similarity is less than a reference value among images (610, 650), the processor (120) may determine that the corresponding products are defective.

[0073] Among the target items, the items reclassified as defective may correspond to a new type of defective item. The processor (120) can detect a new type of defective item by undergoing a reclassification process with the target items.

[0074] According to an embodiment, the processor (120) can compare whether the target image corresponds to the first data. If the target image does not correspond to the first data, the processor (120) can reclassify the target product as defective.

[0075] According to an embodiment, the processor (120) can determine the similarity between a target image and a first data. The processor (120) can determine the similarity between a target image and a first data according to a preset method. Based on the similarity between the target image and the first data, the processor (120) can reclassify the target product as defective or normal. According to an embodiment, the processor (120) can reclassify the target product as defective if the similarity is less than a preset threshold value, and reclassify the target product as normal if the similarity is greater than or equal to the threshold value. According to an embodiment disclosed in this document, the processor (120) determines the similarity between the target image and the first data and reclassifies the target product as defective or normal based on the similarity, but is not limited thereto. According to another embodiment, the processor (120) may determine the dissimilarity between the target image and the first data and reclassify the target product as defective or normal based on the dissimilarity.

[0076] The processor (120) may determine the target image at the time the target product is reclassified as defective as the second data. The second data may correspond to an image for classifying the product as defective. The second data may include an image of a new type of defective product that is not included in the previously stored training data. According to an embodiment, the processor (120) may include the second data in the training data. By including the second data in the training data, the processor (120) can learn the new type of defect. The processor (120) may retrain the artificial intelligence model (161) based on the images (e.g., images (650)) newly included in the second data.

[0077] According to an embodiment, the processor (120) may determine the target image as second data when the target image does not correspond to the first data. According to another embodiment, the processor (120) may determine the target image as second data when the similarity of the target image is less than a preset threshold value.

[0078] FIG. 7 is a drawing showing a battery defect detection method according to one embodiment disclosed in this document.

[0079] Referring to FIG. 7, a battery defect detection method according to one embodiment disclosed in this document may include the step of acquiring an image of a target product (S100), the step of classifying a target product using stored training data (S110), the step of acquiring a target image corresponding to a target product (S120), and the step of determining the characteristics of the target image using the first data and the target image (S130).

[0080] Below, a method for detecting battery defects is described in detail with reference to FIG. 1. For the convenience of explanation, content that overlaps with previously explained content is omitted or briefly explained.

[0081] In step S100, the battery defect detection device (101) can acquire an image of the target product. According to an embodiment, the target product may include a secondary battery. According to an embodiment, in step S100, the battery defect detection device (101) can acquire an image of the target product in multiple channels. Step S100 can be performed by the image acquisition unit (143) of the battery defect detection device (101).

[0082] In step S110, the battery defect detection device (101) can classify a target product using stored training data. According to an embodiment, the battery defect detection device (101) can compare an image of a target product with previously stored training data and classify the target product based on the comparison result. According to an embodiment, the battery defect detection device (101) can classify a target product based on an artificial intelligence model (161) to which previously stored training data is applied. Step S110 can be performed by the processor (120) of the battery defect detection device (101).

[0083] According to an embodiment, the training data may include at least one piece of data for classifying a target product as normal or defective. The first data may refer to data among the training data for classifying a target product as normal. According to an embodiment, the first data may include an image of a good product. According to an embodiment, the battery defect detection device (101) may learn criteria for classifying a target product as normal using the first data.

[0084] In step S110, the battery defect detection device (101) can classify a target product as normal or defective based on training data. In step S710, a target product classified as normal can be referred to as a target product. According to the embodiment, the battery defect detection device (101) can classify a target product as normal or defective based on training data, and a new type of defective product that has not been learned by the training data may be classified as normal. At this time, the target product may include not only good products but also new types of defective products.

[0085] In step S120, the battery defect detection device (101) can acquire a target image corresponding to a target product. According to an embodiment, the battery defect detection device (101) can select a target image corresponding to a target product from among the images of the acquired products. Step S120 can be performed by the processor (120) of the battery defect detection device (101).

[0086] In step S130, the battery defect detection device (101) can determine the characteristics of the target image using the first data and the target image. The first data may refer to data for classifying a target product as normal among the training data. According to an embodiment, the first data may include an image of a good product. According to an embodiment, the battery defect detection device (101) can determine the characteristics of the target image using the first data and the target image based on an artificial intelligence model (165) to which the first data is applied. Step S130 may be performed by the processor (120) of the battery defect detection device (101).

[0087] In step S130, the battery defect detection device (101) can determine the characteristics of the target image. According to an embodiment, the battery defect detection device (101) can determine whether the target image corresponds to the first data. According to another embodiment, the battery defect detection device (101) can determine the similarity between the target image and the first data.

[0088] The battery defect detection device (101) can reclassify the target product as normal or defective based on the characteristics of the target image determined in step S130. A specific method for reclassifying the target product as normal or defective based on the characteristics of the target image determined by the battery defect detection device (101) will be described later in FIGS. 8 and 9.

[0089] FIG. 8 is a drawing showing a battery defect detection method according to one embodiment disclosed in this document.

[0090] Referring to FIG. 8, a battery defect detection method according to one embodiment disclosed in this document may further include the steps of determining whether a target image corresponds to first data (S200), reclassifying a target product as defective if the target image does not correspond to first data (S210), determining the target image as second data (S220), and / or including the second data in training data (S230).

[0091] In the following, a method for detecting battery defects will be described in detail with reference to FIGS. 1 and FIGS. 7. For the convenience of explanation, content that overlaps with previously explained content will be omitted or briefly explained.

[0092] In step S200, the battery defect detection device (101) can determine whether the target image corresponds to the first data. Step S200 may substantially correspond to a specific example of step S130 of FIG. 7.

[0093] In step S200, if the battery defect detection device (101) determines that the target image does not correspond to the first data, the battery defect detection device (101) may reclassify the target item as defective (S210). According to the embodiment, the target item is an item classified as normal by the battery defect detection device (101) based on the training data, but the target item may also include a new type of defective item not included in the training data. In step S200, the battery defect detection device (101) may detect a new type of defective item included in the target item by determining whether the target item corresponds to the first data. Step S200 may be performed by the processor (120) of the battery defect detection device (101).

[0094] If it is determined in step S200 that the target image corresponds to the first data, the battery defect detection device (101) can reclassify the target item as normal.

[0095] If it is determined in step S200 that the target image does not correspond to the first data, the battery defect detection device (101) may reclassify the target product as defective (S210).

[0096] In step S210, the target product reclassified as defective by the battery defect detection device (101) may correspond to a new type of defective product not included in the training data. Step S210 may be performed by the processor (120) of the battery defect detection device (101).

[0097] In step S220, the battery defect detection device (101) may determine the target image of the target product that the battery defect detection device (101) reclassified as defective in step S210 as the second data. According to an embodiment, the second data may include an image of a new type of defective product in the battery defect detection device (101). Step S220 may be performed by the processor (120) of the battery defect detection device (101).

[0098] In step S230, the battery defect detection device (101) may include second data in the training data. According to an embodiment, the second data may correspond to an image for the battery defect detection device (101) to classify a target product as defective. According to an embodiment, the second data may include an image of a new type of defective product that is not included in the previously stored training data. In step S230, the battery defect detection device (101) may learn the new type of defect by including the second data in the training data. Step S230 may be performed by the processor (120) of the battery defect detection device (101).

[0099] FIG. 9 is a drawing showing a battery defect detection method according to one embodiment disclosed in this document.

[0100] Referring to FIG. 9, a battery defect detection method according to one embodiment disclosed in this document may further include a step (S300) of determining the similarity between a target image and a first data and a step (S310) of reclassifying a target product as defective or normal based on the similarity.

[0101] In the following, a method for detecting battery defects will be described in detail with reference to FIG. 1 and FIG. 7 to 8. For the convenience of explanation, content that overlaps with previously described content will be omitted or briefly explained.

[0102] In step S300, the battery defect detection device (101) can determine the similarity between the target image and the first data. Step S300 may substantially correspond to a specific example of step S130 of FIG. 7. Step S300 may be performed by the processor (120) of the battery defect detection device (101).

[0103] In step S300, the battery defect detection device (101) can determine the similarity between the target image and the first data according to a preset method. According to an embodiment, the battery defect detection device (101) can determine the similarity between the target image and the first data by quantifying it. According to an embodiment disclosed in this document, the battery defect detection device (101) determines the similarity between the target image and the first data, but is not limited thereto and may also determine the dissimilarity between the target image and the first data.

[0104] In step S310, the battery defect detection device (101) can reclassify the target product as defective or normal based on the similarity between the target image and the first data. The battery defect detection device (101) can reclassify the target product as defective or normal based on the degree of similarity between the target image and the first data. According to an embodiment, the battery defect detection device (101) can set a reference value for the similarity between the target image and the first data. According to an embodiment, the battery defect detection device (101) can reclassify the target product as defective or normal by comparing the similarity determined in step S300 with the reference value.

[0105] According to an embodiment, the battery defect detection device (101) can reclassify a target product corresponding to a target image as normal when the similarity between the target image and the first data is greater than or equal to a preset reference value.

[0106] According to an embodiment, the battery defect detection device (101) can reclassify a target product corresponding to a target image as defective when the similarity between the target image and the first data is less than a preset threshold value. At this time, the target product reclassified as defective may correspond to a new type of defective product.

[0107] FIG. 10 is a drawing showing a battery defect detection method according to one embodiment disclosed in this document.

[0108] Referring to FIG. 10, a battery defect detection method according to one embodiment disclosed in this document may further include a step of determining the similarity between a target image and a first data (S400), a step of comparing the similarity between the target image and the first data with a preset reference value (S410), a step of determining the target image as second data when the similarity between the target image and the first data is less than the reference value (S420), and / or a step of including the second data in the training data (S430).

[0109] In the following, a method for detecting battery defects will be described in detail with reference to FIG. 1 and FIG. 7 to 9. For the convenience of explanation, content that overlaps with previously described content will be omitted or briefly explained.

[0110] Step S400 may correspond to substantially the same step as Step S300 of FIG. 4. Steps S410, S420, and / or S430 may be performed simultaneously with Step S310 of FIG. 4 or before and after Step S310.

[0111] In step S410, the battery defect detection device (101) can determine whether the similarity between the target product and the first data is less than a preset reference value. The reference value may correspond to a value set as a standard for the battery defect detection device (101) to reclassify the target product as normal or defective based on the similarity between the target image and the first data.

[0112] In step S410, the target image that the battery defect detection device (101) determines to have a similarity greater than or equal to a reference value may include an image of a good product. According to an embodiment, the battery defect detection device (101) may not perform additional operations on the target image that it determines to have a similarity greater than or equal to a reference value. According to another embodiment, the battery defect detection device (101) may add and include the target image that it determines to have a similarity greater than or equal to a reference value in the first data. According to the embodiments disclosed in this document, the battery defect detection device (101) can increase the precision of defect detection by adding and learning the target image that it determines to have a similarity greater than or equal to a reference value in the first data.

[0113] In step S420, the battery defect detection device (101) may determine a target image that is determined to have a similarity less than a reference value as second data. According to an embodiment, the target image determined as second data may include an image of a new type of defective product. According to an embodiment, in step S420, the battery defect detection device (101) may determine a target image corresponding to a new type of defective product as second data.

[0114] In step S430, the battery defect detection device (101) may include second data in the training data. According to an embodiment, the second data may correspond to an image for the battery defect detection device (101) to classify a target product as defective. According to an embodiment, the second data may include an image of a new type of defective product that is not included in the previously stored training data. In step S430, the battery defect detection device (101) may learn the new type of defect by including the second data in the training data.

[0115] FIG. 11 is a drawing showing a battery defect detection system according to one embodiment disclosed in this document.

[0116] Referring to FIG. 11, a battery defect detection system (1000) according to one embodiment disclosed in this document may include an inspection device (1100) and a new defect detection device (1200). Here, the inspection device (1000) may be based on an artificial intelligence model (161), and the new defect detection device (1100) may be based on an artificial intelligence model (165).

[0117] In the following, a battery defect detection system will be described in detail with reference to FIG. 1 and FIG. 7 to 10. For the convenience of explanation, content that overlaps with previously described content will be omitted or briefly explained.

[0118] The inspection device (1100) can acquire an image of a target product and classify the target product using the image of the target product and previously stored training data. According to an embodiment, the inspection device (1100) may include an image acquisition unit (1110) and a processor (1120).

[0119] The inspection device (1100) can acquire an image of a target product. According to an embodiment, the inspection device (1100) can acquire an image of a target product through an image acquisition unit (1110). According to an embodiment, the image acquisition unit (1110) can acquire an image from a remote image acquisition device that captures the target product to acquire an image of the target product. According to an embodiment, the remote image acquisition device can be implemented as a camera. According to an embodiment, the remote image acquisition device can acquire an image of the target product in multiple channels.

[0120] The inspection device (1100) can classify a target product using an image of the target product and previously stored training data. According to an embodiment, the inspection device (1100) can classify a target product through a processor (1120). According to an embodiment, the processor (1120) can compare an image of the target product with previously stored training data. According to an embodiment, the processor (1120) can classify a target product based on a deep learning model to which previously stored training data is applied.

[0121] The processor (1120) can classify a target product as normal or defective based on previously stored training data. The processor (1120) classifies the target product based on previously stored training data and can set the target product classified as normal as the target product. According to an embodiment, the processor (1120) can classify a new type of defective product among the target products that is not included in the training data as normal, and the target product may include both a good product and a new type of defective product.

[0122] The new defect detection device (1200) can acquire a target image corresponding to a target product classified as normal by the inspection device (1100), and determine the characteristics of the target image using the first data for classifying the target product as normal among the training data and the target image. According to an embodiment, the new defect detection device (1200) may include a target image acquisition unit (1210) and a processor (1220).

[0123] The new defect detection device (1200) can receive a target product from the inspection device (1100). The new defect detection device (1200) can acquire a target image corresponding to the target product. According to an embodiment, the new defect detection device (1200) can acquire a target image through a target image acquisition unit (1210).

[0124] According to an embodiment, the new defect detection device (1200) may receive a target image corresponding to a target product from the inspection device (1100). The target image acquisition unit (1210) may select a target image corresponding to a target product from among the images of the target product previously acquired by the inspection device (1100). According to an embodiment, the target image acquisition unit (1210) may be integrated with a processor (1220) and implemented as a single module.

[0125] According to another embodiment, the target image acquisition unit (1210) can acquire a target image by taking an image of a target product received from the inspection device (1100). In this case, the target image acquisition unit (1210) can be implemented as a camera.

[0126] The new defect detection device (1200) can determine the characteristics of the target image using the first data and the target image. The first data may refer to data among the training data for the inspection device (1100) to classify the target product as normal. According to an embodiment, the new defect detection device (1200) can share the training data with the inspection device (1100). According to an embodiment, the new defect detection device (1200) can determine the characteristics of the target image through the processor (1220).

[0127] According to an embodiment, the processor (1220) can determine whether the target image corresponds to the first data. According to another embodiment, the processor (1220) can determine the similarity between the target image and the first data.

[0128] The processor (1220) can determine the characteristics of the target image and detect a new type of defect based on the determined characteristics. According to an embodiment, the processor (1220) can determine the target image as second data if the target image does not correspond to first data. According to another embodiment, the processor (1220) can determine the target image as second data if the similarity between the target image and the first data is less than a preset threshold value. According to the embodiments disclosed in this document, the target product corresponding to the target image determined to be second data may correspond to a new type of defective product. The processor (1220) can detect a new type of defective product by detecting the target product corresponding to the target image.

[0129] According to an embodiment, the processor (1220) may include the second data in the training data. By including the second data in the training data, the processor (1220) can learn a new type of defect. The new defect detection device (1200) may share the second data with the inspection device (1100). According to an embodiment, the inspection device (1100) can learn the second data, and by learning the second data, the precision of defect detection can be increased.

[0130] Terms such as "include," "compose," or "have" as used above, unless specifically stated otherwise, mean that the relevant component may be inherent; therefore, they should be interpreted as allowing for the inclusion of additional components rather than excluding them. All terms, including technical or scientific terms, have the same meaning as generally understood by those skilled in the art to which the embodiments disclosed in this document pertain, unless otherwise defined. Commonly used terms, such as those defined in advance, should be interpreted in accordance with their meaning in the context of the relevant technology and, unless explicitly defined in this document, should not be interpreted in an ideal or overly formal sense.

[0131] The foregoing description is merely an illustrative explanation of the technical concept disclosed in this document, and a person skilled in the art to which the embodiments disclosed in this document pertain can make various modifications and variations within the scope of the essential characteristics of the embodiments disclosed in this document. Accordingly, the embodiments disclosed in this document are intended to explain, not limit, the technical concept of the embodiments disclosed in this document, and the scope of the technical concept disclosed in this document is not limited by these embodiments. The scope of protection of the technical concept disclosed in this document shall be interpreted by the claims below, and all technical concepts within an equivalent scope shall be interpreted as being included within the scope of rights of this document.

Claims

Claim 1 A battery defect detection device comprises a communication module, a processor, a first artificial intelligence model, a second artificial intelligence model, and a memory for storing instructions, wherein when the instructions are executed by the processor, the battery defect detection device acquires an image of a target product using the communication module, inputs the image of the target product into the first artificial intelligence model to classify the target product, and if the target product is classified as normal, inputs the image of the target product into the second artificial intelligence model to determine whether the image of the target product corresponds to the first data among the training data of the first artificial intelligence model for classifying the target product as normal, and the image of the target product is acquired in multiple channels by an image acquisition device. Claim 2 A battery defect detection device according to claim 1, wherein the battery defect detection device is configured to reclassify the target product as defective when the instructions are executed by the processor and the image does not correspond to the first data. Claim 3 A battery defect detection device according to claim 1, wherein, when the instructions are executed by the processor, the battery defect detection device is configured to determine the image as second data for classifying the target product as defective if the image does not correspond to the first data. Claim 4 A battery defect detection device according to claim 3, wherein the instructions are configured such that when executed by the processor, the battery defect detection device trains the first artificial intelligence model based on the second data. Claim 5 delete Claim 6 A battery defect detection device according to claim 1, wherein, when the instructions are executed by the processor, the battery defect detection device is configured to obtain a similarity between the image and the first data through the second artificial intelligence model and reclassify the target product as defective or normal based on the similarity. Claim 7 A battery defect detection device according to claim 6, wherein, when the instructions are executed by the processor, the battery defect detection device determines the image as second data for classifying the target product as defective when the similarity of the image is less than a preset reference value. Claim 8 A battery defect detection device according to claim 7, wherein the instructions are configured such that when executed by the processor, the battery defect detection device learns the first artificial intelligence model based on the second data. Claim 9 A method of operation for a battery defect detection device comprising: acquiring an image of a target product; inputting the image of the target product into a first artificial intelligence model to classify the target product; and, when the target product is classified as normal, inputting the image of the target product into a second artificial intelligence model to determine whether the image of the target product corresponds to a first data among the training data of the first artificial intelligence model for classifying the target product as normal, wherein the image of the target product is acquired through multiple channels. Claim 10 A method of operation according to claim 9, further comprising the operation of reclassifying the target product as defective when the image does not correspond to the first data. Claim 11 A method of operation according to claim 9, further comprising the operation of determining the image as second data for classifying the target product as defective when the image does not correspond to the first data. Claim 12 A method of operation according to claim 11, further comprising the operation of training the first artificial intelligence model based on the second data. Claim 13 delete Claim 14 A method of operation according to claim 9, further comprising the operation of obtaining a similarity between the image and the first data through the second artificial intelligence model, and the operation of reclassifying the target product as defective or normal based on the similarity. Claim 15 A method of operation according to claim 14, further comprising the operation of determining the image as second data for classifying the target product as defective when the similarity of the image is less than a preset reference value. Claim 16 A method of operation according to claim 15, further comprising the operation of training the first artificial intelligence model based on the second data. Claim 17 A battery defect detection system comprising: a battery defect detection device that acquires an image of a target product and classifies the target product using a first artificial intelligence model trained with the image of the target product and previously stored training data; and a novel defect detection device that, when the target product is classified as normal, inputs the image of the target product into a second artificial intelligence model to determine whether the image of the target product corresponds to first data among the training data of the first artificial intelligence model for classifying the target product as normal, and an image acquisition device that acquires the image of the target product in a multi-channel manner.

Images