Track type inspection method for temperature in battery production formation stage

Abstract

The invention provides a temperature rail type inspection method for a battery production formation stage. The method comprises the following steps: acquiring a visible light image and an infrared image with equal focal lengths of a target battery; removing the background of the visible light image to obtain a contour visible light image of the target battery; fusing the infrared image and a contour visible light image of the target battery into a fused image, and segmenting the fused image according to the contour of the target battery in the contour visible light image to obtain a segmented image; and according to data information of the BMS system, thermodynamic inverse solution is carried out on the segmented image, and internal temperature information and external temperature information of the target battery are obtained. According to the method, the visible light image is obtained, the distance parameter of the infrared image is obtained through correction according to the parameters such as the focal length when the visible light image is obtained, the background of the visible light image is removed, the image is fused, the background of the fused image is segmented according to the contour of the target battery, and the fused image of the target battery is accurately obtained. And inverse solution is carried out to obtain battery temperature information.

Description

Technical Field

[0001] The invention relates to the field of battery production, and in particular to a temperature track inspection method in a battery production formation stage. Background Art

[0002] With the advantages of long storage and cycle life, high energy density, wide application range, and no memory storage, lithium-ion batteries have been rapidly developed and applied in the fields of new energy vehicles, ships, aircraft, energy storage power stations, etc. In the production process of lithium batteries, the battery temperature distribution is prone to unevenness during the formation stage, and there will be a large number of batteries piled up during the storage stage of lithium batteries, which may cause safety hazards to the production process. In the formation stage of a single battery, the battery will be given a certain current to stimulate the active substances of the positive and negative electrodes of the battery, and finally the electrochemical process that enables the battery to have discharge capacity is called formation. After the battery production is completed, the battery will be stored in a large number of warehouses. In these two processes, dangers are prone to occur, so it is necessary to monitor the temperature distribution of the battery in the formation and storage of the battery in real time to ensure the safety and stability of the battery formation and storage process.

[0003] At present, the temperature monitoring of battery modules mainly adopts contact measurement methods such as thermocouples, thermistors and fiber grating sensors. The required wire harness increases sharply with the increase of temperature monitoring points, which is difficult to apply to occasions with a large number of batteries. Infrared thermal imaging not only has a wide measurement range and high efficiency, but also can monitor the temperature of all positions on the battery surface. However, the measurement accuracy of infrared thermal imaging is greatly affected by the environment, measurement distance, etc., and the resolution is low, and it cannot distinguish the background temperature and battery temperature well. Summary of the invention

[0004] In order to solve the problems existing in the background technology, the present invention proposes a temperature orbital inspection method in the formation stage of battery production.

[0005] A method for temperature orbital inspection in the formation stage of battery production, comprising the steps of:

[0006] S100, acquiring a visible light image and an infrared image of the target battery at equal focal lengths;

[0007] S200, performing image recognition on the visible light image, and obtaining a visible light image of the outline of the target battery after removing the background;

[0008] S300, after fusing the infrared image with the contour visible light image of the target battery into a fused image, segment the fused image according to the contour of the target battery in the contour visible light image to obtain a segmented image;

[0009] S400: According to the data information of the BMS system, the segmented image is thermodynamically inversely solved to obtain the internal temperature information and the external temperature information of the target battery.

[0010] Based on the above, in step S100, after the visible light image of the target battery is acquired, the distance parameter of the infrared image to be acquired is corrected according to the focal length information when the visible light image is acquired.

[0011] Based on the above, step S100 includes the following steps:

[0012] S110, constructing a track frame and a mobile module slidably arranged on the track frame corresponding to the formation cabinet, and symmetrically arranging a visible light camera and an infrared thermal imager on the mobile module;

[0013] S120, a stop point is set on the track frame corresponding to the center of each battery position in the formation cabinet, and the symmetrical central axis of the visible light camera and the infrared thermal imager on the mobile module is set corresponding to the stop point;

[0014] S130, after the mobile module moves to the target battery position, the visible light camera automatically focuses on the target battery and takes a photo to obtain a visible light image, and sends the focal length information to the PC end;

[0015] S140, the PC side corrects the distance parameter information from the infrared thermal imager to the target battery according to the focal length information, and then obtains an infrared image;

[0016] S150: The PC receives the acquired visible light image and infrared image.

[0017] Based on the above, in step S200, the battery outline and the background are segmented in the visible light image by using fuzzy clustering.

[0018] Based on the above, in step S300, a cascaded fusion module and a semantic segmentation module are constructed. The fusion module fuses the contour visible light image and the infrared image, and the semantic segmentation module segments the fused image to obtain a segmented image; wherein, after obtaining the fused image, the fused image is compensated by the content loss module; after obtaining the segmented image, the semantic loss module uses semantic loss to guide high-level semantic information to flow back to the fusion module for compensation.

[0019] Based on the above, the content loss module includes the content loss function, the content loss function L con Including strength loss L int and texture loss L texture :

[0020] L con =L int +aL texture

[0021]

[0022]

[0023] Where H and W are the height and width of the image, respectively, ‖·‖2 represents the L2 norm, max(·) represents the element-by-element maximum value selection, ▽ represents the Canny gradient operator, |·| represents the absolute operation, and I f ∈R H×W×3 represents the fused image, I ir ∈R H×W×1 represents the infrared image after registration, I vi ∈R H×W×3 Represents the registered visible light image.

[0024] Based on the above, the semantic loss module includes the semantic loss function, the semantic loss function L semantic Including the main semantic loss function L main And the auxiliary semantic loss function L aux :

[0025] L semantic =L main +λL aux

[0026]

[0027]

[0028] Among them, λ is a constant that balances the main semantic loss and the auxiliary semantic loss, L so ∈R H×W×C Represents the segmentation label L s ∈(1,C) H×W The transformed one-hot encoding; That is I s ∈R H×W×C , represents the segmentation result output by the segmentation network, and C represents the number of channels of the image; That is I sa ∈R H×W×C , represents the auxiliary segmentation result.

[0029] The present invention has outstanding substantial features and significant progress compared to the prior art. Specifically, the present invention acquires a visible light image, corrects the distance parameter of the infrared image according to the focal length and other parameters when the visible light image is acquired, removes the background of the visible light image, fuses the image, and performs background segmentation on the fused image according to the contour of the target battery. After accurately acquiring the fused image of the target battery, the temperature information of the battery is obtained by reverse analysis, which has the advantages of saving time and effort. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 It is a schematic block diagram of the process of the present invention;

[0031] Figure 2 It is a structural schematic diagram of the mobile module of the present invention;

[0032] Figure 3 A schematic diagram of constructing a global heterogeneous graph of the present invention;

[0033] Figure 4 It is a structural schematic diagram of the fusion module of the present invention;

[0034] Figure 5 It is a structural schematic diagram of the GRDB module of the present invention.

[0035] Explanation of the reference numerals: 1. Visible light camera; 2. Infrared thermal imager. DETAILED DESCRIPTION

[0036] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0037] like Figure 1 As shown, a temperature orbital inspection method for a battery production formation stage includes the following steps: S100, obtaining an isofocal visible light image and an infrared image of a target battery; S200, performing image recognition on the visible light image, removing the background and obtaining a contour visible light image of the target battery; S300, fusing the infrared image with the contour visible light image of the target battery into a fused image, and then segmenting the fused image according to the contour of the target battery in the contour visible light image to obtain a segmented image; S400, performing thermodynamic inverse analysis on the segmented image according to data information of a BMS system to obtain internal temperature information and external temperature information of the target battery.

[0038] Specifically, a track frame and a track system are provided corresponding to the formation cabinet, and a slide rail is provided on the track frame, on which a mobile module is slidably provided, and the track system is used to control the movement of the mobile module on the track frame, take pictures of different batteries, etc. (the track frame, the control method of the track system, the mobile structure of the mobile module, etc. are all common structures in the prior art and will not be described in detail). A fixed battery position for placing the battery is provided on the formation cabinet corresponding to the battery, and a stop point of the mobile module is provided on the vertical center line of the corresponding battery position on the track frame by storing experimental data or providing inductive detection elements such as contact switches. Figure 2As shown, the mobile module is symmetrically provided with a visible light camera 1 and an infrared thermal imager 2. When the mobile module stops at the corresponding battery position and takes pictures, the symmetry axes of the visible light camera 1 and the infrared thermal imager 2 are aligned with the vertical center line (the line connecting the symmetry axes and the vertical center line is perpendicular to the track frame). When taking pictures, the vertical distance between the visible light camera and the target battery is consistent with the vertical distance between the infrared thermal imager and the target battery. A wireless communication module such as a WIFI module can be provided in the mobile module for communication connection with a host computer such as a PC, and the shooting parameters, visible light images, infrared images, etc. of the visible light camera are sent to the host computer.

[0039] After the mobile module moves to the battery position, the visible light camera automatically focuses and takes pictures to obtain the visible light image of the target battery, and sends the focal length and other shooting parameter information to the PC. According to the focal length and the distance from the mobile module to the battery, the distance parameter from the infrared thermal imager to the battery is corrected, and the infrared thermal imager takes an infrared image of the target battery. The obtained visible light image and infrared image are sent to the PC. When the mobile module can move, the mobile module moves to other battery positions, and then repeats the process of focusing and acquiring images, so as to achieve the function of real-time monitoring.

[0040] Since there is a certain distance between the mobile module and the battery, and the infrared image resolution is low, the battery and the background cannot be distinguished in the infrared image. The surface temperature of each single battery during the formation process is distinguished by fusion of infrared image and visible light image.

[0041] In the visible light image, fuzzy clustering is used for image recognition and segmentation to remove the background in the visible light image and leave the outline of the target battery. The outline visible light image can then be fused with the infrared image. Since the positions of the visible light camera and the infrared thermal imager are fixed on the mobile module, the outline of the battery is determined in the infrared image based on the affine transformation. In this embodiment, a cascaded fusion module and a semantic segmentation module are constructed. The fusion module fuses the outline visible light image and the infrared image, and the semantic segmentation module segments the fused image to obtain a segmented image. After obtaining the fused image, the fused image is compensated by a content loss module. After obtaining the segmented image, the semantic loss module uses semantic loss to guide high-level semantic information back to the fusion module for compensation, such as Figure 3 As shown in Figure 2, the performance of the high-level visual task of segmenting batteries and background on fused images is effectively improved.

[0042] In order to promote the fusion model of visible light and infrared images to integrate more meaningful image information and improve visual quality and quantitative indicators, a content loss module is designed. The content loss module includes content loss function, content loss function L con Including strength loss L int and texture loss Ltexture The exact definition of content loss is as follows:

[0043] L con =L int +aL texture

[0044] Among them, L int Constrain the overall apparent intensity of the fused image, L texture Texture forces the fused image to contain finer-grained texture details.

[0045] Strength loss L int The difference between the fused image and the source image at the pixel level is measured, and the intensity loss of infrared and visible light images is defined as:

[0046]

[0047] Where H and W are the height and width of the image, respectively, ∥□∥2 represents the L2 norm, and max(·) represents the element-by-element maximum selection. However, the intensity loss only provides a coarse-grained distribution constraint for model learning. The texture loss is introduced to force the fused image to contain more fine-grained texture information. The texture loss is defined as:

[0048]

[0049] Where ▽ represents the Canny gradient operator, which measures the fine-grained texture information of the image. |·| represents the absolute operation, I f ∈R H×W×3 represents the fused image, I ir ∈R H×W×1 represents the infrared image after registration, I vi ∈R H×W×3 Represents the registered visible light image.

[0050] The segmentation module performs real-time semantic segmentation on the fused image and outputs the segmentation result I s ∈R H×W×C And auxiliary segmentation result I sa ∈R H×W×C The semantic loss includes the main semantic loss and the auxiliary semantic loss. The main semantic loss is expressed as:

[0051]

[0052] The auxiliary semantic loss is expressed as:

[0053]

[0054] Where L so ∈R H×W×C Represents the segmentation label Ls ∈(1,C) H×W The transformed one-hot encoding,; That is I s ∈R H×W×C , represents the segmentation result output by the segmentation network, and C represents the number of channels of the image; That is I sa ∈R H ×W×C , represents the auxiliary segmentation result.

[0055] Finally, the semantic loss is expressed as:

[0056] L semantic =L main +λL aux

[0057] where λ is a constant that balances the main semantic loss and the auxiliary semantic loss.

[0058] The training of the image fusion model is guided by a joint loss, defined as:

[0059] L joint =L con +βL semantic

[0060] where β is a hyperparameter that characterizes the semantic importance of missing semantics.

[0061] Fusion modules such as Figure 4 and Figure 5 As shown, Conv is a convolutional layer, and ReLU is a RectifiedLinearUnit (rectified linear unit), which is a commonly used activation function and is defined as follows:

[0062] f(x)=max(0,x)

[0063] The numbers "16, 32, 48" in the figure indicate the number of feature maps. The fusion module includes a feature extraction part and an image reproduction part. Since the battery volume is relatively small compared to the formation cabinet, the feature extraction part contains two GRDBs (gradient residual dense blocks) to extract fine-grained features, and the features output by the two GRDBs are combined and input into the image reproduction part. Figure 5 As shown in the figure, the main line uses dense connections, the residual flow integrates the gradient operator, and the Canny operator is used as the gradient operator. The gradient is calculated while the filtering operation is included. Except for the 1×1 convolution layer, the step size is set to 1, so there is no need to introduce downsampling operations, and the size of the fused image is consistent with the source image.

[0064] Based on the segmented image obtained after fusion and semantic segmentation, combined with the data information of the BMS system, the segmented image is inversely solved through the existing thermodynamic model to obtain the internal and external temperatures of the battery module. The internal and external temperature of each battery can be quickly and accurately determined to ensure that the formation stage and storage stage of the battery production process are within a safe range, preventing dangerous situations such as explosions of unqualified battery products.

[0065] It will be apparent to those skilled in the art that the invention is not limited to the details of the exemplary embodiments described above and that the invention can be implemented in other specific forms without departing from the spirit or essential features of the invention. Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description, and it is intended that all variations falling within the meaning and scope of the equivalent elements of the claims be included in the invention. Any reference numeral in a claim should not be considered as limiting the claim to which it relates.

Claims

1. A temperature orbital inspection method for battery production formation stage, characterized in that: Includes steps: S100, acquiring a visible light image and an infrared image of the target battery at equal focal lengths; S200, performing image recognition on the visible light image, and obtaining a visible light image of the outline of the target battery after removing the background; S300, after fusing the infrared image with the contour visible light image of the target battery into a fused image, segment the fused image according to the contour of the target battery in the contour visible light image to obtain a segmented image; S400: According to the data information of the BMS system, the segmented image is thermodynamically inversely solved to obtain the internal temperature information and the external temperature information of the target battery.

2. The method for temperature orbital inspection of battery production formation stage according to claim 1, characterized in that: In step S100, after the visible light image of the target battery is acquired, the distance parameter of the infrared image to be acquired is corrected according to the focal length information when the visible light image is acquired.

3. The method for temperature orbital inspection of battery production formation stage according to claim 1, characterized in that: Step S100 includes the steps of: S110, constructing a track frame and a mobile module slidably arranged on the track frame corresponding to the formation cabinet, and symmetrically arranging a visible light camera and an infrared thermal imager on the mobile module; S120, a stop point is set on the track frame corresponding to the center of each battery position in the formation cabinet, and the symmetrical central axis of the visible light camera and the infrared thermal imager on the mobile module is set corresponding to the stop point; S130, after the mobile module moves to the target battery position, the visible light camera automatically focuses on the target battery and takes a photo to obtain a visible light image, and sends the focal length information to the PC end; S140, the PC side corrects the distance parameter information from the infrared thermal imager to the target battery according to the focal length information, and then obtains an infrared image; S150: The PC receives the acquired visible light image and infrared image.

4. The method for temperature orbital inspection of battery production formation stage according to claim 1, characterized in that: In step S200, the battery outline and the background are segmented in the visible light image by using fuzzy clustering.

5. The method for temperature orbital inspection of battery production formation stage according to claim 1, characterized in that: In step S300, a cascaded fusion module and a semantic segmentation module are constructed. The fusion module fuses the contour visible light image and the infrared image, and the semantic segmentation module segments the fused image to obtain a segmented image. After obtaining the fused image, the fused image is compensated by the content loss module. After obtaining the segmented image, the semantic loss module uses semantic loss to guide high-level semantic information to flow back to the fusion module for compensation.

6. The method for temperature orbital inspection of battery production formation stage according to claim 5, characterized in that: The content loss module includes the content loss function, the content loss function L con Including strength loss L int and texture loss L texture : <h2 style=";text-align:left;direction:ltr">L<h2 style=";text-align:left;direction:ltr"> con <h2 style=";text-align:left;direction:ltr"> =L<h2 style=";text-align:left;direction:ltr"> int <h2 style=";text-align:left;direction:ltr"> +aL<h2 style=";text-align:left;direction:ltr"> texture Where H and W are the height and width of the image, respectively, ‖·‖2 represents the L2 norm, max(·) represents the element-by-element maximum value selection, ▽ represents the Canny gradient operator, |·| represents the absolute operation, and I f ∈R H×W×3 represents the fused image, I ir ∈R H×W×1 represents the infrared image after registration, I vi ∈R H×W×3 Represents the registered visible light image.

7. The method for temperature orbital inspection of battery production formation stage according to claim 5, characterized in that: The semantic loss module includes the semantic loss function, the semantic loss function L semantic Including the main semantic loss function L main and auxiliary semantic loss function L aux : THE semantic =L main +λL aux Among them, λ is a constant that balances the main semantic loss and the auxiliary semantic loss, L so ∈R H×W×C Represents the segmentation label L s ∈(1,C) H×W The transformed one-hot encoding, That is I s ∈R H×W×C , represents the segmentation result output by the segmentation network; C represents the number of channels of the image; That is I sa ∈R H×W×C , represents the auxiliary segmentation result.

Images