TWM653513U - System for inspecting graphic code quality - Google Patents

Abstract

A system for inspecting a graphic code quantity is provided. The system includes a computer device, a machine station and a camera device. The machine station uses a fixture to fix a graphic code test object having a graphic code. The computer device drives a light source to provide a light and activates the camera device to take a picture of the test object so as to generate a graphic code image. After that, when the system obtains the graphic code image, a graphic-identification model is used to identify category of the graphic code and identify a quality level of the graphic code.

Description

Graphics code quality inspection system

The specification discloses a graphic code quality detection technology, in particular, a graphic code quality detection system that uses a deep learning method to identify the graphic code classification and then performs optical detection based on the graphic code classification.

The common use of barcodes (including one-dimensional barcodes, two-dimensional barcodes, such as QR (quick response code) codes, etc.) is to write information according to a specific coding format, so that users can use barcode readers to read the information recorded therein, such as product numbers, product information, web addresses, and various identification information. The quality of barcodes will affect the accuracy and availability of the read data. However, the barcode printing process often has poor printing such as reprinting, missing printing, wrong printing, or blurred printing due to mechanical equipment and external environment. In addition, the quality of the surface of various objects or paper on which the barcode is printed also affects the success or failure of data reading.

The prior art has proposed a barcode detection technology, wherein the machine includes a base, a control console, a conveyor mechanism installed on the base, and a barcode reader installed near the conveyor mechanism. The conveyor mechanism uses a driving element therein to drive the detection barcode tape, which enables the barcode reader to read the data of a single barcode on the barcode tape in sequence, and then judge whether the barcode is poorly printed based on the result of reading the barcode. If an error result is generated based on the reading of the barcode, it is marked as a poorly printed barcode.

In order to ensure the quality of barcodes, a quality inspection process is required when barcodes are printed. The book proposes a graphic code quality inspection system using deep learning methods. The inspection method is applicable to various graphic codes such as one-dimensional barcodes and two-dimensional barcodes that can be represented graphically and scanned to identify information.

According to the embodiment of the graphic code quality detection system, the graphic code quality detection system includes a computer device, a machine and a photographic device. The machine is provided with a fixture for fixing a graphic code object to be tested having a graphic code. The computer device drives the light source to provide light, and the photographic device is activated to shoot the graphic code object to be tested to obtain a graphic code detection image. After obtaining the graphic code detection image, the graphic code detection image is identified to obtain the classification of the graphic code and distinguish the quality level of the graphic code.

The computer device uses software and hardware to implement functional modules, including an image capture module with a camera (including a lens, a photosensitive element and other electronic components) and a light source to capture graphics to generate graphics detection images; a deep learning module that uses a deep neural network to learn graphics of multiple classifications to train a graphics recognition model, and the graphics recognition model is used to recognize and classify the graphics; and an optical detection module that executes a graphics quality detection process corresponding to the classification of the graphics.

Thus, in the image quality detection method operated by the system, after the image acquisition module obtains the image of the image detection, the image recognition model is used to identify the image of the image detection, and the classification of the image code can be obtained. The optical detection module detects the image of the image detection according to the classification of the image code to distinguish the quality level of the image code.

Furthermore, the classification of the graphic codes includes a variety of one-dimensional and two-dimensional graphic codes represented by graphics. Thus, in the process of training the graphic recognition model by the deep learning module, image data of a variety of graphic codes are obtained and divided into a training set used for the deep neural network and a verification set used to verify the graphic recognition model.

The deep learning module can use a residual network to learn a variety of graphics images of different sizes, colors, and distortions, and use supervised learning methods to train image recognition models for various graphics.

Preferably, the classification of the graphic code is the classification specified by the International Organization for Standardization (ISO), so that when the graphic recognition model recognizes the graphic code detection image, it will obtain the classification of the graphic code that meets the standards of the International Organization for Standardization. The optical detection module then performs quality detection based on the quality items specified by the International Organization for Standardization to obtain multiple quality levels of the graphic code.

When the test results are generated, the module can display multiple quality levels of the graphic code, especially multiple quality levels that meet the standards of the International Organization for Standardization.

Furthermore, the detection result can be fed back via an electrical signal, a warning light or a computer screen of a computer device to indicate whether the graphic code meets the quality standard, and the computer screen can display a graphical user interface, which is divided into a graphic code image preview area, a quality level display area and a quality parameter display area for displaying multiple quality levels of the graphic code.

The quality parameter items displayed in the quality parameter display area may include decoding, axial inconsistency, grid inconsistency, symbol contrast, modulation ratio, reflection amplitude, inherent pattern damage, unused error correction and overall logo, etc.

To further understand the features and technical contents of this new model, please refer to the following detailed description and drawings of this new model. However, the drawings provided are only used for reference and description and are not used to limit this new model.

10: Graphic code object to be tested

100: Fixture

101: Light source

103:Photographic equipment

105:Computer equipment

201: Image capture module

203:Graphic Recognition Model

205: Deep Learning Module

207: Optical detection module

209: Display module

50: Graphical User Interface

501: Graphics image preview area

503: Operation button

505: Quality level display area

507: Quality parameter display area

40: Image chart

Steps S301~S309: Graphics code quality inspection process

Figure 1 shows a schematic diagram of a hardware configuration implementation example of a quality detection system; Figure 2 shows a diagram of a functional module implementation example of a graphic code quality detection system; Figure 3 shows a flowchart of a graphic code quality detection method implementation example; Figure 4 shows examples of various types of graphic codes; and Figure 5 shows a schematic diagram of a user interface implementation example of a graphic code quality detection system performing detection.

The following is a specific implementation example to illustrate the implementation of this creation. Technical personnel in this field can understand the advantages and effects of this creation from the content disclosed in this manual. This creation can be implemented or applied through other different specific implementation examples. The details in this manual can also be modified and changed based on different viewpoints and applications without deviating from the concept of this creation. In addition, the attached figures of this creation are only for simple schematic illustration and are not depicted according to actual size. Please note in advance. The following implementation method will further explain the relevant technical content of this creation in detail, but the disclosed content is not used to limit the scope of protection of this creation.

It should be understood that although the terms "first", "second", "third" and so on may be used in this article to describe various components or signals, these components or signals should not be limited by these terms. These terms are mainly used to distinguish one component from another component, or one signal from another signal. In addition, the term "or" used in this article may include any one or more combinations of the related listed items depending on the actual situation.

The disclosure document proposes a graphic code quality detection system, which adopts deep learning technology and uses a deep learning network to learn various types of graphic code images, including one-dimensional barcodes and two-dimensional barcodes, which are represented by graphics and can obtain information by scanning. The model trained by the deep learning network can accurately classify the graphics code and perform quality detection. The graphic code quality includes the bad conditions such as reprinting, missing printing, wrong printing, blurred printing and paper damage caused by mechanical equipment and environmental factors. In addition, in one embodiment, the graphic code quality detection system further performs quality detection according to the quality items specified by the International Organization for Standardization (ISO) (such as ISO/IEC15415).

FIG1 is a schematic diagram showing a hardware configuration implementation example of the image code quality detection system proposed in the disclosure. The example shown in the figure is for illustrative purposes only and is not intended to limit the implementation method of the image code quality detection system.

The figure shows that the graphic code quality inspection system provides a machine, which is equipped with a fixture 100 for fixing the graphic code object 10 to be tested. The graphic code object 10 can be a graphic code image printed on the surface of a specific material. When performing optical inspection, the computer device 105 serves as a main control computer, connected to the camera device 103, drives the light source 101 to provide light, and activates the camera device 103 to shoot the graphic code object 10 to obtain the graphic code inspection image.

Computer device 105 represents a system for running various forms of image code quality detection methods, in which software and hardware can collaborate to implement various functional modules for running image code quality detection. Please refer to the functional module implementation example diagram in the image code quality detection system shown in Figure 2.

According to the embodiment of the graphic code quality detection system, the main components include an image capture module 201 implemented by software and hardware. The image capture module 201 includes a camera device, such as the camera device 103 in FIG. 1, which includes various lenses, photosensitive elements and other electronic components not shown in the figure, and a light source 101 as shown in FIG. 1, which is used to capture a graphic code printed on a graphic code to be tested to generate a graphic code detection image.

The image code quality detection system proposes a deep learning module 205, in which a processor in a computer system executes a deep learning algorithm and uses a deep neural network to learn multiple types of image codes to train an image recognition model 203. The image recognition model 203 is mainly used to recognize and classify image codes. Furthermore, the deep learning module 205 can obtain a large number of image code images of different categories from the image capture module 201, and divide them into a training dataset for the deep neural network and a validation dataset for verifying the image recognition model.

According to the embodiment, the different classifications of the graphic code images obtained by the deep learning module 205 can be classifications of various one-dimensional and two-dimensional graphic codes represented by graphics under the International Organization for Standardization (ISO) standards. Therefore, in the process of the deep learning module 205 training the graphic recognition model 203, the image data of various graphic codes specified by the International Organization for Standardization are divided into the training set and the verification set.

The deep learning module 205 uses deep learning to learn the training set of graphic code images of various sizes, colors and with damages provided by the graphic code quality detection system under the standards of the International Organization for Standardization. The supervised learning method is used to perform feature learning on the images in the training set. The feature value of each pixel in the graphic code images of various categories is obtained through deep learning. The correlation is established through the pixel feature values, and screening and classification are performed to train the graphic recognition model 203.

To form a training set for the deep neural network, a training set is generated according to the graphic codes of each category under the International Organization for Standardization standard. For example, FIG. 4 shows a training set of each category (such as Code 128, Code 39, Code 93, EAN-13, QR, and DotCode). A graphic recognition model of each category is established, and then the graphic recognition model 203 is obtained by integrating the training sets.

Furthermore, the recognition results of the graphic recognition model 203 are repeatedly verified with the verification sets generated by each graphic code classification to adjust the operating parameters in the graphic recognition model 203 and establish a graphic recognition model 203 that can accurately recognize the graphic code classification.

It is worth mentioning that in the process of training a deep neural network model, as the number of layers increases, the gradient may become smaller, resulting in the inability to effectively learn the model to be trained. Therefore, a residual network (Residual Network, ResNet) can be used. The residual network (ResNet) is a deep convolutional neural network that can learn the difference between the image features obtained by deep learning and the image of the original input graphic code, that is, the residual. By learning the difference between the output and input images, the complexity of the training model can be reduced, and various types of graphic code image features can be effectively learned.

The graphic code quality detection system includes an optical detection module 207, which can execute the corresponding graphic code quality detection process according to the classification of the graphic code identified by the graphic recognition model 203. According to the embodiment, the graphic code quality detection system uses the optical imaging principle to produce a graphic code detection image, and uses the universal graphic recognition model 203 to perform recognition according to the defined multiple graphic code classifications to detect various quality levels of the graphic code, and the quality levels can be based on the quality levels specified by the International Organization for Standardization.

Finally, the display module 209 of the image code quality detection system displays the detection result through the user interface. For example, the detection result can reply whether the image code meets the quality specification through interfaces such as electrical signals, warning lights, and computer screens of computer devices. In one embodiment, the display module 209 provides a display interface, which can display the detection results of multiple quality levels of the image code on the display interface in the form of graphical charts.

FIG3 shows a flow chart of an embodiment of the detection method implemented by multiple functional modules in the above-mentioned graphic code quality detection system.

Initially, the image capture module is used to capture the graphic code printed on the surface of a specific material to obtain a graphic code detection image (step S301). Then, the graphic recognition model trained by using a deep neural network to learn multiple types of graphic codes is used to recognize and classify the graphic code detection image to obtain the classification of the graphic code (step S303).

Then, the corresponding code quality detection process (step S305) is performed according to the classification of the identified code using an optical detection method. According to one embodiment, after the classification of the code is obtained, the optical detection module can scan the code detection image to obtain a reflectivity curve. One of the methods is to use the Scan Reflectance Profile (SRP) as the detection basis. The corresponding detection method is more in line with the barcode quality level specified by the International Organization for Standardization (ISO/IEC15416) in 2000, which is a universal quality standard for printed linear barcodes.

Next, quality inspection is performed according to the quality items specified by the International Organization for Standardization. The quality level of the code is obtained by inspecting the code inspection image (step S307). The inspection result is then generated and output (step S309). The displayed inspection result can be used to respond to whether the code meets the quality specifications through interfaces such as electrical signals, warning lights, and computer screens.

According to the above-mentioned embodiment, the classification of graphic codes may include various graphic codes of one dimension and two dimensions specified by the International Organization for Standardization. The example diagram of various types of graphic codes shown in FIG. 4 may be referred to next.

FIG4 shows an image chart 40 of various graphic codes specified by the International Organization for Standardization. Common graphic codes can be divided into multiple categories (e.g., 15 categories). The figure only lists some examples, including Code 128, Code 39, Code 93, EAN-13, QR, and DotCode. According to the illustrated example, in the process of deep learning training the graphic recognition model, image data of various graphic codes specified by the International Organization for Standardization as shown in the figure are obtained and divided into a training set and a verification set for use in a deep neural network.

After the image code recognition is completed, the classification of the image code that meets the standards of the International Organization for Standardization can be obtained. Then, the quality test is performed according to the quality items specified by the International Organization for Standardization using an optical test method to obtain multiple quality levels of the image code. Finally, the test results can be displayed on a display. The test results can refer to the user interface implementation example of the image code quality test system performing the test shown in Figure 5.

FIG. 5 shows a graphical user interface 50 displayed on a computer screen, which can be divided into a graphic code image preview area 501, a quality level display area 505 (in this example, the quality level is A), and a quality parameter display area 507 that displays multiple quality levels of the graphic code in a graphical manner.

The graphical user interface 50 can be run on the control computer of the system. The graphical user interface 50 also provides operation buttons 503, such as configuration, standby, photo taking and exit function buttons as shown in the figure, so that the operator can use them to perform the test. The quality parameter display area 507 can display the ISO quality parameter embodiment of the graphic code quality detection system as shown in Table 1 below.

Table 1 shows the ISO15415 quality parameters, including decoding, axial non-uniformity, grid non-uniformity, symbol contrast, modulation, reflectance margin, fixed pattern damage, unused error correction, and overall marking. This chart allows users to confirm the quality level of the graphic code at a glance, providing a convenient way to check during large-scale testing.

In summary, according to the above-mentioned embodiment of the graphic code quality detection system, the system uses a deep learning method to learn the characteristics of various graphic codes through a computer system, establishes a classification model, and can classify the graphic codes under detection, and can further judge the quality of the graphic codes, especially the quality standards specified by the International Organization for Standardization (ISO) can be used for detection.

The above disclosed contents are only the preferred feasible embodiments of the present invention, and do not limit the scope of the patent application of the present invention. Therefore, all equivalent technical changes made by using the contents of the specification and drawings of the present invention are included in the scope of the patent application of the present invention.

10: Graphic code object to be tested

100: Fixture

101: Light source

103:Photographic equipment

105:Computer equipment

Claims (10)

A graphic code quality detection system includes: a computer device; and a camera connected to the computer device, the computer device drives a light source to provide light, and the camera is activated to shoot a graphic code object fixed by a fixture of a machine to obtain a graphic code detection image; wherein, in the graphic code quality detection system, after obtaining the graphic code detection image, the graphic code detection image is identified to obtain a classification of a graphic code of the graphic code object to be tested, and to distinguish the quality level of the graphic code. The image code quality detection system as described in claim 1, wherein, in the computer device, the functional modules implemented by software and hardware collaboration include: an image capture module, which generates the image code detection image by photographing the image code; a deep learning module, which uses a deep neural network to learn multiple classifications of image codes to train an image recognition model for recognizing and classifying the image code; and an optical detection module, which executes an image code quality detection process corresponding to the classification according to the classification of the image code to distinguish the quality level of the image code. A graphic code quality detection system as described in claim 2, wherein the classification of the graphic code includes a variety of one-dimensional and two-dimensional graphic codes represented by graphics. As described in claim 3, the image quality detection system, wherein, in the process of the deep learning module training the image recognition model, the image data of the multiple image codes are obtained and divided into a training set used for the deep neural network and a verification set used to verify the image recognition model. As described in claim 4, the deep learning module uses a residual network to learn a variety of different sizes, colors and damaged code images provided by the graphic code quality detection system, and uses a supervised learning method to train the graphic recognition model for various graphics. As described in claim 2, the graphic code quality detection system, wherein the graphic recognition model is used to identify the graphic code detection image to obtain the classification of the graphic code that complies with the standards of an international standardization organization, and then the optical detection module is used to perform quality detection according to the quality items specified by the international standardization organization to obtain multiple quality levels of the graphic code. The graphic code quality detection system as described in claim 6 further includes a display module for displaying a detection result generated by the optical detection module, wherein the detection result displays multiple quality levels of the graphic code in a graph. A graphic code quality detection system as described in claim 7, wherein the detection result returns whether the graphic code meets the quality standard through an electrical signal, a warning light or a computer screen of the computer device. A graphic code quality detection system as described in claim 8, wherein the computer screen displays a graphical user interface, which is divided into a graphic code image preview area, a quality level display area, and a quality parameter display area for displaying the multiple quality levels of the graphic code. A graphic code quality detection system as described in claim 9, wherein the quality parameter items displayed in the quality parameter display area include decoding, axial inconsistency, grid inconsistency, symbol contrast, modulation ratio, reflection amplitude, inherent graphic damage, unused error and overall flag.

Images