Apparatus and method for inspecting battery

Abstract

A method for inspecting a battery, according to an embodiment of the present invention, relates to a method for inspecting the quality of a battery in a manufacturing process, the method comprising the steps of: acquiring an image capturing at least a portion of the exterior of the battery to preprocess the image; detecting one or more defect candidate regions in the preprocessed image by using one or more detection algorithms; extracting position information of the defect candidate regions and shape feature information of the defect candidate regions; and inputting, into a pre-trained machine learning model, information related to the detection algorithms, the position information of the defect candidate regions, and the shape feature information of the defect candidate regions, to determine whether corresponding defect candidate shapes are defective.

Description

Battery inspection device and method

[0001] This application claims the benefit of the filing date of Korean Patent Application No. 10-2024-0125276 filed with the Korean Intellectual Property Office on September 13, 2024, and all contents disclosed in the document of said Korean patent application are incorporated into this specification.

[0002] The present invention relates to a battery inspection device and method, and more specifically, to a battery inspection device and method for inspecting the quality of a battery during a manufacturing process.

[0003] Rechargeable batteries are batteries that can be reused through charging even after discharge. They can be utilized as energy sources for small devices such as mobile phones, tablet PCs, and vacuum cleaners, and are also used as medium-to-large energy sources for personal mobility devices, automobiles, and Energy Storage Systems (ESS) for smart grids. Depending on the system requirements, rechargeable batteries are used in the form of assemblies, such as battery modules in which multiple battery cells are connected in series and parallel, or battery packs in which battery modules are connected in series and parallel.

[0004] Batteries can be manufactured through a process that broadly includes electrode, assembly, formation, and other processes, and defects are checked for on the manufactured battery cells or the products of each process using inspection equipment. A commonly used technique for defect inspection is visual inspection, which can be performed by acquiring images of the product using an optical system and detecting defects using a detection algorithm.

[0005] However, defect inspection methods utilizing images of the battery exterior carry a potential risk of under-inspection. Furthermore, modifying inspection algorithms to adjust the detection capability of the tester results incurs time and cost losses, as the process involves repetitive attempts and error corrections that require manually changing variables to verify detection results.

[0006] The objective of the present invention to solve the above-mentioned problems is to provide a battery inspection method for inspecting the quality of a battery during the manufacturing process.

[0007] Another objective of the present invention to solve the above-mentioned problems is to provide a battery inspection device that utilizes the battery inspection method.

[0008] A battery inspection method according to an embodiment of the present invention for achieving the above objective is a method for inspecting the quality of a battery during a manufacturing process, and may include the steps of: acquiring and preprocessing an image of the exterior of at least a portion of the battery; detecting one or more candidate defect regions in the preprocessed image using one or more detection algorithms; extracting location information of the candidate defect regions and shape feature information of the candidate defect regions; and inputting the information related to the detection algorithm, the location information of the candidate defect regions, and the shape feature information of the candidate defect regions into a pre-trained machine learning model to perform a defect determination for the corresponding candidate defect shape.

[0009] The shape feature information of the above defect candidate region may include one or more of the area, length, diameter, roundness, defect factor, and circularity of each defect candidate shape.

[0010] The information related to the detection algorithm may include the type of detection algorithm used to detect the defect candidate region and information about the lighting used when applying the detection algorithm.

[0011] The machine learning model above can be pre-trained using training data including information related to the detection algorithm, location information of the defect candidate region, and shape feature information for the defect candidate region.

[0012] The shape feature information used as training data can be matched with a label of the defect type for the corresponding shape.

[0013] The above detection algorithm is a rule-based algorithm and may include one or more of a dent detection algorithm, a contamination detection algorithm, a scratch detection algorithm, and a leak detection algorithm.

[0014] The battery inspection method described above may further include the step of visualizing and outputting the defect determination flow of the machine learning model; and the step of ranking the degree of contribution to the defect determination and visualizing and outputting the defect determination result.

[0015] The above machine learning model may include one or more of a decision tree model, a random forest model, and a Support Vector Model (SVM).

[0016] The information regarding the lighting above may include information about the location where the lighting is installed and the type of lighting.

[0017] The battery inspection method described above may further include a step of performing a final defect determination using the area and length of feature information for a defect candidate shape determined as defective by the machine learning model.

[0018]

[0019] A battery inspection device according to an embodiment of the present invention for achieving the above other purpose is a device for inspecting the quality of a battery during a manufacturing process, and may include at least one processor; a memory for storing at least one command executed through the at least one processor, and the at least one command may include: a command to acquire and preprocess an image of the exterior of at least a part of the battery; a command to detect one or more defect candidate regions in the preprocessed image using one or more detection algorithms; a command to extract location information of the defect candidate regions and shape feature information of the defect candidate regions; and a command to input information related to the detection algorithm, location information of the defect candidate regions, and shape feature information of the defect candidate regions into a pre-trained machine learning model to perform a defect determination for the corresponding defect candidate shape.

[0020] The shape feature information of the above defect candidate region may include one or more of the area, length, diameter, roundness, defect factor, and circularity of each defect candidate shape.

[0021] The information related to the detection algorithm may include the type of detection algorithm used to detect the defect candidate region and information about the lighting used when applying the detection algorithm.

[0022] The machine learning model above can be pre-trained using training data including information related to the detection algorithm, location information of the defect candidate region, and shape feature information for the defect candidate region.

[0023] The shape feature information used as training data can be matched with a label of the defect type for the corresponding shape.

[0024] The above detection algorithm is a rule-based algorithm and may include one or more of a dent detection algorithm, a contamination detection algorithm, a scratch detection algorithm, and a leak detection algorithm.

[0025] The above at least one command may further include a command to visualize and output the defect determination flow of the machine learning model; and a command to rank the degree of contribution to defect determination and visualize and output the defect determination result.

[0026] The above machine learning model may include one or more of a decision tree model, a random forest model, and a Support Vector Model (SVM).

[0027] The information regarding the lighting above may include information about the location where the lighting is installed and the type of lighting.

[0028] The above at least one command may further include a command to perform a final defect determination using the area and length of feature information for a defect candidate shape determined as a defect by the machine learning model.

[0029] According to the embodiment of the present invention as described above, the accuracy of the inspection for determining whether a battery being manufactured is defective can be increased. Accordingly, the risk to the producer can be reduced by reducing misclassification and over-inspection.

[0030] In addition, since defect judgment priorities and models are visualized, maintenance of the inspection algorithm is easy, and time and cost losses for modifying audit logic can be reduced.

[0031] FIG. 1 is a flowchart of the operation sequence of a battery inspection method according to an embodiment of the present invention.

[0032] FIG. 2 is a diagram illustrating the process of detecting defective candidate regions and extracting shape feature information according to an embodiment of the present invention.

[0033] Figure 3 shows an example of training data for a machine learning model according to an embodiment of the present invention.

[0034] Figure 4 shows an example of a rule-based algorithm used for defect detection according to an embodiment of the present invention.

[0035] FIG. 5 illustrates an example of label matching that matches defect shape information and defect type according to an embodiment of the present invention.

[0036] Figure 6 is a diagram showing an example of a defect indication method using machine learning among defect detection methods according to an embodiment of the present invention.

[0037] FIG. 7 is a perspective view of an imaging device according to an embodiment of the present invention.

[0038] FIG. 8 is a cross-sectional view of an imaging device according to an embodiment of the present invention.

[0039] FIG. 9a is a drawing showing the external appearance of an imaging device according to another embodiment of the present invention.

[0040] FIG. 9b is a multi-channel illumination configuration diagram of an imaging device according to another embodiment of the present invention.

[0041] FIG. 10 is a block diagram of a battery inspection device according to an embodiment of the present invention.

[0042] The present invention is susceptible to various modifications and may have various embodiments; specific embodiments are illustrated in the drawings and described in detail in the detailed description. However, this is not intended to limit the invention to specific embodiments, and it should be understood that the invention includes all modifications, equivalents, and substitutions that fall within the spirit and scope of the invention. Similar reference numerals have been used for similar components in the description of each drawing.

[0043] Terms such as first, second, A, B, etc., may be used to describe various components, but said components shall not be limited by said terms. These terms are used solely for the purpose of distinguishing one component from another. For example, without departing from the scope of the present invention, the first component may be named the second component, and similarly, the second component may be named the first component. The term "and / or" includes a combination of a plurality of related described items or any of a plurality of related described items.

[0044] When it is stated that one component is "connected" or "connected" to another component, it should be understood that while it may be directly connected or connected to that other component, there may also be other components in between. On the other hand, when it is stated that one component is "directly connected" or "directly connected" to another component, it should be understood that there are no other components in between.

[0045] The terms used in this application are used merely to describe specific embodiments and are not intended to limit the invention. The singular expression includes the plural expression unless the context clearly indicates otherwise. In this application, terms such as "comprising" or "having" are intended to specify the presence of the features, numbers, steps, actions, components, parts, or combinations thereof described in the specification, and should be understood as not precluding the existence or addition of one or more other features, numbers, steps, actions, components, parts, or combinations thereof.

[0046] Unless otherwise defined, all terms used herein, including technical or scientific terms, have the same meaning as generally understood by those skilled in the art to which the present invention pertains. Terms such as those defined in commonly used dictionaries should be interpreted as having a meaning consistent with their meaning in the context of the relevant technology, and should not be interpreted in an ideal or overly formal sense unless explicitly defined in this application.

[0047]

[0048] Hereinafter, preferred embodiments according to the present invention will be described in detail with reference to the attached drawings.

[0049]

[0050] FIG. 1 is a flowchart of the operation sequence of a battery inspection method according to an embodiment of the present invention.

[0051] The subject of inspection according to the present invention may be a battery of the type such as cylindrical, pouch, or prismatic during the manufacturing process, in particular, a cylindrical battery of various specifications. For example, a cylindrical battery may include an electrode assembly, a battery can, a plate-shaped cap, and terminals. A cylindrical battery may additionally include an insulating gasket, a current collector, an insulator, a sealing gasket, etc.

[0052] A method for inspecting the appearance of a product according to an embodiment of the present invention acquires and preprocesses an image of the object to be inspected (S100, S110). An image of at least a part of the battery to be inspected can be acquired through an imaging device, i.e., an optical system, and a plurality of regions of interest (ROI) may be included within a single image.

[0053] The image preprocessing process (S110) is a process of emphasizing features of an image of a target object before inspection to facilitate inspection, and is a process of converting an image acquired through an imaging device so that defects can be easily identified through a process of correction or synthesis. The image preprocessing process may also include a process of generally amplifying the brightness of the image of the target object. Methods such as image synthesis, binarization, normalization, blurring, and stretching may be used in the preprocessing process.

[0054] Here, image synthesis can be effectively utilized for detecting defects such as dents and scratches by expressing depth in the image using multi-channel images, which will be examined later. In particular, photometric stereo synthesis and phase shift synthesis can be used as image synthesis methods.

[0055] Furthermore, binarization is a process that resets pixel values ​​for objects and backgrounds within an image to only two values, 1 and 0; through image correction via binarization, defects on the product's exterior can be transformed to become more visible. Additionally, normalization is a process that distributes pixel values ​​concentrated in specific parts of an image evenly across the entire area. Blurring, also known as smoothing, is used to soften rough-looking images or as a preprocessing step to remove the effects of noise present in the image. Moreover, stretching can be used as a preprocessing method to compensate for the state of distorted images.

[0056] For the preprocessed image, one or more defect candidate regions are detected using a rule-based detection algorithm (S120), and location information of the detected defect candidate regions and shape feature information of the defect candidate regions are extracted (S130).

[0057] Here, the feature information of a defect candidate shape according to an embodiment of the present invention may include one or more of the area, length, diameter, circularity, degree of ellipse, and degree of squareness of each defect candidate shape. In addition, the location information of a defect candidate region may include the x-coordinate and y-coordinate of each region on an image.

[0058] Subsequently, information related to the detection algorithm, location information of the defect candidate region, and shape feature information for the defect candidate region are input into a pre-trained machine learning model, and the defect status of the shape can be determined based on the output of the machine learning model (S200).

[0059] The information regarding the detection algorithm may include the type of detection algorithm used to detect the defect candidate and information regarding the lighting used when applying the detection algorithm. Here, the detection algorithm may use a rule-based detection algorithm. Rule-based detection algorithms that can effectively detect typical defect types of batteries may include a dent detection algorithm, a contamination detection algorithm, a scratch detection algorithm, a leak detection algorithm, etc.

[0060] In addition, the information regarding the lighting used when applying the detection algorithm may include information about the location where the lighting is installed and the type of lighting.

[0061] Meanwhile, although not illustrated in FIG. 1, the inspection method according to the present invention may further include the step of visualizing and outputting the flow of determining whether a machine learning model is defective in relation to the determination of defects by the machine learning model; and the step of ranking the degree of contribution to the determination of defects and visualizing and outputting the result of determining whether a defect is defective.

[0062] Meanwhile, the machine learning model is in a pre-trained state using a large amount of training data that includes information related to the detection algorithm used to detect defect candidate shapes, location information of the defect candidate region, and shape feature information for the defect candidate region.

[0063] For defect candidate shapes determined as defective by the learning model (example of S210), additional defect determination can be performed using the area and length of feature information (S300). On the other hand, shapes that are not determined as defective by the learning model are determined to be good products (S220).

[0064] In the additional defect determination process (S300), a defect can be determined based on the length and area (or size) of the defect candidate shape. The size of the shape can be determined by the outer diameter of the maximum circumscribed circle or by using the area of ​​the defect. Generally, since foreign matter or contamination has the characteristic of appearing darker compared to the surrounding area, the size and area of ​​the dark region can be used to ultimately determine whether it is a defect by utilizing this characteristic. For example, since scratches or abrasions have a straight line shape, a linearity factor (struct factor) that quantifies the straight-line characteristics can be used to distinguish them from foreign matter such as threads.

[0065]

[0066] FIG. 2 is a diagram illustrating the process of detecting defective candidate regions and extracting shape feature information according to an embodiment of the present invention.

[0067] As described above, to detect defect candidate regions according to an embodiment of the present invention, an external image of an inspection target is obtained using an imaging device. For one or more potential candidate regions (20) included in the obtained image, defect candidate regions are detected using one or more rule-based detection algorithms. Here, multiple detection algorithms may be applied to a single potential candidate region.

[0068] Rule-based detection algorithms may include, for example, dent detection algorithms, contamination detection algorithms, scratch detection algorithms, and leak detection algorithms. Dent detection algorithms are suitable for detecting defects such as dents or scratches. Contamination detection algorithms are suitable for detecting various types of contamination. Additionally, scratch detection algorithms are suitable for detecting various types of scratch defects, and leak detection algorithms are suitable for detecting defects such as liquid leakage.

[0069] If there is a detection algorithm that executes multiple detection algorithms on a potential candidate region (20) and determines the region as defective, shape feature information can be extracted for the detection algorithm and the shape detected by the algorithm.

[0070] Here, the shape feature values ​​of the defect candidate region may include the shape area, width, height, maximum diameter, curcularity, defect factor, rectangularity, struct factor, etc.

[0071] In addition, brightness information of candidate defect areas is collected during the process of extracting shape feature information. Since changes in shape feature values ​​of candidate defect areas can occur due to variations in lighting conditions over time, brightness information can be one of the important factors in determining defects.

[0072]

[0073] Figure 3 shows an example of training data for a machine learning model according to an embodiment of the present invention.

[0074] As explained above, the machine learning model according to the present invention can be pre-trained using training data including information related to a detection algorithm, location information of a defect candidate region, and shape feature information of a defect candidate region.

[0075] The rule-based algorithm used as a detection algorithm in the present invention is intended to detect defect types based on the form, depth, length, height, etc. of contamination. By training a machine learning model using the name of the rule-based algorithm used for contamination detection and the location and feature information of the battery area where the defect occurred, and by determining the defect using the trained model, the accuracy of the determination can be improved.

[0076] Here, the battery image may include several regions of interest (ROIs) distinguished according to shape characteristics. In the example of FIG. 3, the contamination detection area of ​​the battery may include battery region 1 related to the top surface of the battery, battery region 2 related to the side surface of the battery, battery region 3 related to the bottom surface of the battery, etc.

[0077] At this time, depending on the position and angle at which the battery is placed during inspection, and the location of the battery area, the type of lighting used can be an important factor in detection. The lighting used for battery inspection may consist of, for example, 5 to 10 lights depending on the battery area, and different lighting may be used for rule-based algorithms or detection areas.

[0078] In the table of Figure 3, the name of the algorithm used for each defect detection and the light number can be identified. The location where the light is installed and the type of light can be identified by each light number.

[0079]

[0080] Figure 4 shows an example of a rule-based algorithm used for defect detection according to an embodiment of the present invention.

[0081] Referring to the table in Fig. 4, rule-based detection algorithms may include, for example, a dent detection algorithm, a contamination detection algorithm, a scratch detection algorithm, a leak detection algorithm, etc.

[0082] Dent detection algorithms are suitable for detecting defects such as dents or scratches. Dents can include various types, such as general dents, floor dents, and extensive scratches, and different detection algorithms and lighting can be used depending on each type. Additionally, contamination detection algorithms are suitable for detecting various types of contamination, such as general contamination and heavy contamination. Scratch detection algorithms are suitable for detecting various types of scratch defects, and leak detection algorithms are suitable for detecting defects such as liquid leakage.

[0083] In the present invention, the accuracy of defect detection can be improved by using a detection algorithm and lighting suitable for various types of defects, and by using information regarding the detection algorithm and lighting used for training a machine learning model.

[0084]

[0085] FIG. 5 illustrates an example of label matching that matches defect shape information and defect type according to an embodiment of the present invention.

[0086] The label matching method illustrated in FIG. 5 can be implemented in the form of separate software and performed by a labeling system. The labeling system according to the present invention can be implemented in the form of various types of computing devices. Software implemented to perform labeling according to the present invention can be stored in the memory of the labeling system and performed by a processor.

[0087] Referring to the left screen of FIG. 5, the labeling system according to an embodiment of the present invention is a labeling system that matches shape feature information with defect types by utilizing images of defect candidate groups captured under various lighting conditions. The labeling system according to an embodiment of the present invention can select a defect type (contamination, particle, scratch, dent, rust) corresponding to a presented defect candidate image (indicated by a unique ID), and can label the selected defect type to the shape feature information possessed by each defect candidate image. As previously explained, the labeled shape feature information can be used as training data to train a machine learning model, along with information related to a detection algorithm and location information of the defect candidate region.

[0088] Meanwhile, the machine learning model according to the present invention may include one or more of a decision tree model, a random forest model, and a Support Vector Model (SVM).

[0089]

[0090] Figure 6 is a diagram showing an example of a defect indication method using machine learning among defect detection methods according to an embodiment of the present invention.

[0091] As described above, according to the defect detection method of the embodiment of the present invention, one or more defect candidate regions are detected in a preprocessed image using one or more rule-based detection algorithms (using a 2D appearance inspector), and location information of the defect candidate regions and shape feature information of the defect candidate regions are extracted (S61).

[0092] According to an embodiment of the present invention, a machine learning model that has been trained can perform a defect determination for a defect candidate shape when information related to a detection algorithm, location information of the defect candidate region, and shape feature information for the defect candidate region are input.

[0093] At this time, the embodiment of the present invention can visualize the flow of a machine learning model determining defects (S62). More specifically, the flow of how defect candidate feature values ​​in a pre-trained machine learning model are subdivided and determined as specific defects according to a range of values ​​can be visualized in real time with partial emphasis.

[0094] Afterwards, when the defect judgment is completed, the final judgment result is displayed, and a certain number of defect information that contributed significantly to the process of subdividing defects into specific defects during the machine learning judgment process is graphed, and the degree of contribution to the defect judgment for the visualized defect shape features is ranked and visualized (S63).

[0095]

[0096] FIG. 7 is a perspective view of an imaging device according to an embodiment of the present invention, and FIG. 8 is a cross-sectional view of an imaging device according to an embodiment of the present invention.

[0097] Referring to FIGS. 7 and 8, an imaging device (200) according to one embodiment of the present invention may include at least one of a housing (210), a first light (220), a first mirror (231), a second mirror (232), a camera (240), a third mirror (250), a second light (260), a transfer means (270), and an inversion means (280).

[0098] In the example of FIG. 7, the housing (210) is positioned on the side of the battery (10) and surrounds at least the first region (A1) of the battery (10). For example, as shown in FIG. 7, the battery (10) may be positioned in a lying position with its side facing upward, and the housing (210) may be positioned on the top of the battery (10). The shape of the housing (210) is not particularly limited, and the inner surface of the housing (210) may have a dome shape.

[0099] The first light (220) is positioned on the inner surface of the housing (210) and irradiates light onto the battery (10). Although the first light (220) is shown in FIG. 8 as being positioned protruding from the inner surface of the housing (210), the first light (220) may be positioned embedded in the inner surface of the housing (210) so that at least a portion is exposed. Specifically, the first light (220) irradiates light onto the side of the battery (10). The first light (220) can irradiate light onto the entire area of ​​the first region (A1) of the battery (10). As described above, the inner surface of the housing (210) may have a dome shape, and the first light (220) positioned on the inner surface of the housing (210) may be positioned in an overall round shape to irradiate light onto the side of the battery (10).

[0100] The first mirror (231) and the second mirror (232) may each be placed on each side of the battery (10) to reflect a portion of the first area (A1) of the battery (10). Referring to FIG. 8, the first mirror (231) may be placed on the left side of the battery (10), and the second mirror (232) may be placed at an angle on the right side of the battery (10).

[0101] Meanwhile, the camera (240) captures at least a portion of the first region (A1) of the battery (10), an image reflected in the first mirror (231), and an image reflected in the second mirror (232). That is, the camera (240) captures at least a portion of the first region (A1) of the battery (10). At this time, the portion of the first region (A1) reflected in the first mirror (231) or the second mirror (232) includes all areas of the first region (A1) that cannot be captured by the camera (240) when the first mirror (231) or the second mirror (232) is not present. Therefore, with the presence of the first mirror (231) and the second mirror (232), the image captured by the camera (240) can include an image of the entire area of ​​the first region (A1) of the battery (10).

[0102] Meanwhile, if the camera (240) is not positioned facing the side of the battery (10) to directly photograph at least a portion of the first region (A1) of the battery (10), the image reflected in the first mirror (231), and the image reflected in the second mirror (232), the imaging device (200) may further include a third mirror (250). In this case, the camera (240) indirectly photographs at least a portion of the first region (A1), the image reflected in the first mirror (231), and the image reflected in the second mirror (232) by photographing the image reflected in the third mirror (250). That is, the camera (240) photographs an image including at least a portion of the first region (A1) of the battery (10) reflected in the third mirror (250), the image reflected in the first mirror (231), and the image reflected in the second mirror (232).

[0103] A third mirror (250) is positioned at an angle on the side (200S) of the battery (10) to reflect at least a portion of the first area (A1) of the battery (10), the image reflected in the first mirror (231), and the image reflected in the second mirror (232). Based on the drawing, the third mirror (250) may be positioned at an angle facing between the Y direction (Y) and the Z direction (Z). The third mirror (250) may be positioned so that at least a portion of the first area (A1) of the battery (10), the image reflected in the first mirror (231), and the image reflected in the second mirror (232) are incident and reflected at an angle of about 45°. At this time, the camera (240) may be positioned facing the light reflected from the third mirror (250).

[0104] Meanwhile, the imaging device (200) may include a second light (260) in addition to the first light (220). The second light (260) is positioned on the third mirror (250) to irradiate light onto the battery (10) and may be a coaxial light.

[0105] Additionally, a transfer means (270) transfers the battery (10) in the axial direction (X) of the battery (10) so that the battery (10) passes through the housing (210). The battery (10) is transferred via the transfer means (270), and after reaching the inspection area within the housing (210), it stops for a certain period of time and is photographed by a camera (240), thereby allowing a side inspection of the battery (10) to be performed. The battery (10) that has been inspected is then transferred to a subsequent process via the transfer means (270).

[0106] Meanwhile, the inversion means (280) flips over the battery (10) that has been photographed by the camera (240). By repositioning the battery (10) in a state where it has been rotated 180°, the inversion means (280) allows for a side inspection to be performed on the opposite side of the battery (10) where the side inspection was not performed, using the same device as the aforementioned imaging device (200).

[0107] An imaging device (200) according to one embodiment of the present invention includes a first mirror (231) and a second mirror (232), thereby enabling the acquisition of an entire image of a first area (A1), which is more than half of the side (200S) of the battery (10), through a single shot. That is, an external inspection of the entire area of ​​the side of the battery (10) is possible through two inspections, one for one side and the other for the other side of the battery (10). Furthermore, sufficient visibility of defects can be secured even in areas far from the center of the first area (A1) that is the subject of inspection. Through this, an imaging device capable of performing a side external inspection without rotating or rolling the cylindrical battery can be provided. Additionally, an imaging device capable of preventing image noise and improving image quality and the detection capability of the inspection device can be provided.

[0108] Meanwhile, the imaging device according to the embodiment described through FIGS. 7 and 8 is a structure suitable for inspecting the side of a battery and is presented as an example of an imaging device according to the present invention.

[0109]

[0110] FIG. 9a is a drawing showing the external appearance of an imaging device according to another embodiment of the present invention, and FIG. 9b is a multi-channel lighting configuration diagram of an imaging device according to another embodiment of the present invention.

[0111] The imaging device illustrated in FIG. 9a has a structure suitable for acquiring images of the upper and lower surfaces of a battery, such as the rivet and vent portions of the battery cell, and can acquire various images under various lighting conditions using a multi-illumination device as illustrated in FIG. 9b. Here, the multi-illumination device may include, for example, dome illumination, low angle illumination, coaxial illumination, ring illumination, and dark field illumination. Thus, the imaging device according to the present invention may include a structure for capturing images of various parts of a battery cell, such as the upper and lower surfaces of the battery and the side of the battery.

[0112]

[0113] FIG. 10 is a block diagram of a battery inspection device according to an embodiment of the present invention.

[0114] A battery inspection device (100) according to an embodiment of the present invention may include at least one processor (110), a memory (120) that stores at least one command executed through the processor, and a transmitting and receiving device (130) that is connected to a network and performs communication.

[0115] Here, the at least one command may include: a command to acquire and preprocess an image of the exterior of at least a portion of the battery; a command to detect one or more defect candidate regions in the preprocessed image using one or more detection algorithms; a command to extract location information of the defect candidate regions and shape feature information of the defect candidate regions; and a command to input the information related to the detection algorithm, the location information of the defect candidate regions, and the shape feature information of the defect candidate regions into a pre-trained machine learning model to perform a defect determination for the corresponding defect candidate shapes.

[0116] The shape feature information of the above defect candidate region may include one or more of the area, length, diameter, roundness, defect factor, and circularity of each defect candidate shape.

[0117] The information related to the detection algorithm may include the type of detection algorithm used to detect the defect candidate region and information about the lighting used when applying the detection algorithm.

[0118] The machine learning model above can be pre-trained using training data including information related to the detection algorithm, location information of the defect candidate region, and shape feature information for the defect candidate region.

[0119] The shape feature information used as training data can be matched with a label of the defect type for the corresponding shape.

[0120] The above detection algorithm is a rule-based algorithm and may include one or more of a dent detection algorithm, a contamination detection algorithm, a scratch detection algorithm, and a leak detection algorithm.

[0121] The above at least one command may further include a command to visualize and output the defect determination flow of the machine learning model; and a command to rank the degree of contribution to defect determination and visualize and output the defect determination result.

[0122] The above machine learning model may include one or more of a decision tree model, a random forest model, and a Support Vector Model (SVM).

[0123] The information regarding the lighting above may include information about the location where the lighting is installed and the type of lighting.

[0124] The above at least one command may further include a command to perform a final defect determination using the area and length of feature information for a defect candidate shape determined as a defect by the machine learning model.

[0125] A battery inspection device according to an embodiment of the present invention may be linked with or include an imaging device that captures the appearance of a target object. The imaging device according to the present invention may be provided in the same or similar form as the device shown in FIGS. 7, 8, 9a, and 9b.

[0126] The battery inspection device according to an embodiment of the present invention may also be linked with, for example, the labeling system according to the present invention described through FIG. 5, or include such a system.

[0127] The battery inspection device (100) may also further include an input interface device (140), an output interface device (150), a storage device (160), etc. Each component included in the battery inspection device (100) may be connected by a bus (170) to communicate with each other.

[0128] The storage device (160) can store at least one image produced at each stage of the inspection method according to the present invention, such as a plurality of images captured by the imaging device, an image preprocessed therefrom, and a postprocessed image. The image stored in the storage device (160) may be provided to at least one processor (110) if necessary during the execution of the inspection method according to the present invention.

[0129] Here, the processor (110) may mean a central processing unit (CPU), a graphics processing unit (GPU), or a dedicated processor on which methods according to embodiments of the present invention are performed. The memory (or storage device) may be composed of at least one of a volatile storage medium and a non-volatile storage medium. For example, the memory may be composed of at least one of read-only memory (ROM) and random access memory (RAM).

[0130]

[0131] The operation of the method according to an embodiment of the present invention can be implemented as a computer-readable program or code on a computer-readable recording medium. A computer-readable recording medium includes all types of recording devices in which data that can be read by a computer system is stored. Additionally, the computer-readable recording medium may be distributed across networked computer systems, allowing computer-readable programs or code to be stored and executed in a distributed manner.

[0132] Some aspects of the invention have been described in the context of a device, but may also be described according to a corresponding method, wherein a block or device corresponds to a method step or a feature of a method step. Similarly, aspects described in the context of a method may also be represented by a corresponding block or item or a feature of a corresponding device. Some or all of the method steps may be performed by (or using) a hardware device, such as, for example, a microprocessor, a programmable computer, or an electronic circuit. In some embodiments, one or more of the most important method steps may be performed by such a device.

[0133] Although the present invention has been described with reference to preferred embodiments, those skilled in the art will understand that various modifications and changes can be made to the invention without departing from the spirit and scope of the invention as described in the following claims.

Claims

1. A method for inspecting the quality of a battery during the manufacturing process, A step of acquiring and preprocessing an image of at least a portion of the exterior of the battery; A step of detecting one or more defective candidate regions in the preprocessed image using one or more detection algorithms; A step of extracting location information of the above-mentioned defective candidate region and shape feature information of the above-mentioned defective candidate region; and A battery inspection method comprising the step of inputting information related to the detection algorithm, location information of the defect candidate region, and shape feature information of the defect candidate region into a pre-trained machine learning model to perform a defect determination for the corresponding defect candidate shape.

2. In Claim 1, The shape feature information of the above defective candidate region is, A battery inspection method comprising one or more of the area, length, diameter, defect factor, and circularity of each defect candidate shape.

3. In Claim 1, The information regarding the above detection algorithm is, A battery inspection method comprising information on the type of detection algorithm used to detect the above-mentioned defective candidate region and the lighting used when applying the above-mentioned detection algorithm.

4. In Claim 1, The above machine learning model is, A battery inspection method pre-trained using training data including information related to the detection algorithm, location information of the defect candidate region, and shape feature information for the defect candidate region.

5. In Claim 4, A battery inspection method in which the shape feature information used as training data is matched with a label of a defect type for the corresponding shape.

6. In Claim 1, The above detection algorithm is a rule-based algorithm, and, A battery inspection method comprising one or more of a dent detection algorithm, a contamination detection algorithm, a scratch detection algorithm, and a leak detection algorithm.

7. In Claim 1, A step of visualizing and outputting the defect determination flow of the machine learning model above; and A battery inspection method comprising the additional step of ranking the degree of contribution to defect determination and visualizing and outputting the defect determination result.

8. In Claim 1, The above machine learning model is, A battery inspection method comprising one or more of a decision tree model, a random forest model, and a Support Vector Model (SVM).

9. In Claim 3, A battery inspection method in which information regarding the above lighting includes information regarding the location where the lighting is installed and the type of lighting.

10. In Claim 1, A battery inspection method further comprising the step of performing a final defect determination using the area and length of feature information for a defect candidate shape determined as defective by the machine learning model.

11. A device for inspecting the quality of a battery during the manufacturing process, At least one processor; It includes a memory that stores at least one instruction executed through the above-mentioned at least one processor, and The above at least one command is, A command to acquire and preprocess an image capturing the exterior of at least a portion of the above battery; An instruction to detect one or more defective candidate regions in the preprocessed image using one or more detection algorithms; A command to extract location information of the above-mentioned defective candidate region and shape feature information for the above-mentioned defective candidate region; and A battery inspection device comprising a command to input information related to the detection algorithm, location information of the defect candidate region, and shape feature information of the defect candidate region into a pre-trained machine learning model to perform a defect determination for the corresponding defect candidate shape.

12. In Claim 11, The shape feature information of the above defective candidate region is, A battery inspection device comprising one or more of the area, length, diameter, defect factor, and circularity of each defect candidate shape.

13. In Claim 11, The information regarding the above detection algorithm is, A battery inspection device comprising information on the type of detection algorithm used to detect the above-mentioned defective candidate region and the lighting used when applying the above-mentioned detection algorithm.

14. In Claim 11, The above machine learning model is, A battery inspection device pre-trained using training data including information related to the detection algorithm, location information of the defect candidate region, and shape feature information for the defect candidate region.

15. In Claim 14, A battery inspection device in which the shape feature information used as training data is matched with a label of a defect type for the corresponding shape.

16. In Claim 11, The above detection algorithm is a rule-based algorithm, and, A battery inspection device comprising one or more of a dent detection algorithm, a contamination detection algorithm, a scratch detection algorithm, and a leak detection algorithm.

17. In Claim 11, The above at least one command is, A command to visualize and output the defect determination flow of the above machine learning model; and A battery inspection device further comprising a command to rank the degree of contribution to defect judgment and to visualize and output the defect judgment result.

18. In Claim 11, The above machine learning model is, A battery inspection device comprising one or more of a decision tree model, a random forest model, and a Support Vector Model (SVM).

19. In Claim 13, A battery inspection device that includes information regarding the lighting, such as the location where the lighting is installed and the type of lighting.

20. In Claim 11, The above at least one command is, A battery inspection device further comprising a command to perform a final defect determination using the area and length of feature information for a defect candidate shape determined as defective by the machine learning model.

Images