CN116029315A - Artificial intelligence optical decoding system and method - Google Patents

Abstract

The application discloses artificial intelligence optical decoding system, its characterized in that includes: the image capturing module is used for capturing the image of the bar code label; the image recognition module is electrically connected with the image acquisition module and used for receiving and recognizing the image so as to generate a plurality of character recognition data and a plurality of bar code data. A database for storing a plurality of label templates; the machine learning module is electrically connected with the image identification module and the database, receives the character identification data and the bar code data, establishes the relationship between the character identification data and the bar code data, learns to establish or correct the label template, and coordinates one of the label templates; and the output module is electrically connected with the machine learning module and outputs bar code data according to the matched label template. The artificial intelligence optical decoding system and the method can accurately and automatically match the label template by utilizing machine learning so as to read the bar code label, decode bar code data, automatically output and feed the bar code data into a database.

Description

Artificial intelligence optical decoding system and method

technical field

The present invention relates to an artificial intelligence optical decoding system and method, in particular to an artificial intelligence optical decoding system and method using machine learning.

Background technique

In recent years, optical barcodes (including one-dimensional and two-dimensional barcodes) and optical identification systems have been widely used in industry and daily life. In addition to being used as labels for product-related information (including but not limited to product serial numbers, production history, etc.), they are also often Relevant signs applied to the production system or logistics system to control the manufacturing process or transmission process. Since optical barcodes are easy to produce and identify, and have international standardization, more and more information is marked with optical barcodes. Often there are multiple optical barcodes on an object to record various related information.

The collection of conventional optical barcode label data mostly uses a manual scanner to read the optical barcode data sequentially. It will cause troubles for users to scan, and it is easy to cause reading errors, which reduces the efficiency of operations.

Contents of the invention

In view of the above-mentioned problems to be solved and their reasons, the present invention proposes an artificial intelligence optical decoding system and method that uses machine learning to accurately and automatically match label templates to read barcode labels, decode barcode data, and automatically output, And then fed into the database.

Therefore, one aspect of the present invention proposes an artificial intelligence optical decoding system suitable for barcode labels, including: an image capture module for capturing images of barcode labels; an image recognition module electrically connected to the image capture module The group receives and recognizes images to generate a plurality of character recognition data and a plurality of barcode data. The database is used to store multiple label templates; the machine learning module is electrically connected to the image recognition module and the database, receives character recognition data and barcode data, establishes the relationship between character recognition data and barcode data, and learns to create Or modify the label template and match one of the label templates; and the output module is electrically connected to the machine learning module, and outputs the barcode data according to the matched label template.

According to an embodiment of the present invention, the artificial intelligence optical decoding system further includes an image correction module, which is electrically connected to the image capture module and the image recognition module, and performs image correction after receiving the image from the image capture module, and The corrected image is sent to the image recognition module for recognition.

According to another embodiment of the present invention, in the artificial intelligence optical decoding system, the barcode data is selected from the group consisting of dimensional barcode, quick response matrix code, PDF417 barcode and data matrix.

According to another embodiment of the present invention, the artificial intelligence optical decoding system wherein the character recognition data further includes a plurality of barcode-related characters and at least one characteristic character, wherein each barcode data corresponds to one of the barcode-related characters, and Each of the label templates includes a plurality of fields and at least one template characteristic character, the fields are respectively associated with barcode-related characters, and the template characteristic characters are associated with characteristic characters.

Another aspect of the present invention proposes an artificial intelligence optical decoding method, which is suitable for decoding a barcode label, including: capturing the image of the barcode label; identifying the image to generate multiple character identification data and multiple barcode data; connecting the data library, wherein the database stores a plurality of label templates; establishes the relationship between the character recognition data and the barcode data to learn to create or modify the label templates, and matches one of the label templates; and according to the matched label templates, the barcode Data output.

According to an embodiment of the present invention, after capturing the image of the barcode label in the artificial intelligence optical decoding method, it further includes: performing image correction on the image.

According to another embodiment of the present invention, the artificial intelligence optical decoding method wherein the barcode data is selected from the group consisting of dimensional barcode, quick response matrix code, PDF417 barcode and data matrix.

According to another embodiment of the present invention, the artificial intelligence optical decoding method wherein the character recognition data further includes a plurality of barcode-related characters and at least one characteristic character, wherein each barcode data corresponds to one of the barcode-related characters, and Each of the label templates includes a plurality of fields and at least one template characteristic character, the fields are respectively associated with barcode-related characters, and the template characteristic characters are associated with characteristic characters.

Description of drawings

In order to make the above and other objects, features, advantages and embodiments of the present invention more comprehensible, the accompanying drawings are described as follows:

FIG. 1 is a device schematic diagram of an artificial intelligence optical decoding system according to an embodiment of the present invention.

FIG. 2 is a block diagram of an artificial intelligence optical decoding system according to an embodiment of the present invention.

FIG. 3 is a schematic diagram of a barcode label according to an embodiment of the present invention.

FIG. 4 is a flowchart of an artificial intelligence optical decoding method according to an embodiment of the present invention.

Explanation of reference signs

1: Logistics transmission device 2: Cargo delivery platform

4: Goods 41: Upper surface

410: barcode label 11: computer

13: Cargo sensing frame 14: Display device

20: Artificial intelligence optical decoding system

22: Image capture module 24: Image correction module

26: Image Recognition Module 28: Machine Learning Module

30: Database 32: Output module

34: target media 100: barcode label

110C, 112C, 114C, 116C, 118C, 120C, 122C, 124C, 126C, 128C: Barcode

110A, 112A, 114A, 116A, 118A, 120A, 122A, 124A, 126A, 128A: Barcode related characters

110B, 112B, 114B, 116B, 118B, 120B, 122B, 124B, 126B, 128B: barcode characters

102, 104, 106, 108: characteristic characters

S1, S2, S3, S4, S5, S6, S7: steps

Detailed ways

For the sake of easy understanding of the described embodiments, a number of technical details will be provided below. Of course, not all embodiments require these technical details. Meanwhile, some well-known structures or elements are only drawn schematically in the drawings to appropriately simplify the contents of the drawings.

In order to make the description of the disclosure more detailed and complete, the following provides illustrative descriptions of the implementation aspects and specific embodiments of the present invention; but this is not the only form of implementing or using the specific embodiments of the present invention. The description covers features of various embodiments as well as method steps and their sequences for constructing and operating those embodiments. However, other embodiments can also be used to achieve the same or equivalent functions and step sequences.

Please refer to FIG. 1 , which is a device schematic diagram of an artificial intelligence optical decoding system according to an embodiment of the present invention. The artificial intelligence optical decoding system of the present invention can usually be set in a logistics conveying device 1. The logistics conveying device 1 includes a cargo conveying platform 2 for placing a cargo 4. Generally speaking, the cargo 4 will be placed on the top of the barcode label 410. Surface 41 is placed upwards. A computer 11 can be set beside the logistics conveying device 1 for setting the artificial intelligence optical decoding system of the present invention. On the top of the cargo delivery platform 2, a cargo sensing frame 13 will be installed to set up related devices for the detection of the cargo 4, including the image capture module (not shown) in the artificial intelligence optical decoding system of the present invention, infrared sensing A device or a motion sensor is used to detect that the goods 4 have been delivered to its detection range to activate the artificial intelligence optical decoding system of the present invention, and may also include appropriate lighting to assist the image capture module. In addition, the logistics conveying device 1 can also be equipped with a display device 14, which can display relevant information of the goods, including captured images, reading results of barcode labels, and related detection information, etc. The display device includes a liquid crystal display (LCD display).

Please refer to FIG. 2 , which is a block diagram of an artificial intelligence optical decoding system according to an embodiment of the present invention. The artificial intelligence optical decoding system 20 of the present invention is mainly used to decode barcode labels, and includes an image capture module 22, including a digital camera, or other CCD modules that can be used to capture images. The image capture module 22 is used to capture the image of the barcode label. As shown in FIG. . An image correction module 24 is electrically connected to the image capture module 22, and performs image correction after receiving an image from the image capture module 22. Generally speaking, the positions or angles of the goods delivered by the goods conveying platform are different each time, and the size of the goods is also different. Therefore, in the images of the goods captured by the image capture module 22, The image sizes and angles of barcode labels vary. Through the image correction module 24, the original image captured by the image capture module 22 can be cropped, scaled, rotated and other image corrections can be performed to obtain a proper corrected image for subsequent identification.

The image recognition module 26 is electrically connected to the image capture module 24, receives the corrected image and recognizes the image, so as to generate a plurality of character identification data and a plurality of barcode data. The database 30 is used to store a plurality of label templates, and of course, can also store the results of barcode label interpretation. The machine learning module 28 is electrically connected to the image recognition module 26 and the database 30, receives the character recognition data and barcode data transmitted by the image recognition module 26, establishes the relationship between the character recognition data and the barcode data, and learns to establish or Modify the label templates in the database 30, and match one of the label templates in the database 30, so that these barcode data can be output correctly. The output module 32 is electrically connected to the machine learning module 28, and accurately outputs the barcode data according to the matching label template. As shown in the figure, the output result can be stored in the database 30, or connected to the Internet or B Ethernet, output the barcode data to the target medium 34, such as the database of the client.

Next, please refer to FIG. 3 , which is a schematic diagram of a barcode label according to an embodiment of the present invention. Through this barcode label sample, it will be explained how the artificial intelligence optical decoding system of the present invention can accurately decode and output the barcode data of the barcode label through machine learning. As shown in Figure 3, in general, barcode labels 100 usually print many messages, including barcodes 110C, 112C, 114C, 116C, 118C, 120C, 122C, 124C, 126C, 128C; barcode related characters 110A, 112A, 114A , 116A, 118A, 120A, 122A, 124A, 126A, 128A; barcode characters 110B, 112B, 114B, 116B, 118B, 120B, 122B, 124B, 126B, 128B, and feature characters 102, 104, 106, 108. As mentioned above, the artificial intelligence optical decoding system of the present invention can convert the above-mentioned barcodes 110C-128C and barcode-related characters 110A- 128A, barcode characters 110B～128B and characteristic characters 102, 104, 106, 108 are identified one by one, and character identification data (including barcode related characters, barcode characters and characteristic characters) and barcode data (barcode ). The character recognition data can be generated through optical character recognition method (Optical Character Recognition, OCR). The barcode data is not limited to the one-dimensional barcode (IDBarcode) in this embodiment, but also includes quick response matrix code (QR Code), PDF417 barcode and data matrix (Data Matrix).

In addition to receiving the character recognition data and barcode data sent by the image recognition module, the machine learning module of the present invention also records the relative positions of the character recognition data and barcode data in the image. According to an embodiment of the present invention, when the image recognition module performs recognition, it will record the position of the recognized image in the image, and send it to the machine learning module along with character recognition data and barcode data.

It can be seen from Figure 3 that the barcode characters correspond to the barcode, that is, the barcode characters are the content of the barcode interpretation. For example, the characters and numbers decoded from the barcode 112C are the barcode character 112B "CL8068403359524". The barcode-related characters usually correspond to the barcode and the barcode characters. For example, the barcode-related character 110A is "(V)SUPPLIER", which is the supplier, and the corresponding barcode 110C and barcode character 110B "04195" represent is the supplier number. For another example, the related character 120A of the barcode is "(Q)QTY" representing the quantity Quantity, and the corresponding barcode 120C and the character 120B "696" indicate that the quantity is 696. However, each company has different formats or different abbreviation codes for the related characters of the barcode, but the common feature is that the related characters of the barcode and the barcode or the barcode characters are arranged adjacent to each other. Therefore, the machine learning module of the present invention associates these barcode-related characters, barcodes, and the content of the barcode characters with their relative positions to form samples. The associated samples of these barcodes/barcode characters/barcode-related characters in the entire barcode label will also be associated with the characteristic characters to create samples. Furthermore, some label templates are preset in the database of the present invention, that is to clearly define the meaning of the barcode at the corresponding position in the barcode label, and the data fields that should be output after decoding. For example, the decoded value of the barcode 110C represents the number of the supplier, and the data field that should be output is "SUPPLIER No.", the value decoded by the barcode 120C represents the quantity, and the data field that should be output is "Quantity", etc. . For the machine learning module of the present invention, each barcode definition, corresponding position, and corresponding data field is a sample, which is an association of data, and each label template is a collection of samples. In addition to recording the meaning represented by the barcode at the corresponding position and the data fields that should be output after decoding, these label templates will also record some characteristics of the label template (template characteristic characters), such as which supplier or manufacturer The barcode label of the goods, the barcode label of the goods in which area, etc. As shown in Figure 3, in one embodiment of the present invention, the barcode label also includes some characteristic characters, such as the characteristic character 102 "INTEL" represents that the barcode label is marked as the goods of INTEL company, and the characteristic character 104 "RoHs COMPLIANT, el" means that the goods comply with European RoHs environmental protection standards, and the destination of the goods should be Europe. Characteristic character 106 "ASSEMBLED IN CHINA" means that the goods are assembled in China. For the machine learning module of the present invention, the association between these feature characters and label templates is also a kind of sample, so each label template is a combination of many samples, and these preset label templates in the database are machine Sample database for learning. Through the sample and association of feature characters, the machine learning module can filter out closer label templates.

Those who are familiar with this technology should know that for the above-mentioned barcode labels, each manufacturer may change the position of the barcode or even change the content of the barcode due to different periods and different needs. If the label template needs to be manually adjusted every time it is changed, it will cause inefficiency. , and more and more label template versions are accumulated, and it is not easy to manage. However, the ultimate goal of all barcode reading systems is to output the barcode reading to the correct data field, and the machine learning module of the present invention, through the method of machine learning, can establish the read character recognition data and barcode data The correlation and samples are cross-learned with the correlation and samples of the preset label templates in the database, and the label template corresponding to the barcode label can be correctly matched, and then the decoded barcode is correctly output. Even for the manufacturer's changes to the barcode label, the machine learning module of the present invention can be used to modify the label template in the database, and even create a new label template. According to an embodiment of the present invention, for example, if INTEL company changes the relative position of barcode 120C/barcode character 120B/barcode related character 120A and barcode 128C/barcode character 128B/barcode related character 128A because of demand , that is, the positions of the two barcodes are reversed. The machine learning module of the present invention will match other barcodes/barcode characters/barcode related characters, and feature characters and their associations to the original label template which is the closest , and recognize that the position of the two barcodes has changed, so the content of the barcode can still be accurately output, and the original label template can be corrected, or a new version of the label template can be created. According to another embodiment of the present invention, for another example, if INTEL company changes the barcode-related character 120A corresponding to the barcode 120C due to demand, such as changing from "QTY" to "NO.", the machine learning method of the present invention The module will also infer the barcode 120C corresponding to the barcode related character 120A by correlating with other samples in the closest label template, and the field corresponding to the decoded barcode data is "Quantity".

Machine learning models that can be processed in the machine learning module of the present invention include but are not limited to: supervised learning (linear regression, logistic regression, decision tree, support vector machine, K-neighbor algorithm), cluster analysis, pattern recognition ( K-means algorithm, ensemble learning AdaBoost, Bayesian classifier), dimensionality reduction and metric learning (principal component analysis, autoencoder, linear discriminant analysis), structure prediction, anomaly detection, artificial neural network (feedforward neural network, radial basis function network, recurrent neural network), reinforcement learning type (Markafe chain, Q-learning, Monte Carlo, SARSA (State Action RewardState Action Learning), variational method).

Next, please refer to FIG. 4 , which is a flowchart of an artificial intelligence optical decoding method according to an embodiment of the present invention. As shown in FIG. 4 , the artificial intelligence optical decoding method of the present invention is used to decode a barcode label, such as the barcode label shown in FIG. 3 . As in step S1, firstly the image capture module captures the image of the barcode label. Next, in step S2, the image correction module performs image correction on the image, and sends the corrected image to the image recognition module. In step S3, the image recognition module performs image recognition on the corrected image. As mentioned above, the image recognition includes character recognition (step S4 ) and barcode recognition (step S5 ), so as to generate a plurality of character recognition data and a plurality of barcode data. As in step S6, these character recognition data and barcode data will be sent to the machine learning module, and the machine learning module will connect to a database, wherein the database stores a plurality of label templates; the machine learning module recognizes the data and barcode data (including its location) establishes the relationship between character recognition data and barcode data to learn to create or modify label templates, and automatically match one of the label templates. Next, as in step S7, output the barcode data to the correct field in the label template according to the matching label template.

As mentioned above, in one embodiment of the present invention, the barcode data in the artificial intelligence optical decoding method includes one or a combination of one-dimensional barcode, quick response matrix image code, PDF417 barcode and data matrix. According to another embodiment of the present invention, the character identification data further includes barcode related characters and characteristic characters. Wherein, each barcode data corresponds to one of the barcode-related characters, and each label template includes a plurality of fields and at least one template characteristic character, the fields are respectively associated with the barcode-related characters or barcode, and the template The characteristic characters are associated with the characteristic characters, and are matched by the machine learning module.

In summary, the artificial intelligence optical decoding system and method of the present invention utilizes machine learning to accurately and automatically match label templates to read barcode labels, decode barcode data, and automatically output and then automatically feed into the database , can improve the efficiency of optical decoding.

The present invention is only disclosed in preferred embodiments herein, but anyone skilled in the art should understand that the above embodiments are only used to describe the present invention, and are not intended to limit the scope of patent rights claimed by the present invention. All changes or substitutions that are equal or equivalent to the above-mentioned embodiments should be interpreted as falling within the spirit or scope of the present invention. Therefore, the protection scope of the present invention should be determined by the claims.

Claims (8)

1. An artificial intelligence optical decoding system suitable for bar code labels, comprising:

the image capturing module is used for capturing an image of the bar code label;

the image identification module is electrically connected with the image acquisition module and used for receiving and identifying the image so as to generate a plurality of character identification data and a plurality of bar code data;

a database for storing a plurality of label templates;

the machine learning module is electrically connected with the image identification module and the database, receives the character identification information and the bar code information, establishes the relation between the character identification information and the bar code information, learns to establish or correct the label template, and matches one of the label templates; and

and the output module is electrically connected with the machine learning module and outputs the bar code data according to the matched label template.

2. The system of claim 1, further comprising an image correction module electrically connected to the image capturing module and the image recognition module, wherein the image correction module receives the image from the image capturing module, and then performs image correction and transmits the corrected image to the image recognition module for recognition.

3. The optical decoding system of claim 1, wherein the barcode data is selected from the group consisting of dimensional barcodes, fast response matrix codes, PDF417 barcodes and data matrices.

4. The system of claim 1, wherein the character recognition data further comprises a plurality of bar code related characters and at least one feature character, wherein each of the bar code data corresponds to one of the bar code related characters, and wherein each of the tag templates comprises a plurality of fields and at least one template feature character, wherein the fields are associated with the bar code related characters, and wherein the template feature characters are associated with the feature characters.

5. An artificial intelligence optical decoding method, which is suitable for decoding bar code labels, comprises the following steps:

capturing an image of the bar code label;

identifying the image to generate a plurality of character identification data and a plurality of bar code data;

connecting a database, wherein the database stores a plurality of label templates;

establishing a relation between the character identification information and the bar code information so as to learn to establish or correct the label template and to match one of the label templates; and

outputting the bar code data according to the matched label template.

6. The method of claim 5, wherein capturing the image of the barcode label further comprises: and performing image correction on the image.

7. The method of claim 5, wherein the barcode data is selected from the group consisting of a dimensional barcode, a fast response matrix pattern code, a PDF417 barcode, and a data matrix.

8. The method of claim 5, wherein the character recognition data further comprises a plurality of bar code related characters and at least one feature character, wherein each of the bar code data corresponds to one of the bar code related characters, and wherein each of the tag templates comprises a plurality of fields and at least one template feature character, wherein the fields are associated with the bar code related characters, and wherein the template feature characters are associated with the feature characters.

Images