New energy battery cell transfer system with posture adjustment function

Abstract

The invention relates to the technical field of intelligent control, and particularly discloses a new energy battery cell transfer system with a posture adjustment function, which utilizes a computer vision technology based on deep learning to carry out image analysis on a real-time posture image of a new energy battery cell in a transfer process. Global attitude semantic information and edge contour detail information of the battery cell are captured, and the current attitude of the battery cell is comprehensively described based on interactive fusion characteristics of the global attitude semantic information and the edge contour detail information, so that an attitude adjustment instruction of the battery cell is generated. Therefore, intelligent identification and accurate adjustment of battery cell postures can be realized, the battery cells are effectively prevented from being damaged in the transfer process, the transfer efficiency is improved, and the safety of battery production, storage and use processes is ensured.

Description

Technical Field

[0001] The present application relates to the field of intelligent control technology, and more specifically, to a new energy battery cell transport system with a posture adjustment function. Background Art

[0002] With the rapid development of the new energy industry, new energy batteries, as its core components, are increasingly used in electric vehicles, energy storage systems, portable electronic devices and other fields. In the production, assembly and subsequent maintenance of new energy batteries, the transportation of battery cells is an indispensable link.

[0003] The new energy battery cell transfer system is mainly used to achieve fast, accurate and safe transfer of battery cells in the battery production, storage, charging and replacement process, and usually includes transfer equipment, storage facilities, control systems and safety monitoring devices. During the cell transfer process, due to the irregular shape of the cell and the precision limitations of the robotic arm, conveyor belt or other transfer equipment, the battery cell is prone to posture deviations, such as tilting and flipping, which will not only affect the battery assembly efficiency, but may also damage the internal structure of the cell, reducing battery performance and service life.

[0004] However, many existing new energy battery cell transfer systems still require manual participation in the detection and adjustment of the battery cell posture, which not only increases production costs and low adjustment efficiency, but also increases the possibility of human error.

[0005] Therefore, a new energy battery cell transport system with posture adjustment function is expected. Summary of the invention

[0006] In order to solve the above technical problems, the present application is proposed. The embodiment of the present application provides a new energy battery cell transportation system with a posture adjustment function, which uses computer vision technology based on deep learning to perform image analysis on real-time posture images of new energy battery cells during transportation, captures the global posture semantic information and edge contour detail information of the battery cell, and comprehensively describes the current posture of the battery cell based on the interactive fusion characteristics of the two, so as to generate posture adjustment instructions for the battery cell. In this way, intelligent recognition and precise adjustment of the battery cell posture can be achieved, effectively preventing damage to the battery cell during transportation, improving transportation efficiency, and ensuring the safety of battery production, storage and use.

[0007] Accordingly, according to one aspect of the present application, a new energy battery cell transport system with a posture adjustment function is provided, comprising:

[0008] The battery cell posture real-time monitoring module is used to obtain the real-time posture image of the new energy battery cell during the transportation process;

[0009] A multi-level image feature extraction module, used for performing multi-level feature extraction on the real-time posture image of the battery cell to obtain a semantic feature vector of the battery cell posture image and a HOG feature vector of the battery cell posture image;

[0010] A feature interactive fusion module, used for interactively fusing the semantic feature vector of the battery cell posture image and the HOG feature vector of the battery cell posture image to obtain a multi-scale representation feature vector of the battery cell posture;

[0011] The adjustment instruction generation module is used to generate a cell posture adjustment instruction based on the multi-scale characterization feature vector of the cell posture.

[0012] In the above-mentioned new energy battery cell transport system with posture adjustment function, the multi-level image feature extraction module includes: an image preprocessing unit, which is used to perform image preprocessing on the real-time posture image of the battery cell to obtain a detail-enhanced battery cell posture image; a battery cell posture semantic feature extraction unit, which is used to extract the global semantic features of the detail-enhanced battery cell posture image to obtain the semantic feature vector of the battery cell posture image; and a battery cell posture edge detail feature extraction unit, which is used to extract the edge detail features of the detail-enhanced battery cell posture image to obtain the HOG feature vector of the battery cell posture image.

[0013] In the above-mentioned new energy battery cell transport system with posture adjustment function, the image preprocessing unit is used to: perform gamma correction on the real-time posture image of the cell to obtain the detail-enhanced cell posture image.

[0014] In the above-mentioned new energy battery cell transport system with posture adjustment function, the cell posture semantic feature extraction unit is used to: input the detail enhanced cell real-time posture image into the image semantic feature extractor based on the VIT model to obtain the cell posture image semantic feature vector.

[0015] In the above-mentioned new energy battery cell transport system with posture adjustment function, the cell posture edge detail feature extraction unit is used to: calculate the gradient direction histogram of the detail enhanced cell real-time posture image to obtain the cell posture image HOG feature vector.

[0016] In the above-mentioned new energy battery cell transport system with posture adjustment function, the feature interaction fusion module is used to: input the semantic feature vector of the cell posture image and the HOG feature vector of the cell posture image into a multi-scale posture feature combiner based on a bilinear pooling layer to obtain the multi-scale representation feature vector of the cell posture.

[0017] In the above-mentioned new energy battery cell transport system with posture adjustment function, the adjustment instruction generation module includes: a feature optimization unit, which is used to perform feature optimization on the multi-scale characterization feature vector of the battery cell posture to obtain an optimized multi-scale characterization feature vector of the battery cell posture; a decoding unit, which is used to pass the optimized multi-scale characterization feature vector of the battery cell posture through a decoder-based posture adjustment instruction generator to obtain the battery cell posture adjustment instruction, and the battery cell posture adjustment instruction includes a rotation angle, a flipping direction and a translation direction.

[0018] In the above-mentioned new energy battery cell transport system with posture adjustment function, the feature optimization unit is used to: perform feature compatibility evaluation on the multi-scale characterization feature vector of the battery cell posture based on implicit correlation to obtain the optimized multi-scale characterization feature vector of the battery cell posture.

[0019] In the above-mentioned new energy battery cell transport system with posture adjustment function, the feature optimization unit is used to: calculate the mean and variance of the multi-scale characterization feature vector of the battery cell posture; calculate the difference between the eigenvalue of the predetermined position of the multi-scale characterization feature vector of the battery cell posture and the mean of the multi-scale characterization feature vector of the battery cell posture to obtain a first intermediate calculation value; multiply the decoding parameter matrix by the multi-scale characterization feature vector of the battery cell posture, and then subtract the multi-scale characterization feature vector of the battery cell posture by the position to obtain a first intermediate vector, and calculate the Froben i of the first intermediate vector. us norm to obtain a first norm value; add the first norm value to the variance of the multi-scale characterization feature vector of the battery cell posture and then take the square root to obtain a second intermediate calculation value; divide the first intermediate calculation value by the second intermediate calculation value to obtain a third intermediate calculation value; calculate the square of each position in the multi-scale characterization feature vector of the battery cell posture and then add them up to obtain a first sum value; take the square root of the first sum value and divide it by the scale of the decoding parameter matrix to calculate the natural exponential function value to obtain a fourth intermediate calculation value, wherein the scale of the decoding parameter matrix is ​​equal to the product of the width value of the decoding parameter matrix and its height value; multiply the fourth intermediate calculation value by the weighted hyperparameter and then add the third intermediate calculation value to obtain the eigenvalue of the predetermined position of the optimized battery cell posture multi-scale characterization feature vector.

[0020] Compared with the prior art, the new energy battery cell transport system with posture adjustment function provided by the present application uses computer vision technology based on deep learning to perform image analysis on the real-time posture image of the new energy battery cell during the transport process, captures the global posture semantic information and edge contour detail information of the battery cell, and comprehensively describes the current posture of the battery cell based on the interactive fusion characteristics of the two, thereby generating posture adjustment instructions for the battery cell. In this way, intelligent recognition and precise adjustment of the battery cell posture can be achieved, effectively preventing damage to the battery cell during the transport process, improving transport efficiency, and ensuring the safety of the battery production, storage and use process. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other purposes, features and advantages of the present application will become more apparent. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0022] Figure 1 It is a block diagram of a new energy battery cell transport system with posture adjustment function according to an embodiment of the present application.

[0023] Figure 2 Schematic diagram of the architecture of a new energy battery cell transport system with posture adjustment function according to an embodiment of the present application.

[0024] Figure 3 It is a block diagram of a multi-level image feature extraction module in a new energy battery cell transportation system with posture adjustment function according to an embodiment of the present application.

[0025] Figure 4 It is a block diagram of an adjustment instruction generation module in a new energy battery cell transport system with a posture adjustment function according to an embodiment of the present application. DETAILED DESCRIPTION

[0026] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings 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.

[0027] The terms "first", "second", "third", "fourth", etc. (if any) in the specification and claims of the present invention and the above drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the numbers used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in sequences other than those illustrated or described herein.

[0028] It should be understood that in various embodiments of the present invention, the size of the sequence number of each process does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0029] It should be understood that in the present invention, "including" and "having" and any variations are intended to cover non-exclusive inclusions. For example, a process, method, system, product or apparatus comprising a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to these processes, methods, products or apparatuses.

[0030] Depending on the context, "if" as used herein may be interpreted as "when" or "when" or "in response to determining" or "in response to detecting."

[0031] The technical solution of the present invention is described in detail with specific embodiments below. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described in detail in some embodiments.

[0032] As mentioned in the above background technology, in the context of the increasing global environmental awareness, the new energy battery industry is undergoing a profound change. However, despite the significant progress in battery technology, some links in the production process of battery cells still have the problem of inefficiency. In particular, the cell transport system still generally relies on manual detection and adjustment of the cell posture. The transport system with manual participation not only requires a large amount of human resources, which significantly increases the production cost, but also in the high-intensity repetitive labor, the fatigue and distraction of workers may lead to a decrease in adjustment efficiency. More importantly, this manual operation mode inevitably introduces the risk of human error, which may pose a potential threat to the quality and safety of the battery. In response to the above technical problems, the technical concept of the present application is to use computer vision technology based on deep learning to perform image analysis on the real-time posture image of the new energy battery cell during the transportation process, capture the global posture semantic information and edge contour detail information of the cell, and comprehensively describe the current posture of the battery cell based on the interactive fusion characteristics of the two, so as to generate the posture adjustment instruction of the battery cell. In this way, the intelligent identification and precise adjustment of the battery cell posture can be achieved, effectively preventing the battery cell from being damaged during transportation, improving transportation efficiency, and ensuring the safety of battery production, storage and use.

[0033] Figure 1 It is a block diagram of a new energy battery cell transport system with posture adjustment function according to an embodiment of the present application. Figure 2 Schematic diagram of the architecture of a new energy battery cell transport system with posture adjustment function according to an embodiment of the present application. Figure 1 and Figure 2 As shown, according to the embodiment of the present application, the new energy battery cell transportation system 100 with posture adjustment function includes: a cell posture real-time monitoring module 110, which is used to obtain the real-time posture image of the new energy battery cell during the transportation process; a multi-level image feature extraction module 120, which is used to perform multi-level feature extraction on the real-time posture image of the cell to obtain a cell posture image semantic feature vector and a cell posture image HOG feature vector; a feature interactive fusion module 130, which is used to interactively fuse the cell posture image semantic feature vector and the cell posture image HOG feature vector to obtain a cell posture multi-scale representation feature vector; an adjustment instruction generation module 140, which is used to generate a cell posture adjustment instruction based on the cell posture multi-scale representation feature vector.

[0034] In the above-mentioned new energy battery cell transport system 100 with posture adjustment function, the battery cell posture real-time monitoring module 110 is used to obtain the real-time posture image of the new energy battery cell during the transportation process. It should be understood that the battery cell may change its posture due to vibration, collision or improper operation during the transportation process, such as angle tilt, front and back flipping, etc., which will affect the stability and safety of the battery cell. In order to be able to make timely adjustments to the improper posture of the battery cell, in the technical solution of the present application, the image information of the battery cell is captured in real time to understand the current posture of the battery cell, so as to analyze and identify the battery cell posture using an image processing algorithm, and generate posture adjustment instructions for abnormal postures of the battery cell without manual intervention to prevent potential safety problems.

[0035] In the above-mentioned new energy battery cell transport system 100 with posture adjustment function, the multi-level image feature extraction module 120 is used to perform multi-level feature extraction on the real-time posture image of the battery cell to obtain the semantic feature vector of the battery cell posture image and the HOG feature vector of the battery cell posture image. It should be understood that, considering that in the process of recognizing the posture of the battery cell, the global information and local detail information in the real-time posture image of the battery cell are of great significance for the description of the posture of the battery cell. Among them, the global semantic features usually include information such as the overall shape, size and orientation of the battery cell, which helps to understand the global posture of the battery cell; while the edge detail features can reflect the contour, edge and possible local deformation of the battery cell, which helps to identify the slight posture changes and edge states of the battery cell. Therefore, in the technical solution of the present application, a multi-level image feature extraction method is adopted, and the global semantic features and edge detail features of the detail-enhanced battery cell posture image are extracted respectively, so as to provide a richer feature description for subsequent posture recognition.

[0036] Figure 3 FIG. 1 is a block diagram of a multi-level image feature extraction module in a new energy battery cell transport system with a posture adjustment function according to an embodiment of the present application. Figure 3 As shown, the multi-level image feature extraction module 120 includes: an image preprocessing unit 121, used to perform image preprocessing on the real-time posture image of the battery cell to obtain a detail-enhanced battery cell posture image; a battery cell posture semantic feature extraction unit 122, used to extract the global semantic features of the detail-enhanced battery cell posture image to obtain the semantic feature vector of the battery cell posture image; a battery cell posture edge detail feature extraction unit 123, used to extract the edge detail features of the detail-enhanced battery cell posture image to obtain the HOG feature vector of the battery cell posture image.

[0037] Specifically, the image preprocessing unit 121 is used to perform image preprocessing on the real-time posture image of the battery cell to obtain a detail-enhanced battery cell posture image. It should be understood that, considering that in practical applications, the real-time posture image of the battery cell may have problems such as uneven illumination and noise interference, which affect the accuracy of subsequent feature extraction. Therefore, before performing image feature extraction, the real-time posture image of the battery cell is further image preprocessed to enhance the local details of the image and improve the contrast of the image, so that the subsequent feature extraction is more accurate and stable. In a specific example of the present application, the processing method for performing image preprocessing on the real-time posture image of the battery cell is to perform gamma correction on the real-time posture image of the battery cell to obtain the detail-enhanced battery cell posture image. Specifically, gamma correction is a common image enhancement technology, which can effectively improve the local contrast of the image by adjusting the brightness curve of the image, so that the detail features of the image are more prominent, which is conducive to subsequent feature extraction.

[0038] Specifically, the cell posture semantic feature extraction unit 122 is used to extract the global semantic features of the detail enhanced cell posture image to obtain the cell posture image semantic feature vector. In a specific example of the present application, the processing method for extracting the global semantic features of the detail enhanced cell posture image is to input the detail enhanced cell real-time posture image into the image semantic feature extractor based on the VIT model to obtain the cell posture image semantic feature vector. It should be understood that the VIT (Vision Transformer) model is an advanced deep learning model, which can effectively capture the main semantic information in the image from a global perspective by introducing the Transformer structure into computer vision tasks. Specifically, the VIT model divides the detail enhanced cell real-time posture image into multiple small, non-overlapping regions (tokens), and then converts these tokens into high-dimensional vectors through embedded coding, and then performs sequence processing through a multi-layer self-attention mechanism, which can effectively capture the global context information of the image, fully understand the global semantic features such as the overall shape, orientation and size of the cell, and provide key information support for the recognition of the cell posture.

[0039] Specifically, the cell attitude edge detail feature extraction unit 123 is used to extract the edge detail features of the detail enhanced cell attitude image to obtain the HOG feature vector of the cell attitude image. In a specific example of the present application, the processing method for extracting the edge detail features of the detail enhanced cell attitude image is to calculate the gradient direction histogram of the detail enhanced cell real-time attitude image to obtain the HOG feature vector of the cell attitude image. Those of ordinary skill in the art should know that the HOG (Histogram of Oriented Gradients) feature is a classic edge and detail descriptor that can effectively capture edge information and local structural features in an image. Specifically, the calculation process of the HOG feature generally includes calculating the gradient of the image, normalizing the gradient in a direction, constructing a direction histogram in each cell, and enhancing rotational invariance and suppressing background noise through cell and block normalization. In the technical solution of the present application, by calculating the gradient direction histogram of the detail-enhanced real-time posture image of the battery cell, the sensitivity of recognizing small posture changes and local deformation features of the battery cell can be enhanced, which helps to improve the accuracy of posture recognition. In addition, the edge contour information and local structural features of the battery cell are expressed in numerical form, which facilitates subsequent multi-level feature fusion analysis.

[0040] In the above-mentioned new energy battery cell transport system 100 with posture adjustment function, the feature interactive fusion module 130 is used to interactively fuse the semantic feature vector of the cell posture image and the HOG feature vector of the cell posture image to obtain a multi-scale characterization feature vector of the cell posture. It should be understood that by fusing the semantic feature vector of the cell posture image and the HOG feature vector of the cell posture image, the information complementarity and interactive enhancement of the global semantic features and edge detail features of the cell can be achieved to obtain a more comprehensive and accurate representation of the cell posture features, and further improve the accuracy of cell posture recognition. In a specific example of the present application, the interactive fusion processing method of the semantic feature vector of the cell posture image and the HOG feature vector of the cell posture image is to input the semantic feature vector of the cell posture image and the HOG feature vector of the cell posture image into a multi-scale posture feature combiner based on a bilinear pooling layer to obtain the multi-scale characterization feature vector of the cell posture. It should be known to those skilled in the art that the bilinear pooling layer is a feature fusion technology in deep learning, which can effectively understand the association between different modalities and enhance the interaction between features. Here, the bilinear pooling layer calculates the high-order interaction between the semantic feature vector of the battery cell posture image and the HOG feature vector of the battery cell posture image, fully explores and utilizes the intrinsic connection between the two features, and can deeply integrate the global semantic macro understanding of the battery cell posture with the fine description of the edge details to form a multi-scale feature representation, thereby improving the robustness to changes in battery cell posture and recognition accuracy.

[0041] In the above-mentioned new energy battery cell transport system 100 with posture adjustment function, the adjustment instruction generation module 140 is used to generate a cell posture adjustment instruction based on the multi-scale characterization feature vector of the cell posture. Figure 4 FIG. 1 is a block diagram of an adjustment instruction generation module in a new energy battery cell transport system with a posture adjustment function according to an embodiment of the present application. Figure 4 As shown, the adjustment instruction generation module 140 includes: a feature optimization unit 141, which is used to perform feature optimization on the multi-scale characterization feature vector of the battery cell posture to obtain an optimized multi-scale characterization feature vector of the battery cell posture; a decoding unit 142, which is used to pass the optimized multi-scale characterization feature vector of the battery cell posture through a decoder-based posture adjustment instruction generator to obtain the battery cell posture adjustment instruction, and the battery cell posture adjustment instruction includes a rotation angle, flipping and translation direction.

[0042] Specifically, the feature optimization unit 141 is used to perform feature optimization on the multi-scale characterization feature vector of the battery cell posture to obtain an optimized multi-scale characterization feature vector of the battery cell posture. In particular, in the technical solution of the present application, the use of the neural network model runs through the entire battery cell posture recognition and adjustment process. First, the image semantic feature extractor uses the visual transformer (VI T) model, which is a deep learning model that extracts deep semantic features of the image by dividing the image into small blocks (tokens) and applying a self-attention mechanism. This processing method enables the model to capture the details and contextual information in the image, providing a rich semantic basis for subsequent feature union. Next, the multi-scale posture feature combiner adopts a technology based on a bilinear pooling layer, which is a pooling technology that can process and fuse features from different modalities. Finally, the posture adjustment instruction generator uses a decoder, which is the part of the neural network responsible for converting the features extracted by the encoder into specific outputs. However, considering that the decoder-based posture adjustment instruction generator performs feature mining and decoding regression on the multi-scale characterization feature vector of the battery cell posture based on the decoding parameter matrix, if the intrinsic implicit correlation between the multi-scale characterization feature vector of the battery cell posture and the decoding parameter matrix can be further considered, the prior information of the decoder can be further utilized to modulate the characteristic manifold of the multi-scale characterization feature vector of the battery cell posture to reduce the possibility of special instance disturbances of its eigenvalues.

[0043] Based on this, in a specific example of the present application, the cell attitude multi-scale characterization feature vector is subjected to a feature compatibility evaluation based on implicit correlation to obtain an optimized cell attitude multi-scale characterization feature vector. Specifically, the feature optimization unit 141 is used to: calculate the mean and variance of the cell attitude multi-scale characterization feature vector; calculate the difference between the feature value of a predetermined position of the cell attitude multi-scale characterization feature vector and the mean of the cell attitude multi-scale characterization feature vector to obtain a first intermediate calculation value; multiply the decoding parameter matrix by the cell attitude multi-scale characterization feature vector and then subtract the cell attitude multi-scale characterization feature vector by position to obtain a first intermediate vector, and calculate the Froben i of the first intermediate vector. us norm to obtain a first norm value; add the first norm value to the variance of the multi-scale characterization feature vector of the battery cell posture and then take the square root to obtain a second intermediate calculation value; divide the first intermediate calculation value by the second intermediate calculation value to obtain a third intermediate calculation value; calculate the square of each position in the multi-scale characterization feature vector of the battery cell posture and then add them up to obtain a first sum value; take the square root of the first sum value and divide it by the scale of the decoding parameter matrix to calculate the natural exponential function value to obtain a fourth intermediate calculation value, wherein the scale of the decoding parameter matrix is ​​equal to the product of the width value of the decoding parameter matrix and its height value; multiply the fourth intermediate calculation value by the weighted hyperparameter and then add the third intermediate calculation value to obtain the eigenvalue of the predetermined position of the optimized battery cell posture multi-scale characterization feature vector.

[0044] That is, the feature compatibility of the multi-scale characterization feature vector of the battery cell posture is evaluated based on implicit correlation using the following feature optimization formula to obtain an optimized multi-scale characterization feature vector of the battery cell posture, wherein the feature optimization formula is:

[0045]

[0046] Where V is the multi-scale characterization feature vector of the cell posture, M is the decoding parameter matrix, β represents the weighted hyperparameter, and v i is the eigenvalue of the i-th eigenvector in the multi-scale characterization feature vector of the battery cell posture, N is the scale of the decoding parameter matrix, the scale of the decoding parameter matrix is ​​equal to the product of the width value and the height value of the decoding parameter matrix, L is the length of the multi-scale characterization feature vector of the battery cell posture, ‖·‖ F represents the Frobenius norm of the multi-scale characterization feature vector of the battery cell posture, μ represents the mean of the feature set composed of all eigenvalues ​​of the multi-scale characterization feature vector of the battery cell posture, σ 2 represents the variance of the feature set composed of all eigenvalues ​​of the multi-scale characterization feature vector of the battery cell posture, v i ′is the eigenvalue of the i-th eigenvalue in the multi-scale characterization eigenvector of the optimized cell posture.

[0047] That is, in the above embodiment, the feature compatibility evaluation based on implicit correlation is performed on the multi-scale characterization feature vector of the battery cell posture, which first maps the multi-scale characterization feature vector of the battery cell posture to the high-dimensional feature space where the decoding parameter matrix is ​​located, and uses the compatibility perception factor between the mapping result and the original input as the intrinsic implicit correlation descriptor between the network output feature and the decoding parameter matrix. Then, the statistical characteristics of the feature group of the multi-scale characterization feature vector of the battery cell posture are used as the inherent modal information of the multi-scale characterization feature vector of the battery cell posture, and the intrinsic implicit correlation descriptor is used as the instance perturbation differential adjustment variable to perform data dynamic standardization on the attribute values ​​of each position of the multi-scale characterization feature vector of the battery cell posture. At the same time, the class-standardized modulus length data of the multi-scale characterization feature vector of the battery cell posture is used as a bias term to constrain the acceptable range of the group distribution of the multi-scale characterization feature vector of the battery cell posture. In this way, it is possible to suppress the attribute influence of the interference distribution of abnormal samples as much as possible, and improve the expression accuracy of the multi-scale characterization feature vector of the battery cell posture.

[0048] Specifically, the decoding unit 142 is used to pass the optimized cell posture multi-scale characterization feature vector through a decoder-based posture adjustment instruction generator to obtain the cell posture adjustment instruction, and the cell posture adjustment instruction includes a rotation angle, flipping and translation direction. That is, the multi-layer neural network structure of the decoder is used to realize feature decoding and conversion of the optimized cell posture multi-scale characterization feature vector to generate accurate cell posture adjustment instructions. First, by inputting the optimized cell posture multi-scale characterization feature vector into the hidden layer of the decoder, the feature is activated by a nonlinear activation function to mine the deep structural information in the feature vector. Then, each layer of the decoder gradually restores the spatial resolution of the feature through deconvolution or upsampling operations, and adaptively reshapes the optimized cell posture multi-scale characterization feature vector to meet the output expectations of the decoder, thereby converting it into a specific cell posture adjustment instruction, including a rotation angle, flipping parameters and translation direction, to achieve automatic control of the cell posture.

[0049] In summary, according to the embodiment of the present application, a new energy battery cell transport system with a posture adjustment function is explained, which uses computer vision technology based on deep learning to perform image analysis on real-time posture images of new energy battery cells during transportation, captures the global posture semantic information and edge contour detail information of the battery cell, and comprehensively describes the current posture of the battery cell based on the interactive fusion characteristics of the two, thereby generating posture adjustment instructions for the battery cell. In this way, intelligent recognition and precise adjustment of the battery cell posture can be achieved, effectively preventing damage to the battery cell during transportation, improving transportation efficiency, and ensuring the safety of battery production, storage and use.

[0050] The basic principle of the present invention is described above in conjunction with specific embodiments. However, it should be pointed out that the advantages, strengths, effects, etc. mentioned in the present invention are only examples and not limitations, and it cannot be considered that these advantages, strengths, effects, etc. must be possessed by each embodiment of the present invention. In addition, the specific details of the above embodiments are only for the purpose of illustration and facilitation of understanding, rather than limitation, and the above details do not limit the present invention to being implemented by adopting the above specific details.

[0051] In the above embodiments, the description of each embodiment has its own emphasis. For the parts that are not described or recorded in detail in a certain embodiment, please refer to the relevant description of other embodiments. In the several embodiments provided by the present invention, it should be understood that the disclosed system can be implemented in other ways. For example, the system embodiment described above is only schematic. For example, the module division is only a logical function division, and there may be other division methods in actual implementation. The modules described as separate components may or may not be physically separated, and the components displayed as modules may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment.

[0052] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in terms of function in the above description. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0053] In addition, each functional module in each embodiment of the present invention may be integrated into one processing unit, each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of hardware plus software functional modules.

[0054] 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 to a figure in a claim should not be considered as limiting the claim to which it relates.

[0055] In addition, it is obvious that the word "comprising" does not exclude other units or steps, and the singular does not exclude the plural. The multiple units stated in the system claims can also be implemented by one unit through software or hardware.

[0056] Finally, it should be noted that the above description has been given for the purpose of illustration and description. In addition, the above embodiments are only used to illustrate the technical solution of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that the technical solution of the present invention can be modified or replaced by equivalents without departing from the spirit and scope of the technical solution of the present invention.

Claims

1. A new energy battery cell transport system with posture adjustment function, characterized in that: include: The battery cell posture real-time monitoring module is used to obtain the real-time posture image of the new energy battery cell during the transportation process; A multi-level image feature extraction module, used for performing multi-level feature extraction on the real-time posture image of the battery cell to obtain a semantic feature vector of the battery cell posture image and a HOG feature vector of the battery cell posture image; A feature interactive fusion module, used for interactively fusing the semantic feature vector of the battery cell posture image and the HOG feature vector of the battery cell posture image to obtain a multi-scale representation feature vector of the battery cell posture; The adjustment instruction generation module is used to generate a cell posture adjustment instruction based on the multi-scale characterization feature vector of the cell posture.

2. The new energy battery cell transport system with posture adjustment function according to claim 1 is characterized in that: The multi-level image feature extraction module comprises: An image preprocessing unit, used for performing image preprocessing on the real-time posture image of the battery cell to obtain a detail-enhanced posture image of the battery cell; A cell posture semantic feature extraction unit, used for extracting the global semantic features of the detail-enhanced cell posture image to obtain a semantic feature vector of the cell posture image; The cell posture edge detail feature extraction unit is used to extract the edge detail features of the detail-enhanced cell posture image to obtain the HOG feature vector of the cell posture image.

3. The new energy battery cell transport system with posture adjustment function according to claim 2 is characterized in that: The image preprocessing unit is used for: Gamma correction is performed on the real-time battery cell posture image to obtain the detail-enhanced battery cell posture image.

4. The new energy battery cell transport system with posture adjustment function according to claim 3 is characterized in that: The cell posture semantic feature extraction unit is used to: The detail-enhanced battery cell real-time posture image is input into an image semantic feature extractor based on a VIT model to obtain a semantic feature vector of the battery cell posture image.

5. The new energy battery cell transport system with posture adjustment function according to claim 4 is characterized in that: The cell posture edge detail feature extraction unit is used to: The gradient direction histogram of the detail-enhanced battery cell real-time posture image is calculated to obtain the HOG feature vector of the battery cell posture image.

6. The new energy battery cell transport system with posture adjustment function according to claim 5 is characterized in that: The feature interaction fusion module is used to: The semantic feature vector of the battery cell posture image and the HOG feature vector of the battery cell posture image are input into a multi-scale posture feature combiner based on a bilinear pooling layer to obtain a multi-scale representation feature vector of the battery cell posture.

7. The new energy battery cell transport system with posture adjustment function according to claim 6 is characterized in that: The adjustment instruction generating module comprises: A feature optimization unit, used for performing feature optimization on the multi-scale characterization feature vector of the battery cell posture to obtain an optimized multi-scale characterization feature vector of the battery cell posture; A decoding unit is used to pass the optimized multi-scale characterization feature vector of the battery cell posture through a decoder-based posture adjustment instruction generator to obtain the battery cell posture adjustment instruction, wherein the battery cell posture adjustment instruction includes a rotation angle, a flipping direction, and a translation direction.

8. The new energy battery cell transport system with posture adjustment function according to claim 7 is characterized in that: The feature optimization unit is used to: The cell posture multi-scale characterization feature vector is evaluated for feature compatibility based on implicit correlation to obtain the optimized cell posture multi-scale characterization feature vector.

9. The new energy battery cell transport system with posture adjustment function according to claim 8, characterized in that: The feature optimization unit is used to: Calculating the mean and variance of the multi-scale characterization feature vector of the battery cell posture; Calculating a difference between a feature value at a predetermined position of the multi-scale characterization feature vector of the battery cell posture and a mean value of the multi-scale characterization feature vector of the battery cell posture to obtain a first intermediate calculation value; Multiplying the decoding parameter matrix by the cell posture multi-scale characterization feature vector and then subtracting the multi-scale characterization feature vector from the cell posture multi-scale characterization feature vector by position to obtain a first intermediate vector, and calculating the Frobenius norm of the first intermediate vector to obtain a first norm value; Add the first norm value to the variance of the multi-scale characterization feature vector of the battery cell posture and then take the square root to obtain a second intermediate calculation value; dividing the first intermediate calculated value by the second intermediate calculated value to obtain a third intermediate calculated value; Calculating the square of each position in the multi-scale characterization feature vector of the battery cell posture and then accumulating and summing them to obtain a first total value; Calculating a natural exponential function value after taking a square root of the first sum value and dividing it by the scale of the decoding parameter matrix to obtain a fourth intermediate calculation value, wherein the scale of the decoding parameter matrix is ​​equal to the product of a width value of the decoding parameter matrix and a height value thereof; The fourth intermediate calculated value is multiplied by the weighted hyperparameter and then added to the third intermediate calculated value to obtain the eigenvalue of the predetermined position of the multi-scale characterization eigenvector of the optimized battery cell posture.

Images