High-speed online marking method based on dot matrix marking method

Abstract

This invention discloses a high-speed online marking method based on dot matrix marking technology. The method involves pre-collecting all product information and corresponding marking information to be marked; training a neural network model to identify product types using product images; pre-collecting images of each product's packaging and marking the marking positions on the images; using the neural network model to determine the product type on the marking machine and comparing the marking positions using image comparison technology; processing the products based on the position comparison results; the dot matrix marking machine reading the marking image corresponding to the product type from the data storage device and marking the image onto the product packaging using the dot matrix marking method; and using image analysis technology to determine if there are any missed or incorrect markings. This achieves fully automated, unattended marking, significantly reducing the cost of manual control and supervision.

Description

Technical Field

[0001] This invention belongs to the field of dot matrix marking and relates to automatic marking control technology, specifically a high-speed online marking method based on dot matrix marking. Background Technology

[0002] Marking is the process of adding text, images, or other markings to products during production, according to relevant national regulations or the company's own management needs. Examples of marking information include production date, expiration date, and product number. This process is called marking. Currently, the most common marking method is dot matrix marking. However, because different product types require different marking images, current dot matrix marking machines have problems such as being unable to automatically identify product types, correct marking positions, or verify marking images. Furthermore, product identification, marking position correction, and marking image verification all require manual labor, making fully automated marking impossible.

[0003] To address this, a high-speed online marking method based on dot matrix marking is proposed. Summary of the Invention

[0004] This invention aims to solve at least one of the technical problems existing in the prior art. To this end, this invention proposes a high-speed online marking method based on dot matrix marking, which realizes fully automatic marking without human monitoring, thereby greatly reducing the cost of manual control and supervision.

[0005] To achieve the above objectives, according to an embodiment of the first aspect of the present invention, a high-speed online marking method based on a dot matrix marking method is proposed, comprising the following steps:

[0006] Step 1: Collect all product information that needs to be tagged and the corresponding tagging information in advance;

[0007] The product information includes product categories and several product images for each product; the marking information includes the marking pattern to be applied to each product and the steps for marking using a dot matrix marking method;

[0008] Furthermore, the dot matrix marking machine uses a data storage device to save the correspondence between each product and the marked image;

[0009] Step 2: Train a neural network model to identify product categories using product images;

[0010] Step 3: Collect images of each product's packaging in advance, and mark the areas on the images where the markings will be made; these images will serve as the standard marking images, and the marked areas will serve as the standard marking locations.

[0011] Step 4: Use a neural network model to determine the product type on the marking machine, and use image comparison technology to compare whether the marking position of the product is correct; process the product according to the position comparison results;

[0012] Step 5: The dot matrix marking machine reads the marking image corresponding to the product type from the data storage device and uses the dot matrix marking method to mark the image onto the product packaging;

[0013] Step Six: Use image analysis technology to determine if there are any omissions or errors in the markings;

[0014] The training of the neural network model for recognizing product types from product images includes the following steps:

[0015] Step S1: Label each product image with a number;

[0016] Step S2: Input the product image into the neural network model and train the neural network model;

[0017] Step S3: Set a prediction accuracy threshold K in advance based on practical experience; stop training when the prediction accuracy of the neural network model is greater than the prediction accuracy threshold K; and mark the trained neural network model as M.

[0018] The neural network model determines the product type in the following way:

[0019] The dot matrix labeling machine starts its onboard image capture device to acquire images of the products to be labeled in real time; then it inputs the product images into the neural network model M to obtain the output product prediction labels, and then reads the product category corresponding to the digital labels based on the product prediction digital labels;

[0020] The method for comparing whether the product labeling position is correct is as follows:

[0021] Before starting the marking process, the dot matrix marking machine uses its built-in image capture device to acquire an image of the product. Based on the actual positional relationship between the marking device and the product, it analyzes the pre-marking position in the product image. The size of the product in the captured image is then scaled to match the standard marking image. Finally, by comparing the pre-marking position with the standard marking position, it determines whether the pre-marking position is correct.

[0022] The method for processing goods based on location comparison results is as follows:

[0023] For products with correct pre-marking positions, mark them according to the standard dot matrix marking method;

[0024] For products with incorrect pre-marking positions, temporarily suspend labeling and notify the labeling management personnel that some products cannot be labeled.

[0025] The method for determining whether the marking results are missing or incorrect is as follows:

[0026] The dot matrix marking machine uses an image capture device to acquire images of the marked products; and uses image comparison technology to compare whether the images contain the corresponding marked images of the products.

[0027] If not included, notify the labeling management personnel that there are missing labeled items;

[0028] If the product image contains tagged images, extract the tagged images from the product image and count the number of small circular indentations in each row of the tagged images; pre-set the error ratio threshold H and the row number ratio threshold R0 based on practical experience; mark the total number of rows of small circular indentations in the tagged standard image as L; mark each row of small circular indentations as l; mark the number of small circular indentations in the l-th row of the tagged product image as Nl; mark the number of small circular indentations in the l-th row of the tagged standard image as Sl; mark the error rate of the small circular indentations in the l-th row as Rl; then the error rate... The total number of rows whose statistical error rate Rl is less than the error threshold H is marked as X; then the proportion of failed rows is... If the proportion of failed rows If the value is >R0, the management personnel will be notified that there are products that failed to be labeled; otherwise, no action will be taken.

[0029] Compared with the prior art, the beneficial effects of the present invention are:

[0030] This invention pre-collects the correspondence between product information and labeling information, then trains a neural network model to identify product types by training product images. Before labeling with a dot matrix labeling machine, the product type is identified, and the corresponding labeling image is matched. The product labeling position is further verified using image comparison technology. If the labeling position is inaccurate, labeling is temporarily suspended. After the product is labeled, the printed image is analyzed to automatically determine whether there are any omissions or errors in labeling.

[0031] Currently, there are existing technologies that use image recognition technology to identify the objects to be marked and then mark them accordingly. However, this method only considers marking the goods to reduce human intervention. In the actual marking process, it is also necessary to consider the problem that the marking position may change due to the deflection of the goods during transportation, as well as the situation of missed or incorrect marking due to the failure of the marking device. The discovery of the above problems often still needs to be done manually. Therefore, the above methods cannot completely realize unmanned and automated marking.

[0032] This invention combines neural network technology and image comparison technology, and proposes a calculation method to determine whether the labeled image is erroneous, thereby realizing fully automatic labeling without human monitoring, which greatly reduces the cost of human control and supervision. Attached Figure Description

[0033] Figure 1 This is a flowchart of the present invention. Detailed Implementation

[0034] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0035] like Figure 1 As shown, a high-speed online marking method based on dot matrix marking includes the following steps:

[0036] Step 1: Collect all product information that needs to be tagged and the corresponding tagging information in advance;

[0037] Preferably, the product information includes product categories and several product images for each product; the marking information includes the marking pattern to be applied to each product and the steps for marking using a dot matrix marking method;

[0038] Furthermore, the dot matrix marking machine uses a data storage device to save the correspondence between each product and the marked image;

[0039] Step 2: Train a neural network model to identify product types using product images; preferably, the neural network model can be a CNN neural network model.

[0040] Step 3: Collect images of each product's packaging in advance, and mark the areas on the images where the markings will be made; these images will serve as the standard marking images, and the marked areas will serve as the standard marking locations.

[0041] Step 4: Use a neural network model to determine the product type on the marking machine, and use image comparison technology to compare whether the marking position of the product is correct; process the product according to the position comparison results;

[0042] Step 5: The dot matrix marking machine reads the marking image corresponding to the product type from the data storage device and uses the dot matrix marking method to mark the image onto the product packaging;

[0043] Step Six: Use image analysis technology to determine if there are any omissions or errors in the markings;

[0044] The most commonly used marking method is dot matrix marking. However, due to the different types of goods, the marking images are different. Current dot matrix marking machines may not be able to automatically identify the type of goods, correct the marking position, or verify the marking images.

[0045] The training of the neural network model for recognizing product types from product images includes the following steps:

[0046] Step S1: Label each product image with a number;

[0047] The preferred method for labeling each product image with a number is as follows:

[0048] If the number of product categories is labeled as N, then the product images are labeled as 1, 2, 3...N according to their categories, with each number being the actual label of the image.

[0049] Step S2: Input the product image into the neural network model and train the neural network model; preferably, the neural network outputs the predicted image label and uses the real label of the image as the prediction target; that is, the prediction is correct if the predicted image label is the same as the real label value; furthermore, the neural network model uses the prediction accuracy as the training target.

[0050] Step S3: Set a prediction accuracy threshold K in advance based on practical experience; stop training when the prediction accuracy of the neural network model is greater than the prediction accuracy threshold K; and mark the trained neural network model as M.

[0051] The neural network model determines the product type in the following way:

[0052] The dot matrix labeling machine starts its onboard image capture device to acquire images of the products to be labeled in real time; then it inputs the product images into the neural network model M to obtain the output product prediction labels, and then reads the product category corresponding to the digital labels based on the product prediction digital labels;

[0053] The method for comparing whether the product labeling position is correct is as follows:

[0054] Before starting the marking process, the dot matrix marking machine uses its built-in image capture device to acquire an image of the product. Based on the actual positional relationship between the marking device and the product, it analyzes the pre-marking position in the product image. The size of the product in the captured image is then scaled to match the standard marking image. Finally, by comparing the pre-marking position with the standard marking position, it determines whether the pre-marking position is correct.

[0055] The method for processing goods based on location comparison results is as follows:

[0056] For products with correct pre-marking positions, mark them according to the standard dot matrix marking method;

[0057] For products with incorrect pre-marking positions, the marking of the products will be temporarily suspended, and the marking management personnel will be notified that some products cannot be marked; preferably, the marking management personnel will be notified by voice notification or by sending a notification message to the marking management personnel's smart mobile device via wireless network.

[0058] In a preferred embodiment, the dot matrix marking method uses one or more small lasers to emit light pulses simultaneously. After passing through a reflector and a focusing lens, one or more laser pulses ablate uniform and fine pits on the surface of the workpiece. The characters and patterns marked by the laser are composed of multiple small round pits.

[0059] Furthermore, the method for determining whether the marking results were missed or mislabeled is as follows:

[0060] The dot matrix marking machine uses an image capture device to acquire images of the marked products; and uses image comparison technology to compare whether the images contain the corresponding marked images of the products.

[0061] If not included, notify the labeling management personnel that there are missing labeled items;

[0062] If the product image contains tagged images, extract the tagged images from the product image and count the number of small circular indentations in each row of the tagged images; pre-set the error ratio threshold H and the row number ratio threshold R0 based on practical experience; mark the total number of rows of small circular indentations in the tagged standard image as L; mark each row of small circular indentations as l; mark the number of small circular indentations in the l-th row of the tagged product image as Nl; mark the number of small circular indentations in the l-th row of the tagged standard image as Sl; mark the error rate of the small circular indentations in the l-th row as Rl; then the error rate... The total number of rows whose statistical error rate Rl is less than the error threshold H is marked as X; then the proportion of failed rows is... If the proportion of failed rows If the value is >R0, the management personnel will be notified that there are products that failed to be labeled; otherwise, no action will be taken.

[0063] The above embodiments are only used to illustrate the technical methods of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical methods of the present invention without departing from the spirit and scope of the technical methods of the present invention.

Claims

1. A high-speed online marking method based on dot matrix marking, characterized in that, Includes the following steps: Step 1: Collect all product information that needs to be tagged and the corresponding tagging information in advance; Step 2: Train a neural network model to identify product categories using product images; Step 3: Collect pictures of each product's packaging in advance, and mark the locations on the pictures where the markings need to be applied; Step 4: Use a neural network model to determine the product type on the marking machine, and use image comparison technology to compare whether the marking position of the product is correct; process the product according to the position comparison results; Step 5: The dot matrix marking machine reads the marking image corresponding to the product type from the data storage device and uses the dot matrix marking method to mark the image onto the product packaging; Step Six: Use image analysis technology to determine if there are any omissions or errors in the markings; The method to check if the product label position is correct is as follows: Before marking begins, the dot matrix marking machine uses its built-in image capture device to acquire an image of the product. Based on the actual positional relationship between the marking device and the product, it analyzes the pre-marking position in the product image and determines whether the pre-marking position is correct by comparing it with the standard marking position. The method for processing goods based on location comparison results is as follows: For products with correct pre-marking positions, mark them according to the standard dot matrix marking method; For products with incorrect pre-marking positions, temporarily suspend labeling and notify the labeling management personnel that some products cannot be labeled. The method for determining whether the marking results are missing or incorrect is as follows: The dot matrix marking machine uses an image capture device to acquire images of the marked products; and uses image comparison technology to compare whether the images contain the corresponding marked images of the products. If not included, notify the labeling management personnel that there are missing labeled items; If the product image contains a tagged image, extract the tagged image from the product image and count the number of small round indentations in each row of the tagged image; pre-set the error ratio threshold H and the row number ratio threshold R0 based on practical experience; mark the total number of rows of small round indentations in the standard tagged image as L; mark each row of small round indentations as l; mark the number of small round indentations in the l-th row of the tagged product image as Nl; mark the number of small round indentations in the l-th row of the standard tagged image as Sl. Let Rl be the error rate of the small concave dots in the l-th row; then the error rate... The total number of rows whose statistical error rate Rl is less than the error threshold H is marked as X; then the proportion of failed rows is... If the failure rate r > R0, then notify the management personnel that there are products that failed to be labeled.

2. The high-speed online marking method based on dot matrix marking according to claim 1, characterized in that, The product information includes product categories and several product images for each product; the marking information includes the marking pattern to be applied to each product and the steps for marking using a dot matrix marking method.

3. The high-speed online marking method based on dot matrix marking according to claim 1, characterized in that, The training of the neural network model for recognizing product types from product images includes the following steps: Step S1: Label each product image with a number; Step S2: Input the product image into the neural network model and train the neural network model; Step S3: Set the prediction accuracy threshold K in advance based on actual experience; stop training when the prediction accuracy of the neural network model is greater than the prediction accuracy threshold K; and mark the trained neural network model as M.

4. The high-speed online marking method based on dot matrix marking as described in claim 1, characterized in that, The neural network model determines the product type in the following way: The dot matrix labeling machine starts its onboard image capture device to acquire images of the products to be labeled in real time; then it inputs the product images into the neural network model M to obtain the output product prediction labels, and then reads the product category corresponding to the digital labels based on the product prediction digital labels.

Images