A code spraying recognition method applied to three-dimensional intelligent detection of steel plates

Abstract

The application discloses a code spraying recognition method applied to three-dimensional intelligent detection of steel plates and relates to the code spraying recognition technical field.The method comprises the following steps: acquiring image data of a product with engraved code spraying; pre-processing the image data to separate and obtain image data of a code spraying character region; performing character cutting on the image data of the code spraying character region to obtain image data of a single code spraying character; identifying the image data of the single code spraying character through a machine learning algorithm; and arranging and combining the identified characters to form a complete code spraying section.The method not only realizes automatic identification of code spraying, improves the accuracy and stability of code spraying recognition, simplifies the code spraying recognition process and greatly reduces the labor intensity of operators, but also is not affected by external use environments and is not limited by hardware platforms, so that the portability is high.

Description

Technical Field

[0001] This invention relates to the field of inkjet printing recognition technology, and in particular to an inkjet printing recognition method for three-dimensional intelligent inspection of steel plates. Background Technology

[0002] In industrial production, coding is often used as a carrier for product tracking and inspection. The coding is printed on the product to form a unique code, and obtaining the code has become a necessary means to control the product process.

[0003] Inkjet coding recognition is typically done manually. However, in practice, factors such as worker skill level, fatigue, and physical condition can affect coding recognition, especially in harsh environments where high temperatures greatly increase the physical burden on workers, directly impacting production efficiency. Furthermore, factors such as the product's viewing angle, product size, and the quality of the coding print also increase the difficulty for workers to recognize the codes, extending the recognition cycle. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to overcome the shortcomings of the prior art and provide a coding recognition method for three-dimensional intelligent inspection of steel plates.

[0005] To solve the above technical problems, the technical solution of the present invention is as follows:

[0006] A method for inkjet printing recognition applied to three-dimensional intelligent inspection of steel plates, comprising,

[0007] Acquire image data of products with inkjet printing markings;

[0008] The image data is preprocessed to separate the image data of the inkjet character area;

[0009] The image data of the inkjet-printed character area is cropped to obtain the image data of a single inkjet-printed character; the image data of the single inkjet-printed character is then identified using a machine learning algorithm.

[0010] The identified characters are arranged and combined to form a complete inkjet code segment.

[0011] As a preferred embodiment of the inkjet printing recognition method for three-dimensional intelligent inspection of steel plates according to the present invention, the step of acquiring image data of the inkjet-printed product includes:

[0012] The image data of the inkjet-printed product is obtained by taking a picture of the product with the inkjet print.

[0013] As a preferred embodiment of the inkjet printing recognition method for three-dimensional intelligent inspection of steel plates according to the present invention, the preprocessing of image data to separate the image data of the inkjet character region includes:

[0014] Enhance the contrast of image data;

[0015] Image data is processed using the whole-local mean difference method;

[0016] Image erosion and dilation methods are used to locate and extract the inkjet character region, and the image data of the inkjet character region is obtained.

[0017] As a preferred embodiment of the inkjet printing recognition method for three-dimensional intelligent inspection of steel plates according to the present invention, wherein: the step of cropping the image data of the inkjet character area to obtain the image data of a single inkjet character includes,

[0018] Binarize the image of the inkjet-coded character area;

[0019] The image of the inkjet-printed character area is divided into several rectangular regions by horizontal and vertical projection methods.

[0020] The segmented rectangular regions are then further segmented and filtered using a threshold method.

[0021] The selected rectangular area is compressed from four directions: top, bottom, left, and right, so that the rectangular border fits the side of the character, thus obtaining the image data of a single inkjet character.

[0022] As a preferred embodiment of the inkjet printing recognition method for three-dimensional intelligent inspection of steel plates described in this invention, before compressing the selected rectangular area in four directions (top, bottom, left, and right) to make the rectangular border fit the side of the character, the method further includes:

[0023] Eight-neighbor noise reduction is performed on the selected rectangular region.

[0024] As a preferred embodiment of the inkjet printing recognition method for three-dimensional intelligent inspection of steel plates described in this invention, the step of performing secondary rectangular segmentation and filtering by setting a threshold for the segmented rectangular region includes:

[0025] For the segmented rectangular region, set a width threshold WidThre and then re-segment it;

[0026] Set a length threshold LenThre, and combine it with a width threshold to filter out erroneous bounding boxes;

[0027] Set a distance threshold DisThre to filter out error frames.

[0028] As a preferred embodiment of the inkjet coding recognition method for three-dimensional intelligent inspection of steel plates according to the present invention, wherein: before recognizing the image data of a single inkjet code character using a machine learning algorithm, the method further includes,

[0029] The machine learning algorithm is trained using training samples.

[0030] As a preferred embodiment of the inkjet printing recognition method for three-dimensional intelligent inspection of steel plates according to the present invention, the step of training the machine learning algorithm using training samples includes:

[0031] Images containing inkjet-printed characters are collected to form a training set;

[0032] The inkjet images in the training set are preprocessed to separate the image data of the inkjet character region. The image data of the inkjet character region is then cropped to obtain the image data of a single inkjet character, which is used as a training sample.

[0033] The training samples are used to train the machine learning algorithm.

[0034] The beneficial effects of this invention are:

[0035] This invention not only achieves automated coding recognition, improves the accuracy and stability of coding recognition, simplifies the coding recognition process, and greatly reduces the labor intensity of operators, but also the method is not affected by the external environment or the limitations of the hardware platform, and has strong portability. Attached Figure Description

[0036] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0037] Figure 1 This is a flowchart illustrating the inkjet printing recognition method for three-dimensional intelligent inspection of steel plates provided by the present invention.

[0038] Figure 2 This is a flowchart illustrating step S103 of the inkjet printing recognition method for three-dimensional intelligent inspection of steel plates provided by the present invention.

[0039] Figure 3 This is a schematic diagram illustrating the process of training a machine learning algorithm in the inkjet printing recognition method for three-dimensional intelligent inspection of steel plates provided by the present invention.

[0040] Figure 4This is a flowchart illustrating the inkjet printing recognition method for three-dimensional intelligent inspection of steel plates provided by the present invention. Detailed Implementation

[0041] To make the content of this invention easier to understand, the invention will be further described in detail below with reference to specific embodiments and accompanying drawings.

[0042] Figure 1 This is a flowchart illustrating a coding recognition method for three-dimensional intelligent inspection of steel plates, provided in an embodiment of this application. The method includes steps S101 to S105, which are described in detail below:

[0043] Step S101: Obtain image data of the inkjet-printed product.

[0044] Specifically, the product with inkjet printing is photographed using a camera to obtain image data of the product with inkjet printing.

[0045] In addition, to further reduce the labor intensity of operators, the camera can be fixed on the gantry, and a photoelectric sensor can be installed at a preset position on the steel plate conveying device. When the steel plate moves to the preset position under the action of the conveying device, it triggers a photoelectric signal. The controller can then transmit a control signal to the camera, causing the camera to take a picture, acquire image data, and upload it.

[0046] Step S102: Preprocess the image data to separate the image data of the inkjet character area.

[0047] Specifically, after receiving the image uploaded by the camera, the controller first preprocesses the image by enhancing contrast and adjusting the mean difference between the whole and local areas to solve the problems of cluttered background and uneven lighting distribution in the initial image.

[0048] After denoising, image erosion and dilation methods are used to locate and extract the inkjet character region in the image, thus separating the inkjet character region.

[0049] Step S103: Crop the image data of the inkjet character area to obtain the image data of a single inkjet character.

[0050] See Figure 2 This step specifically includes the following steps:

[0051] Step S103a: Binarize the image of the separated inkjet character area.

[0052] Step S103b: The image of the inkjet-printed character area is divided horizontally and vertically using horizontal and vertical projection methods to form several rectangular areas, and each character is enclosed within a rectangular frame.

[0053] Step S103c: Perform secondary segmentation and filtering of the segmented rectangular regions by setting thresholds.

[0054] Specifically, for each rectangular area, a width threshold WidThre is set to further segment cases where characters are stuck together; then a length threshold LenThre is set, which, combined with the width threshold, filters out incorrect bounding boxes caused by background spots and long lines; finally, a distance threshold DisThre is set to filter out incorrect bounding boxes caused by noise at the edges of the image.

[0055] Step S103d: Perform eight-neighbor denoising on the selected rectangular region.

[0056] Step S103e: Compress the rectangular area from the top, bottom, left and right directions to make the rectangular border fit the side of the character, that is, ensure that the rectangular border strictly wraps the character, and obtain the image data of a single inkjet character.

[0057] Step S104: Recognize the image data of a single inkjet character using a machine learning algorithm.

[0058] Specifically, a machine learning algorithm is used to identify individual characters in the cropped image to obtain character recognition results. In this embodiment, the machine learning algorithm used is the KNN algorithm. This algorithm employs the cosine similarity formula, digitizing the image into (X1, X2, X3…Xn), and then using Formula 1 to calculate the similarity between two images, thereby achieving character image recognition. Formula 1 is:

[0059]

[0060] It should be noted that the algorithm needs to be trained before character recognition. See [link / reference] Figure 3 Specifically, it includes the following steps:

[0061] Step a: Obtain images containing inkjet characters to form a training set.

[0062] Step b: Take the inkjet images from the training set and preprocess them to separate the image data of the inkjet character region. Then, crop the image data of the inkjet character region to obtain the image data of a single inkjet character, and use it as a training sample.

[0063] Step c: Use the training samples to train the machine learning algorithm model.

[0064] Step S105: Arrange and combine the recognized characters to form a complete inkjet segment.

[0065] Specifically, after recognizing all the inkjet characters, the inkjet characters are positioned using the hyphen “-” based on the characteristics of the inkjet printing on the steel plate. The recognition results are then arranged and combined, and any gaps in the combined segments are filled with the character “0” to form a complete inkjet recognition code segment.

[0066] It should be noted that if the identification code segment does not meet the inkjet printing requirements, the image will be taken again.

[0067] Figure 4 This is a flowchart illustrating the inkjet printing recognition method for three-dimensional intelligent inspection of steel plates provided in this application embodiment.

[0068] Therefore, the technical solution of this application not only realizes the automated recognition of inkjet printing, improves the accuracy and stability of inkjet printing recognition, simplifies the inkjet printing recognition process, and greatly reduces the labor intensity of operators, but also the method is not affected by the external use environment or the limitations of the hardware platform, and has strong portability.

[0069] In addition to the above embodiments, the present invention may have other implementation methods; all technical solutions formed by equivalent substitution or equivalent transformation fall within the protection scope claimed by the present invention.

Claims

1. A method for inkjet printing recognition applied to three-dimensional intelligent inspection of steel plates, characterized in that: include, Acquire image data of products with inkjet printing markings; The image data is preprocessed to separate the image data of the inkjet character region. Specifically, this includes: enhancing the contrast of the image data; processing the image data using the whole-local mean difference method; and locating and extracting the inkjet character region using image erosion and dilation methods to separate the image data of the inkjet character region. The image data of the inkjet-printed character area is cropped to obtain the image data of a single inkjet-printed character. Specifically, this includes: binarizing the image of the inkjet-printed character area; performing a first segmentation of the image of the inkjet-printed character area in the horizontal and vertical directions using horizontal and vertical projection methods to form several rectangular areas; performing a second segmentation and filtering of the segmented rectangular areas using a threshold setting method; compressing the filtered rectangular areas in four directions (top, bottom, left, and right) to make the rectangular borders fit the sides of the characters; and cropping the rectangular borders to obtain the image data of a single inkjet-printed character. The image data of a single inkjet character is identified using machine learning algorithms; The identified characters are arranged and combined to form a complete inkjet code segment. After identifying all the inkjet code characters, the inkjet code hyphen "-" is used for positioning based on the characteristics of the steel plate inkjet code. The identification results are arranged and combined, and any missing positions in the combined segment are filled with the character "0" to form a complete inkjet code identification segment.

2. The method according to claim 1, characterized in that: The acquisition of image data of the inkjet-printed product includes... The image data of the inkjet-printed product is obtained by taking a picture of the product with the inkjet print.

3. The method according to claim 1, characterized in that: Before compressing the selected rectangular area in four directions (top, bottom, left, and right) to make the rectangular border fit the side of the character, the process also includes: Eight-neighbor noise reduction is performed on the selected rectangular region.

4. The method according to claim 3, characterized in that: The step of performing secondary rectangular segmentation and filtering by setting a threshold for the segmented rectangular region includes... For the segmented rectangular region, set a width threshold WidThre and then re-segment it; Set a length threshold LenThre, and combine it with a width threshold to filter out erroneous bounding boxes; Set a distance threshold DisThre to filter out error frames.

5. The method according to claim 1, characterized in that: Before the image data of a single inkjet character is identified using a machine learning algorithm, the process also includes: The machine learning algorithm is trained using training samples.

6. The method according to claim 5, characterized in that: The step of training the machine learning algorithm using training samples includes... Images containing inkjet-printed characters are collected to form a training set; The inkjet images in the training set are preprocessed to separate the image data of the inkjet character region. The image data of the inkjet character region is then cropped to obtain the image data of a single inkjet character, which is used as a training sample. The training samples are used to train the machine learning algorithm.

Images