Battery-delivered product identification method, device, program product, and storage medium

Abstract

The application discloses a battery conveying product identification method, device, program product and storage medium, and relates to the technical field of image recognition. The method comprises the following steps: acquiring an overhead view of a battery on a conveying line, transforming the overhead view into a preset coordinate system to obtain a corrected image containing pixel coordinates; intercepting a to-be-identified region corresponding to a nameplate coordinate range from the corrected image; acquiring standard templates corresponding to each candidate model; comparing each pixel coordinate in the nameplate coordinate range with pixel values of all the standard templates at the pixel coordinates; acquiring a brightness value of the to-be-identified region at the pixel coordinates, eliminating pixel coordinates with brightness values beyond a preset brightness interval from a judgment coordinate set to obtain an effective judgment coordinate set; and calculating matching degrees corresponding to the candidate models according to the pixel values of the to-be-identified region at the pixel coordinates contained in the effective judgment coordinate set and the pixel values of the standard templates corresponding to the candidate models, and taking a candidate model with the highest matching degree as an identification result of the battery.

Description

Technical Field

[0001] This application relates to the technical field of image recognition, specifically to a product identification method, device, program product, and storage medium for battery-powered products. Background Technology

[0002] As a core component of equipment such as electric vehicles, energy storage power stations, and emergency power supplies, batteries typically need to be transported between various processes via conveyor lines during the manufacturing process. To facilitate traceability management and differentiated operations in downstream processes, it is usually necessary to identify the product model of each battery as it passes through the conveyor line.

[0003] In related technologies, an image acquisition device is usually installed above the conveyor line to obtain images of the batteries flowing through it. Based on the visual characteristics of the battery nameplate area in the acquired images, the images are compared with the pre-established sample characteristics of each product model to determine the product model corresponding to the flowing battery.

[0004] However, in the actual operating environment of the conveyor line, factors such as lighting sources, ambient reflections, and oil stains, water stains, and scratches on the surface of the battery will affect the acquired images, causing some pixels in the nameplate area to appear visually different from the printed content of the nameplate itself, thus affecting the accuracy of the model comparison results based on visual features. Summary of the Invention

[0005] This application provides a product identification method, device, program product, and storage medium for battery-powered products, which can improve the accuracy of product identification in battery-powered products.

[0006] The first aspect of this application provides a product identification method for battery transportation, specifically including: acquiring a top view of a battery on a transportation line, transforming the top view into a preset coordinate system to obtain a corrected image containing pixel coordinates; extracting the area to be identified from the corrected image corresponding to the nameplate coordinate range according to a preset nameplate coordinate range of the battery in the coordinate system; acquiring images of multiple candidate battery models located within the nameplate coordinate range in the coordinate system, as standard templates corresponding to each candidate model; comparing each pixel coordinate in the nameplate coordinate range with the pixel values ​​at the pixel coordinates of all standard templates, and adding pixel coordinates with pixel value differences to a discrimination coordinate set; for each pixel coordinate in the discrimination coordinate set, acquiring the brightness value of the area to be identified at the pixel coordinates, and removing pixel coordinates with brightness values ​​exceeding a preset brightness range from the discrimination coordinate set to obtain an effective discrimination coordinate set; for each candidate model, calculating the matching degree corresponding to the candidate model based on the pixel values ​​of the area to be identified at the pixel coordinates included in the effective discrimination coordinate set and the pixel values ​​of the standard templates corresponding to the candidate model, and taking the candidate model with the highest matching degree as the identification result of the battery.

[0007] Further, based on the preset nameplate coordinate range of the battery in the coordinate system, the region to be identified corresponding to the nameplate coordinate range is extracted from the calibration image, including: extracting the boundary of the top surface of the battery in the calibration image to obtain the measured contour; obtaining the preset standard contour of the top surface of the battery in the coordinate system; determining the pose offset of the measured contour relative to the standard contour based on the correspondence between the measured contour and the standard contour, the pose offset including translation and rotation components; transforming the pixel coordinates according to the pose offset for each pixel coordinate in the nameplate coordinate range to obtain the actual sampling coordinates corresponding to the pixel coordinates in the calibration image; and combining the pixel values ​​at each actual sampling coordinate in the calibration image as the pixel values ​​at the corresponding pixel coordinates in the nameplate coordinate range to obtain the region to be identified.

[0008] Furthermore, each pixel coordinate in the nameplate coordinate range is compared with the pixel value of all standard templates at the pixel coordinates. Pixel coordinates with pixel value differences are added to the discrimination coordinate set. This includes: pairing all candidate models to obtain multiple candidate model pairs; for each candidate model pair, comparing the pixel values ​​of the two standard templates corresponding to the two candidate models in the candidate model pair pixel by pixel, and adding the pixel coordinates with pixel value differences to the difference coordinate subset corresponding to the candidate model pair; merging the difference coordinate subsets corresponding to all candidate model pairs to obtain the discrimination coordinate set.

[0009] Furthermore, the pixel values ​​of the two standard templates corresponding to the two candidate models included in the candidate model pair are compared pixel by pixel. The pixel coordinates with differences in pixel values ​​are added to the difference coordinate subset corresponding to the candidate model pair. This includes: for each pixel coordinate in the nameplate coordinate range, obtaining the pixel values ​​of the two standard templates at the pixel coordinates, denoted as the first pixel value and the second pixel value; calculating the absolute value of the difference between the first pixel value and the second pixel value; obtaining the pixel values ​​of all pixel positions in the first preset neighborhood of the pixel coordinates of the two standard templates, and using the difference between the maximum and minimum values ​​of the obtained pixel values ​​as the local fluctuation amount corresponding to the pixel coordinates; if the absolute value of the difference is greater than the local fluctuation amount, the pixel coordinates are added to the difference coordinate subset corresponding to the candidate model pair.

[0010] Furthermore, based on the pixel values ​​of the region to be identified at the pixel coordinates contained in the effective discrimination coordinate set and the pixel values ​​of the standard templates corresponding to the candidate models, the matching degree corresponding to the candidate models is calculated, including: for each pixel coordinate in the effective discrimination coordinate set, comparing the pixel value of the region to be identified at the pixel coordinate with the pixel values ​​of the standard templates corresponding to all candidate models at the pixel coordinate, and taking the candidate model corresponding to the standard template whose pixel value is closest to the pixel value of the region to be identified at the pixel coordinate as the winning model corresponding to the pixel coordinate; for each candidate model, counting the number of pixel coordinates in the effective discrimination coordinate set that use the candidate model as the winning model, and taking the matching degree corresponding to the candidate model.

[0011] Further, the pixel value of the region to be identified at the pixel coordinates is compared with the pixel value of the standard template corresponding to all candidate models at the pixel coordinates. The candidate model corresponding to the standard template whose pixel value is closest to the pixel value of the region to be identified at the pixel coordinates is selected as the winning model corresponding to the pixel coordinates. This includes: obtaining all pixel coordinates within the second preset neighborhood of the pixel coordinates as the neighborhood coordinate set corresponding to the pixel coordinates; for each neighborhood pixel coordinate in the neighborhood coordinate set, comparing the pixel value of the region to be identified at the neighborhood pixel coordinates with the pixel value of the standard template corresponding to all candidate models at the neighborhood pixel coordinates, and selecting the candidate model corresponding to the standard template whose pixel value is closest to the pixel value of the region to be identified at the neighborhood pixel coordinates as the locally winning model corresponding to the neighborhood pixel coordinates; for each candidate model, counting the number of neighborhood pixel coordinates in the neighborhood coordinate set that use the candidate model as the locally winning model, as the local support number of the candidate model at the pixel coordinates; and selecting the candidate model with the largest local support number among all candidate models as the winning model corresponding to the pixel coordinates.

[0012] Furthermore, after identifying the candidate model with the highest matching degree as the battery identification result, the process also includes: obtaining a preset working parameter library, which stores the working parameters corresponding to each candidate model; querying the working parameters of the candidate model corresponding to the identification result from the working parameter library based on the identification result, and using them as the target working parameters of the battery; and associating and binding the identification result with the target working parameters to output the battery identification profile.

[0013] In a second aspect, this application provides a battery-powered product identification device, which includes: one or more processors and a memory; the memory is coupled to the one or more processors and is used to store computer program code, which includes computer instructions, and the one or more processors call the computer instructions to cause the battery-powered product identification device to perform the method described in the first aspect and any possible implementation thereof.

[0014] Thirdly, this application provides a computer program product containing instructions that, when the computer program product is run on a product identification device for battery delivery, causes the product identification device for battery delivery to perform the method described in the first aspect and any possible implementation thereof.

[0015] Fourthly, this application provides a computer-readable storage medium including instructions that, when executed on a product identification device for battery delivery, cause the product identification device for battery delivery to perform the method described in the first aspect and any possible implementation thereof.

[0016] Compared with the prior art, the product identification method, equipment, program product and storage medium for battery delivery provided in this application have at least the following beneficial effects: (1) By transforming the top view of the battery on the conveyor line to a preset coordinate system, a corrected image containing pixel coordinates is obtained. The area to be identified is then cropped according to the preset nameplate coordinate range in the coordinate system. This allows the position and orientation differences of different batteries flowing through the conveyor line in the acquired image to be uniformly attributed to the same spatial reference. This enables the area to be identified to correspond pixel-by-pixel with the standard template corresponding to each candidate model under the same coordinate semantics, thus avoiding pixel-by-pixel misalignment caused by position drift from the source.

[0017] (2) By comparing each pixel coordinate in the nameplate coordinate range with the pixel value of all standard templates at the pixel coordinate, the pixel coordinates with pixel value differences are added to the discrimination coordinate set. This allows the subsequent model comparison to only act on the pixel coordinates that are truly distinguishable between candidate models, excluding the non-distinguishing pixel coordinates common to each candidate model from the discrimination range, reducing the dilution effect of common areas on model discrimination, and making the accumulation of matching degree focus on the model difference itself.

[0018] (3) By obtaining the brightness value of the region to be identified at each pixel coordinate in the discrimination coordinate set, and removing the pixel coordinates whose brightness values ​​exceed the preset brightness range from the discrimination coordinate set, an effective discrimination coordinate set is obtained. This allows the pixel coordinates that deviate from the normal imaging range due to factors such as fluctuations in ambient lighting, environmental reflection, and oil stains, water stains, and scratches on the surface of the battery to be identified to not participate in the subsequent matching degree calculation. This identifies and isolates unreliable pixel positions from the region to be identified, reducing the impact of the factors on the model discrimination result.

[0019] (4) By calculating the matching degree of each candidate model on the pixel coordinates contained in the effective discrimination coordinate set, and taking the candidate model with the highest matching degree as the identification result of the battery, the final discrimination can be completed on the pixel subset after the dual filtering of "having model differentiation ability" and "reliable imaging". The identification result benefits from the focus of the discrimination coordinate set on model differences and the elimination of abnormal pixels by the effective discrimination coordinate set, thereby improving the accuracy of product identification of the battery during the transfer process on the conveyor line. Attached Figure Description

[0020] Figure 1 This is a schematic flowchart of a product identification method for battery delivery provided in an embodiment of this application; Figure 2 This is an exemplary hardware structure diagram of a product identification device for battery delivery provided in an embodiment of this application. Detailed Implementation

[0021] To enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments.

[0022] In the description of the embodiments of this application, the words "for example" or "for instance" are used to indicate examples, illustrations, or explanations. Any embodiment or design that is described as "for example" or "for instance" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or design options. Rather, the use of the words "for example" or "for instance" is intended to present the relevant concepts in a specific manner.

[0023] In the description of the embodiments of this application, the term "multiple" means two or more. For example, multiple systems means two or more systems, and multiple screen terminals means two or more screen terminals. Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the indicated technical features. Thus, a feature defined with "first" or "second" may explicitly or implicitly include one or more of that feature. The terms "comprising," "including," "having," and variations thereof all mean "including but not limited to," unless otherwise specifically emphasized.

[0024] In view of the problems existing in product identification on battery conveying lines in related technologies, based on the principles and application effects of manual identification methods and automatic identification methods based on barcodes or QR codes mentioned in the background technology, the shortcomings of existing technologies are summarized as follows: (1) Inconsistent accuracy. Manual identification is prone to misjudgment due to fatigue under long-term high-intensity operation. Barcode or QR code-based identification methods are limited by the printing quality of the label and the surface contamination. During the production and transportation process, the surface of the battery is prone to the adhesion of residual electrolyte, dust and oil, which will cause the label area to be dirty, worn or reflective, thus leading to identification failure or misidentification.

[0025] (2) Poor adaptability. Due to factors such as conveyor belt vibration and loose clamps, the position and orientation of the battery relative to the acquisition equipment will be offset and rotated to a certain extent on the conveyor line. This results in the image of the same model of battery showing different imaging positions at different times. The traditional full-image template matching method is highly sensitive to position drift and it is difficult to maintain stable recognition accuracy.

[0026] (3) Weak differentiation ability. Batteries of different models from the same manufacturer often only have local differences in the nameplate area. Their shapes and colors are highly similar. Traditional pixel-by-pixel full-image comparison strategies are difficult to distinguish between common and different areas, which dilutes the model differentiation signal by a large number of common pixels and affects the recognition accuracy.

[0027] In view of this, embodiments of the present invention provide a product identification method, device, program product, and storage medium for battery conveying. The system design can be carried out from top view acquisition and coordinate correction, nameplate area extraction, candidate model difference analysis, abnormal pixel removal to matching degree calculation and identification result output, so as to efficiently achieve accurate identification of different models of batteries in the battery conveying line operation scenario.

[0028] Figure 1 schematically illustrates a flowchart of a product identification method for battery delivery according to an embodiment of the present invention.

[0029] As shown in Figure 1, the product identification method for battery delivery in this embodiment includes operations S101 to S106.

[0030] In operation S101, a top view of the battery on the conveyor line is obtained, and the top view is transformed into a preset coordinate system to obtain a corrected image containing pixel coordinates.

[0031] Specifically, an industrial camera is installed above the battery conveyor line, with its optical axis perpendicular to the conveyor surface, continuously capturing top-down views of the batteries flowing through the imaging field of view. Considering the limited installation height of cameras in industrial settings, the radial distortion of lenses, and the potential for minute angular deviations between the optical axis and the conveyor surface, directly processing the original top-down view would introduce unpredictable geometric errors. Therefore, the camera is pre-calibrated to obtain its intrinsic parameter matrix and lens distortion coefficients. Simultaneously, a checkerboard calibration plate of known size is placed on the conveyor surface to calibrate the camera's extrinsic parameters relative to the conveyor surface, thereby obtaining the homography transformation matrix mapping from the original image coordinate system to a preset coordinate system based on the conveyor surface. .

[0032] Top view acquired for each frame Through homography transformation matrix The original image is then subjected to distortion correction and geometric correction using lens distortion coefficients to obtain an image containing pixel coordinates. Corrected image : in Corrected pixel coordinates The source coordinates are the same as those in the original image. Each pixel coordinate in the corrected image corresponds to a fixed physical position on the transport surface, ensuring that batteries passing through the same physical position at different times have a consistent pixel coordinate reference in the corrected image, providing a unified spatial reference for subsequent nameplate area positioning and pixel-by-pixel comparison.

[0033] In other embodiments, a top view of the battery can be obtained by synchronously triggering a linear scan camera in conjunction with a conveyor line encoder. Specifically, the rotary encoder on the conveyor line triggers the linear scan camera to scan and image at a fixed number of pulses corresponding to a fixed travel distance, resulting in a continuous, distortion-free scan image along the conveying direction. The pixel resolution corresponding to the conveying direction is determined only by the encoder pulse interval and is no longer affected by the matching relationship between the camera frame rate and the conveying speed, making it particularly suitable for production line scenarios with high or fluctuating conveying speeds. In other embodiments, multiple area scan cameras can be used to cover a wide conveying surface by stitching images together. After geometric correction is performed on the images acquired by each camera, image stitching is performed based on the pre-calibrated overlapping areas of the fields of view of adjacent cameras to obtain a corrected image covering the entire width of the conveying surface.

[0034] In operation S102, based on the preset nameplate coordinate range of the battery in the coordinate system, the region to be identified corresponding to the nameplate coordinate range is extracted from the calibration image.

[0035] In some embodiments, the above operation S102 may further include: extracting the boundary of the top surface of the battery in the calibration image to obtain the measured contour; obtaining a preset standard contour of the top surface of the battery in the coordinate system; determining the pose offset of the measured contour relative to the standard contour according to the correspondence between the measured contour and the standard contour; transforming the pixel coordinates according to the pose offset for each pixel coordinate in the nameplate coordinate range to obtain the actual sampling coordinates corresponding to the pixel coordinates in the calibration image; and combining the pixel values ​​of the calibration image at each actual sampling coordinate as the pixel values ​​at the corresponding pixel coordinates in the nameplate coordinate range to obtain the region to be identified.

[0036] Specifically, during the transportation process, the position and orientation of the battery often deviate from the preset position due to conveyor belt vibration and mechanical clearance of the clamps. Directly cropping a region from the calibration image using the preset nameplate coordinate range will cause a misalignment between the cropped area and the actual location of the battery nameplate. Therefore, boundary extraction is first performed in the calibration image: Gaussian smoothing is applied to the calibration image to suppress noise, then edge detection operators are used to extract edge responses, and morphological closing operations are used to fill broken edges. Finally, the maximum connected edge is extracted as the measured contour of the battery's top surface. .

[0037] Standard contour in the preset coordinate system The boundary of the battery's top surface in its ideal position can be obtained from the battery's mechanical dimension drawings or by collecting data from a sample battery during the calibration phase. The measured profile will then be... With standard profile Perform point set registration to obtain pose offset. ,in , For translation components, The rotation component is used. The pose offset can be obtained by minimizing the sum of squared Euclidean distances between corresponding points on the two contours: in and These are the corresponding first and second profiles on the standard profile and the measured profile, respectively. One sampling point, The total number of sampling points. For is a two-dimensional rotation matrix for the rotation angle.

[0038] After obtaining the pose offset, for each pixel coordinate within the nameplate coordinate range... The actual sampled coordinates corresponding to the pixel coordinates in the corrected image are obtained by transforming the image based on the pose offset. : The pixel values ​​of the corrected image at each actual sampling coordinate are sequentially filled into the corresponding pixel coordinate positions within the nameplate coordinate range, thereby obtaining the region to be identified. When the actual sampling coordinates are non-integer, bilinear interpolation or nearest neighbor interpolation can be used to obtain the corresponding pixel value. Bilinear interpolation is preferred to retain higher grayscale accuracy. Through the above pose compensation operation, the region to be identified... Each pixel coordinate in the model can be strictly matched with the physical position of the standard template corresponding to each candidate model at the same pixel coordinate, thus eliminating the influence of pose deviation for subsequent pixel-by-pixel comparison.

[0039] In other embodiments, significant local features on the top surface of the battery, such as terminals, injection holes, and safety valves, can be used as pose registration references. Specifically, the center coordinates of these local features are located in the corrected image using circle detection or spot detection operators, and rigid body registration is performed between these coordinates and the center coordinates of the corresponding features in the standard pose to obtain the pose offset. Compared to registration methods based on the overall edge contour, this method is more robust to side occlusion of the battery and deformation caused by impacts to the casing. In other embodiments, positioning pin holes on the fixture can be introduced as external reference references. The pose of the battery on the transport line is determined by the imaging center of the positioning pin holes, avoiding feature extraction of the battery body and further reducing the dependence on the uniformity of the battery's appearance.

[0040] In operation S103, images of multiple candidate battery models located within the nameplate coordinate range in the coordinate system are acquired and used as standard templates for each candidate model.

[0041] Specifically, during the system deployment phase, for each candidate model ( , (Total number of candidate models), select a sample battery with good imaging quality for that model, acquire the calibration image and complete pose compensation in the same manner as operations S101 and S102, and crop the image within the nameplate coordinate range as the standard template corresponding to that candidate model. To improve the robustness of the standard template to imaging noise and individual differences, multiple samples can be collected for the same candidate model. The final standard template is obtained by averaging the pixel-level values ​​of the nameplate area images of multiple samples. The standard templates corresponding to all candidate models share the same pixel coordinate reference, forming a template library that corresponds to the region to be identified pixel by pixel.

[0042] In other embodiments, in addition to the pixel-level mean, the standard deviation of the nameplate area image of each sample can be calculated at the pixel level to obtain a pixel-level variance map corresponding to the candidate model. The pixel-level variance map reflects the inherent variability of different samples under the same candidate model at each pixel position, and can serve as an auxiliary reference for the lower bound of the difference judgment threshold in operation S104, so that the model difference judgment is not misjudged as a model difference by inherent variability within the same model, and the true model distinguishing pixels are preserved. In other embodiments, the template library can be version-managed, and independent standard templates can be established for the same candidate model under different production batches and different nameplate printing processes. During the recognition process, the corresponding version of the template can be selected according to the batch information transmitted from the upstream process, further reducing the impact of batch differences on the recognition results.

[0043] In operation S104, the coordinates of each pixel in the nameplate coordinate range are compared with the pixel values ​​of all standard templates at the pixel coordinates, and the pixel coordinates with differences in pixel values ​​are added to the discrimination coordinate set.

[0044] In some embodiments, the above operation S104 may further include: pairing all candidate models in pairs to obtain multiple candidate model pairs; for each candidate model pair, comparing the pixel values ​​of the two standard templates corresponding to the two candidate models included in the candidate model pair pixel by pixel, adding the pixel coordinates with pixel value differences to the difference coordinate subset corresponding to the candidate model pair; merging the difference coordinate subsets corresponding to all candidate model pairs to obtain a discrimination coordinate set.

[0045] Specifically, for all The candidate models are paired up in pairs, resulting in a total of candidate model pairs. For any candidate model pair The two corresponding standard templates are respectively and .

[0046] For each pixel coordinate within the nameplate coordinate range Obtain the pixel values ​​of the two standard templates at the specified pixel coordinates, and denote them as the first pixel value. With the second pixel value And calculate the absolute value of the difference between the two: Considering that even if two standard templates are located in the model difference region at the same pixel coordinates, slight fluctuations in local pixel values ​​may occur due to factors such as camera imaging noise and minor lighting variations, simply using a non-zero difference as the model difference criterion would introduce a large number of spurious difference pixels. Therefore, a local fluctuation value is introduced. An adaptive threshold is used for difference determination. Specifically, the pixel coordinates of the two standard templates are obtained. First preset neighborhood The pixel values ​​at all pixel locations within the area are used, and the difference between the maximum and minimum values ​​among these obtained pixel values ​​is taken as the local fluctuation amount corresponding to that pixel coordinate. The first preset neighborhood Preferred pixel coordinates A 3×3 or 5×5 rectangular neighborhood centered on the center.

[0047] If the absolute value of the difference Greater than local fluctuations If a pixel coordinate is found to have a significant model difference between the two standard templates, then that pixel coordinate is added to the candidate model pair. Corresponding difference coordinate subset This adaptive determination method allows small pixel value differences in flat texture areas to be identified as model differences, while larger differences are required in areas with drastic texture changes to be identified as model differences, thus avoiding local noise being misidentified as model differences.

[0048] The discriminant coordinate set is obtained by finding the union of the corresponding subsets of difference coordinates for all candidate models. : Discriminant coordinate set This method highlights the distinct pixel positions of all candidate models within the nameplate coordinate range, eliminating non-distinguishing areas common to all candidate models (such as areas with consistent battery casing color or nameplate background color), thus allowing the subsequent matching process to focus on the subset of pixels that truly distinguish the models.

[0049] In other embodiments, connected component analysis and morphological opening operations can be performed on the subset of difference coordinates to remove isolated small-area difference pixel clusters, retaining only difference pixel regions with a certain degree of spatial continuity. This processing can effectively remove scattered false difference pixels caused by random noise introduced during the standard template acquisition stage, making the model differences reflected by the discrimination coordinate set more focused on the real graphic and textual differences within the nameplate area. In other embodiments, a separability weight can be assigned to the subset of difference coordinates for each candidate model pair. The separability weight is the average of the absolute values ​​of the pixel value differences between the two standard templates corresponding to the candidate model pair within the subset of difference coordinates. This weight is used to weight the winning votes within the candidate model pair in operation S106, so that key pixels in candidate model pairs that can be clearly distinguished by only a few pixels receive higher judgment weights.

[0050] In operation S105, for each pixel coordinate in the discrimination coordinate set, the brightness value of the region to be identified at the pixel coordinate is obtained, and the pixel coordinates with brightness values ​​exceeding the preset brightness range are removed from the discrimination coordinate set to obtain an effective discrimination coordinate set.

[0051] Specifically, when a battery passes through the imaging area on the conveyor line, its nameplate surface is prone to being contaminated with residual electrolyte, oil, water droplets, or dust. It may also experience localized highlights due to light reflection, causing the brightness value of the corresponding pixel in the area to be identified to deviate significantly from the normal imaging range, thus interfering with model identification. Therefore, a preset brightness range is introduced. Confidence filtering is performed on the pixel coordinates in the discrimination coordinate set.

[0052] Lower limit of preset brightness range and upper limit This can be obtained from the statistical distribution of standard template pixel brightness of all candidate model samples during the system deployment phase. Specifically, the brightness values ​​of the standard template corresponding to each candidate model at the pixel coordinates contained in the discrimination coordinate set are summarized, and the 1st percentile and 99th percentile of this brightness value set are taken as the lower and upper limits, respectively. in Indicates the first The standard template corresponding to each candidate model is located at pixel coordinates. The brightness value at that location, This indicates quantile operations.

[0053] For the discriminant coordinate set Each pixel coordinate in Obtain the brightness value of the region to be identified at the specified pixel coordinates. .like or If a pixel's coordinates are found to be contaminated, obstructed, or abnormally reflective, it is removed from the discrimination coordinate set; otherwise, it is retained. The set of pixel coordinates obtained after removal is the valid discrimination coordinate set. : The above filtering operations effectively determine the coordinate set. The pixel coordinates contained therein not only have the ability to distinguish models, but also correspond to reliable imaging positions in the area to be identified, providing a high-quality subset of pixels for subsequent matching degree calculation.

[0054] In other embodiments, the single filtering criterion based on brightness range can be extended to a multi-dimensional filtering criterion. Specifically, in addition to the brightness channel, the chroma channel value and saturation channel value of the area to be identified at the pixel coordinates can be acquired simultaneously, and corresponding preset ranges can be established for each. The pixel coordinates are only retained when all three fall within the corresponding ranges. This multi-dimensional filtering method has higher recognition sensitivity for contaminants with specific color characteristics, such as oil stains and water droplets. In other embodiments, the local gradient magnitude of the pixel coordinates can be introduced as an auxiliary filtering criterion. Pixel coordinates with local gradient magnitudes significantly lower than the gradient magnitude of the pixel coordinates in the standard template are eliminated to prevent the nameplate area from being completely covered by foreign objects and exhibiting local blurring. Furthermore, when the effective discrimination coordinate set... When the number of pixel coordinates in the battery is lower than the preset minimum pixel count threshold, it can be determined that the current battery nameplate area is covered by too many foreign objects, triggering an alarm signal and marking the battery as pending re-inspection, thus avoiding outputting unreliable identification results when pixel information is severely insufficient.

[0055] In operation S106, for each candidate model, the matching degree of the candidate model is calculated based on the pixel value of the region to be identified at the pixel coordinates contained in the effective discrimination coordinate set and the pixel value of the standard template corresponding to the candidate model. The candidate model with the highest matching degree is taken as the identification result of the battery.

[0056] In some embodiments, the above operation S106 may further include: for each pixel coordinate in the effective discrimination coordinate set, comparing the pixel value of the region to be identified at the pixel coordinate with the pixel value of the standard template corresponding to all candidate models at the pixel coordinate, and taking the candidate model corresponding to the standard template whose pixel value is closest to the pixel value of the region to be identified at the pixel coordinate as the winning model corresponding to the pixel coordinate; for each candidate model, counting the number of pixel coordinates in the effective discrimination coordinate set that use the candidate model as the winning model, and taking the matching degree corresponding to the candidate model.

[0057] Specifically, for the effective discriminant coordinate set Each pixel coordinate in Calculate the pixel value of the region to be identified at the specified pixel coordinates. With each candidate model The corresponding standard template pixel value at this pixel coordinate The absolute value of the difference between them: The candidate model corresponding to the standard template with the smallest absolute value of the above difference is determined as the pixel coordinate. The corresponding winning model : In some embodiments, to further improve the robustness of single-pixel winning determination to random noise, a neighborhood voting mechanism can be introduced on top of the pixel-level winning determination. Specifically, for each pixel coordinate... Obtain its second preset neighborhood. All pixel coordinates within, used as pixel coordinates Corresponding neighborhood coordinate set Second preset neighborhood Preferred pixel coordinates A 3×3 or 5×5 rectangular neighborhood centered on the center.

[0058] For neighborhood coordinate set Each neighboring pixel coordinate The local winning model corresponding to the neighboring pixel coordinates is determined in the manner described above. For each candidate model Statistical neighborhood coordinate set China and Israel The number of neighboring pixel coordinates as the locally winning model, and as the candidate model. In pixel coordinates Local support number at the location : The candidate model with the highest local support count among all candidate models is used as the pixel coordinate. The winning model: Through the neighborhood voting mechanism, the winner of a single pixel position is no longer determined solely by the pixel value at that position, but rather by the model tendency of its local neighborhood, thus suppressing the disturbance of occasional noise and isolated abnormal pixels on the winner determination.

[0059] After obtaining an effective discriminant coordinate set After identifying the winning models corresponding to the pixel coordinates in the model, for each candidate model... ,statistics China and Israel The number of pixel coordinates used to determine the winning model, and the number of candidate models. Corresponding matching degree : The candidate model with the highest matching degree is used as the battery identification result. : The above matching degree calculation method changes the continuous measure of "cumulative gray-level similarity of the whole image" in traditional template matching to a discrete measure of "pixel-by-pixel voting", so that the contribution of each pixel position to the final recognition result is limited to one vote. This fundamentally avoids the problem that a few areas with severe textures dominate the overall matching result due to excessively large absolute differences, thus improving the stability of the recognition result.

[0060] In some embodiments, after identifying the candidate model with the highest matching degree as the battery identification result, the method further includes: obtaining a preset working parameter library, which stores the working parameters corresponding to each candidate model; querying the working parameters of the candidate model corresponding to the identification result from the working parameter library according to the identification result, and using them as the target working parameters of the battery; and associating and binding the identification result with the target working parameters to output the battery identification profile.

[0061] Specifically, the working parameter library stores the working parameters required for downstream processes of each candidate model, indexed by the candidate model identifier. These working parameters include, but are not limited to, rated charge / discharge voltage, rated current, formation process curve, test item list, and packaging specifications. The identification results obtained through operation S106... ,by To retrieve the target battery's working parameters, a query is performed in the working parameter database to obtain the set of working parameters corresponding to the candidate model.

[0062] Recognition results A key-value pair structure is formed with the target operating parameters, and the current battery position on the conveyor line, acquisition timestamp, and unique image number are added to create the battery's identification file. This identification file can be distributed to downstream charging / discharging equipment, testing equipment, or packaging equipment via an industrial fieldbus, enabling downstream processes to automatically load the corresponding operating parameters based on the identification file without manual intervention, achieving fully automated linkage between battery identification and downstream process execution.

[0063] In other implementations, a confidence assessment mechanism can be introduced to improve the tolerance for low-quality samples. Specifically, the confidence assessment mechanism corresponding to the candidate model with the highest matching degree can be used to assess the confidence level of the candidate model. Matching degree with the second highest matching candidate model The ratio, or will and The ratio is used as the recognition confidence level. When the recognition confidence level is lower than a preset confidence level threshold, the current recognition result is considered ambiguous, and the battery is transferred to the re-inspection channel for manual confirmation or re-recognition by a secondary imaging device to avoid outputting erroneous results when the confidence level is insufficient. In other embodiments, the area to be identified, the effective discrimination coordinate set, and the recognition result obtained in each recognition process can be archived and retained. The archived data can be back-analyzed periodically, and the standard template and discrimination coordinate set can be iteratively updated in combination with the actual model information fed back from downstream processes, so that the recognition system has the ability to continuously self-optimize as the production line runs. In addition, when the recognition confidence level of multiple batteries is consistently low, self-inspection processes such as camera lens cleaning and light source brightness calibration can be triggered to eliminate the influence of imaging system state drift on the recognition results from the source.

[0064] As can be seen from the above description, the embodiments of the present invention achieve at least the following technical effects: (1) By transforming the top view of the battery on the conveyor line to a preset coordinate system, a corrected image containing pixel coordinates is obtained. The area to be identified is then cropped according to the preset nameplate coordinate range in the coordinate system. This allows the position and orientation differences of different batteries flowing through the conveyor line in the acquired image to be uniformly attributed to the same spatial reference. This enables the area to be identified to correspond pixel-by-pixel with the standard template corresponding to each candidate model under the same coordinate semantics, thus avoiding pixel-by-pixel misalignment caused by position drift from the source.

[0065] (2) By comparing each pixel coordinate in the nameplate coordinate range with the pixel value of all standard templates at the pixel coordinate, the pixel coordinates with pixel value differences are added to the discrimination coordinate set. This allows the subsequent model comparison to only act on the pixel coordinates that are truly distinguishable between candidate models, excluding the non-distinguishing pixel coordinates common to each candidate model from the discrimination range, reducing the dilution effect of common areas on model discrimination, and making the accumulation of matching degree focus on the model difference itself.

[0066] (3) By obtaining the brightness value of the region to be identified at each pixel coordinate in the discrimination coordinate set, and removing the pixel coordinates whose brightness values ​​exceed the preset brightness range from the discrimination coordinate set, an effective discrimination coordinate set is obtained. This allows the pixel coordinates that deviate from the normal imaging range due to factors such as fluctuations in ambient lighting, environmental reflection, and oil stains, water stains, and scratches on the surface of the battery to be identified to not participate in the subsequent matching degree calculation. This identifies and isolates unreliable pixel positions from the region to be identified, reducing the impact of the factors on the model discrimination result.

[0067] (4) By calculating the matching degree of each candidate model on the pixel coordinates contained in the effective discrimination coordinate set, and taking the candidate model with the highest matching degree as the identification result of the battery, the final discrimination can be completed on the pixel subset after the dual filtering of "having model differentiation ability" and "reliable imaging". The identification result benefits from the focus of the discrimination coordinate set on model differences and the elimination of abnormal pixels by the effective discrimination coordinate set, thereby improving the accuracy of product identification of the battery during the transfer process on the conveyor line.

[0068] The following describes an exemplary product identification device for battery transportation provided in an embodiment of this application. Figure 2 This is an exemplary hardware structure diagram of a product identification device for battery delivery provided in an embodiment of this application.

[0069] In some embodiments, the product identification device for battery delivery is a computer device or includes a computer device. The computer device includes a processor, memory, and a network interface connected via a system bus. The processor of the computer device provides computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The database of the computer device stores data. The network interface of the computer device is used to communicate with other external terminals or servers via a network connection. In some embodiments, the network interface can be a wired network interface; in some embodiments, the network interface can also be a wireless network interface. When the computer program is executed by the processor, it implements the methods in the embodiments of this application.

[0070] Those skilled in the art will understand that Figure 2 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0071] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit it. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this application.

[0072] As used in the above embodiments, depending on the context, the term "when..." can be interpreted as meaning "if...", "after...", "in response to determining...", or "in response to detecting...". Similarly, depending on the context, the phrase "when determining..." or "if (the stated condition or event) is interpreted as meaning "if determining...", "in response to determining...", "when (the stated condition or event) is detected", or "in response to detecting (the stated condition or event)".

[0073] In the above embodiments, implementation can be achieved entirely or partially through software, hardware, firmware, or any combination thereof. When implemented using software, it can be implemented entirely or partially in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that integrates one or more available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium (e.g., solid-state drive), etc.

[0074] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. This program can be stored in a computer-readable storage medium, and when executed, it can include the processes described in the above method embodiments. The aforementioned storage medium includes various media capable of storing program code, such as ROM or random access memory (RAM), magnetic disks, or optical disks.

Claims

1. A product identification method for battery-powered product delivery, characterized by, The method includes: Obtain a top view of the battery on the conveyor line, transform the top view to a preset coordinate system, and obtain a corrected image containing pixel coordinates; Based on the preset nameplate coordinate range of the battery in the coordinate system, the region to be identified corresponding to the nameplate coordinate range is extracted from the calibration image; Images of multiple candidate battery models that are located within the coordinate range of the nameplate in the coordinate system are acquired in advance and used as standard templates for each candidate model. Each pixel coordinate in the nameplate coordinate range is compared with the pixel value of all the standard templates at the pixel coordinate, and the pixel coordinates with pixel value differences are added to the discrimination coordinate set; For each pixel coordinate in the discrimination coordinate set, the brightness value of the region to be identified at the pixel coordinate is obtained, and the pixel coordinates with brightness values ​​exceeding a preset brightness range are removed from the discrimination coordinate set to obtain an effective discrimination coordinate set. For each candidate model, the matching degree of the candidate model is calculated based on the pixel value of the region to be identified at the pixel coordinates contained in the effective discrimination coordinate set and the pixel value of the standard template corresponding to the candidate model. The candidate model with the highest matching degree is taken as the identification result of the battery.

2. The method of claim 1, wherein, The step of extracting the region to be identified from the calibrated image corresponding to the nameplate coordinate range based on the preset nameplate coordinate range in the coordinate system includes: The boundary of the top surface of the battery is extracted from the corrected image to obtain the measured profile; Obtain the standard contour of the top surface of the battery in the preset coordinate system; Based on the correspondence between the measured contour and the standard contour, the pose offset of the measured contour relative to the standard contour is determined, and the pose offset includes translation and rotation components. For each pixel coordinate within the coordinate range of the nameplate, the pixel coordinate is transformed according to the pose offset to obtain the actual sampled coordinates corresponding to the pixel coordinates in the corrected image; The pixel value of the corrected image at each of the actual sampling coordinates is used as the pixel value at the corresponding pixel coordinate in the nameplate coordinate range, and the two values ​​are combined to obtain the region to be identified.

3. The method according to claim 1, characterized in that, The step of comparing each pixel coordinate in the nameplate coordinate range with the pixel values ​​of all the standard templates at that pixel coordinate, and adding the pixel coordinates with pixel value differences to the discrimination coordinate set, includes: All the candidate models are paired up to obtain multiple candidate model pairs; For each candidate model pair, the pixel values ​​of the two standard templates corresponding to the two candidate models in the candidate model pair are compared pixel by pixel, and the pixel coordinates with pixel value differences are added to the difference coordinate subset corresponding to the candidate model pair. The subsets of difference coordinates corresponding to all candidate model pairs are merged to obtain the discrimination coordinate set.

4. The method according to claim 3, characterized in that, The step of comparing pixel values ​​pixel by pixel in the two standard templates corresponding to the two candidate models included in the candidate model pair, and adding the pixel coordinates with pixel value differences to the subset of difference coordinates corresponding to the candidate model pair, includes: For each pixel coordinate in the coordinate range of the nameplate, the pixel values ​​of the two standard templates at the pixel coordinates are obtained and recorded as the first pixel value and the second pixel value. Calculate the absolute value of the difference between the first pixel value and the second pixel value; Obtain the pixel values ​​of all pixel positions within the first preset neighborhood of the pixel coordinates for the two standard templates, and use the difference between the maximum and minimum values ​​of the obtained pixel values ​​as the local fluctuation amount corresponding to the pixel coordinates; If the absolute value of the difference is greater than the local fluctuation amount, then the pixel coordinates are added to the subset of difference coordinates corresponding to the candidate model pair.

5. The method according to claim 1, characterized in that, The step of calculating the matching degree corresponding to the candidate model based on the pixel value of the region to be identified at the pixel coordinates contained in the effective discrimination coordinate set and the pixel value of the standard template corresponding to the candidate model includes: For each pixel coordinate in the effective discrimination coordinate set, the pixel value of the region to be identified at the pixel coordinate is compared with the pixel value of the standard template corresponding to all the candidate models at the pixel coordinate. The candidate model corresponding to the standard template whose pixel value is closest to the pixel value of the region to be identified at the pixel coordinate is taken as the winning model corresponding to the pixel coordinate. For each candidate model, the number of pixel coordinates in the effective discrimination coordinate set that use the candidate model as the winning model is counted, and this number is used as the matching degree corresponding to the candidate model.

6. The method according to claim 5, characterized in that, The step of comparing the pixel value of the region to be identified at the pixel coordinates with the pixel values ​​of the standard templates corresponding to all the candidate models at the pixel coordinates, and selecting the candidate model corresponding to the standard template whose pixel value is closest to the pixel value of the region to be identified at the pixel coordinates as the winning model corresponding to the pixel coordinates, includes: Obtain all pixel coordinates within the second preset neighborhood of the pixel coordinates, and use them as the neighborhood coordinate set corresponding to the pixel coordinates; For each neighboring pixel coordinate in the neighborhood coordinate set, the pixel value of the region to be identified at the neighboring pixel coordinate is compared with the pixel value of the standard template corresponding to all the candidate models at the neighboring pixel coordinate. The candidate model corresponding to the standard template whose pixel value is closest to the pixel value of the region to be identified at the neighboring pixel coordinate is taken as the local winning model corresponding to the neighboring pixel coordinate. For each candidate model, the number of neighboring pixel coordinates in the neighborhood coordinate set that use the candidate model as the locally winning model is counted, and this number is taken as the local support number of the candidate model at the pixel coordinate. The candidate model with the largest number of local support numbers among all the candidate models is taken as the winning model corresponding to the pixel coordinates.

7. The method according to claim 1, characterized in that, After selecting the candidate model with the highest matching degree as the identification result of the battery, the method further includes: Obtain a preset working parameter library, which stores the working parameters corresponding to each candidate model; Based on the identification result, the working parameters of the candidate model corresponding to the identification result are retrieved from the working parameter library and used as the target working parameters of the battery. The identification results are associated and bound with the target operating parameters to output the identification profile of the battery.

8. A product identification device for battery-powered products, characterized in that, The battery-delivered product identification device includes: one or more processors and a memory; the memory is coupled to the one or more processors, the memory is used to store computer program code, the computer program code including computer instructions, and the one or more processors invoke the computer instructions to cause the battery-delivered product identification device to perform the method as described in any one of claims 1-7.

9. A computer program product containing instructions, characterized in that, When the computer program product is run on the product identification device that carries the battery, the product identification device that carries the battery performs the method as described in any one of claims 1-7.

10. A computer-readable storage medium comprising instructions, characterized in that, When the instruction is executed on the product identification device that delivers the battery, the product identification device that delivers the battery performs the method as described in any one of claims 1-7.

Images