WO2023011155A1

WO2023011155A1 - Barcode identification method and apparatus, computer device and storage medium

Info

Publication number: WO2023011155A1
Application number: PCT/CN2022/106288
Authority: WO
Inventors: 姚恒志; 苏步升; 杨泽同; 赵泽林; 刘枢; 吕江波; 沈小勇
Original assignee: 深圳思谋信息科技有限公司; 上海思谋科技有限公司
Priority date: 2021-08-02
Filing date: 2022-07-18
Publication date: 2023-02-09
Also published as: CN113627207A; CN113627207B

Abstract

A barcode identification method and apparatus, a computer device and a storage medium, the method comprising: inputting a barcode image into a pre-built target neural network model to obtain a predicted sequence carrying a code system identifier (21); decoding the predicted sequence to obtain a target prediction sequence comprising coding information (22); and identifying the target prediction sequence according to coding rules and coding information matching the code system identifier so as to obtain an identification result (23).

Description

Barcode recognition method, device, computer equipment and storage medium

This application claims the priority of the Chinese patent application with the application number 202110881245.X and the application name "barcode recognition method, device, computer equipment and storage medium" submitted to the China Patent Office on August 2, 2021, the entire content of which is passed References are incorporated in this application.

technical field

The present application relates to the technical field of image processing, in particular to a barcode recognition method, device, computer equipment and storage medium.

Background technique

In industrial applications, the use of barcodes is becoming more and more widespread, and the demand for the use of barcodes is increasing. The rapid development of bar code technology has derived a variety of types.

In the existing technology, usually a production scene corresponds to a set of barcode recognition schemes; if you need to switch to other barcode recognition schemes, you need to manually set them up; the existing barcode recognition methods are still relatively simple and cannot adapt to changing industrial production scenarios , leading to the low efficiency of existing barcode recognition.

Contents of the invention

According to various embodiments disclosed in this application, a barcode image restoration method, device, computer equipment and storage medium are provided.

In a first aspect, the application provides a barcode recognition method, including:

Inputting the barcode image into the pre-built target neural network model to obtain a predicted sequence output by the target neural network model; wherein, the predicted sequence carries a code system identification;

Decoding the predicted sequence to obtain a target predicted sequence including coding information;

The target prediction sequence is identified according to the encoding rule matched with the coding system identifier and the encoding information, and an identification result is obtained.

In a second aspect, the present application also provides a barcode recognition device, including:

The sequence output module is used to input the barcode image into the pre-built target neural network model to obtain the predicted sequence output by the target neural network model; wherein the predicted sequence carries a code system identification;

A sequence decoding module, configured to decode the predicted sequence to obtain a target predicted sequence containing coding information;

A sequence identification module, configured to identify the target predicted sequence according to the coding rule matched with the coding system identifier and the coding information, and obtain a recognition result.

In a third aspect, the present application also provides a computer device, including a memory and a processor, the memory stores computer-readable instructions, and the processor implements the following steps when executing the computer-readable instructions:

In a fourth aspect, the present application also provides a computer-readable storage medium, on which computer-readable instructions are stored, and when the computer-readable instructions are executed by a processor, the following steps are implemented:

In the fifth aspect, the present application also provides a computer program product, including a computer program, and when the computer program is executed by a processor, the following steps are implemented:

Input the barcode image into the pre-built target neural network model to obtain the predicted sequence output by the target neural network model; wherein, the predicted sequence carries a code system identification;

Decoding the predicted sequence to obtain a target predicted sequence containing encoded information;

According to the encoding rules and encoding information matched with the coding system identifier, the target prediction sequence is identified to obtain the identification result.

The details of one or more embodiments of the application are set forth in the accompanying drawings and the description below. Other features and advantages of the application will be apparent from the description, drawings, and claims.

Description of drawings

In order to better describe and illustrate embodiments and examples of the inventions disclosed herein, reference may be made to one or more of the accompanying drawings. The accompanying specification or examples used to describe the drawings should not be considered as limitations on the scope of any of the disclosed inventions, the presently described embodiments and examples, and the best mode of these inventions currently understood.

Fig. 1 is an application environment diagram of a barcode recognition method in an embodiment.

Fig. 2 is a schematic flowchart of a barcode recognition method in an embodiment.

Fig. 3 is a schematic flowchart of the steps of obtaining a target prediction sequence including encoding information in an embodiment.

Fig. 4 is a schematic flowchart of the step of identifying the target prediction sequence in an embodiment.

Fig. 5 is a schematic flowchart of the steps of training the initial neural network model in one embodiment.

Fig. 6 is a structural block diagram of a barcode recognition device in an embodiment.

Figure 7 is an internal block diagram of a computer device in one embodiment.

Detailed ways

In order to make the purpose, technical solution and advantages of the present application clearer, the present application will be further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described here are only used to explain the present application, and are not intended to limit the present application.

The barcode recognition method provided by this application can be independently applied to the terminal device deployed with the target neural network model; specifically, the terminal device captures a barcode image containing barcode information; the terminal device inputs the barcode image into the pre-deployed terminal device The target neural network model obtains the prediction sequence output by the target neural network model; the prediction sequence carries a code system identification; the terminal device decodes the prediction sequence to obtain the target prediction sequence containing coding information; the terminal device matches the code system identification The coding rules and coding information are used to identify the target prediction sequence and obtain the recognition result.

The barcode recognition method provided in this application can also be applied to the application environment shown in FIG. 1 . Wherein, the terminal 11 communicates with the server 12 through the network. The terminal 11 sends the barcode image to the server 12, and the server 12 inputs the received barcode image into the pre-built target neural network model to obtain the predicted sequence output by the target neural network model; the predicted sequence carries a code system identification; the server 12 pairs the predicted sequence Perform decoding processing to obtain the target prediction sequence including coding information; the server 12 identifies the target prediction sequence according to the coding rules and coding information matching the code system identifier, and obtains the recognition result; the server 12 returns the recognition result to the terminal 11.

Among them, the terminal 11 can be, but not limited to, various code scanning devices, personal computers, notebook computers, smart phones, tablet computers and portable wearable devices, and the server 12 can be realized by an independent server or a server cluster composed of multiple servers .

In some embodiments, as shown in FIG. 2, a barcode recognition method is provided, and the method is applied to the server 12 in FIG. 1 as an example for illustration, including the following steps:

Step 21, inputting the barcode image into the pre-built target neural network model to obtain a predicted sequence output by the target neural network model; wherein the predicted sequence carries a code system identifier.

Among them, the barcode image refers to the image containing the barcode information; the target neural network model refers to the pre-trained neural network model that can identify the barcode information in the barcode image and output the corresponding results; the prediction sequence refers to the target neural network model. The result of the sequence output after the model recognizes the barcode image; the code system identification refers to the code system corresponding to the barcode information in the barcode image recognized by the target neural network model, that is, the code system predicted by the neural network meets the coding requirements of the barcode image.

Specifically, the server inputs the barcode image into the pre-built target neural network model; the target neural network model extracts image features corresponding to the barcode image according to information such as weight parameters determined in the pre-training process, performs convolution operations on the features, pools Output the final prediction result as the prediction sequence. The code system identification can be determined by the number of activations of corresponding neurons in the target neural network model.

Step 22, decoding the prediction sequence to obtain a target prediction sequence including coding information.

Specifically, the server performs an Argmax operation on the predicted sequence, and obtains the decoded sequence by extracting the Top-N maximum position at position i of each sequence;

The decoding process can be performed by various algorithms such as greedy search, beam search, and prefix beam search, which can be determined according to actual needs.

Merge the repeated elements in the sequence after decoding. If the decoded decoding sequence contains continuous and repeated elements, remove the continuous and repeated elements; for example, according to the position of the Top-1 maximum value in the decoding sequence The sequence is expressed as [1,2,106,2,2,2,7,5,3]. It can be seen that three consecutive 2s in the sequence need to be merged, and only one 2 is retained after the merge process, that is, the merged sequence is [ 1,2,106,2,7,5,3].

Finally, according to the coding information, the placeholders in the sequence are further removed to obtain the target prediction sequence containing coding information; for example, CODO128 specifies No. 106 as a placeholder, then 106 needs to be removed from the above sequence, and the transformed sequence becomes into [1,2,2,7,5,3]; so far, the target prediction sequence including the start symbol, content, check digit, and terminator as coding information is obtained.

Step 23, according to the encoding rules and encoding information matched with the code system identifier, identify the target prediction sequence, and obtain the identification result.

The server determines the encoding rule corresponding to the target prediction sequence according to the coding system identifier; identifies the start symbol, check digit, terminator and content in the target prediction sequence according to the coding information, and analyzes the target prediction sequence; for example, the coding rule is CODE128 The target prediction sequence is [104, 35, 19, 99, 91, 24, 56, 78, 89, 106], the server further determines that the target prediction sequence is the CODE128-B sub-coding rule through the start character 104, then according to the CODE128-B sub-coding rule The coding rules further analyze the target prediction sequence to obtain the analysis result. The analysis result can be output as the recognition result after passing the verification.

The above-mentioned barcode recognition method includes: inputting the barcode image into a pre-built target neural network model to obtain a predicted sequence output by the target neural network model; the predicted sequence carries a code system identification; decoding the predicted sequence to obtain a The target prediction sequence: according to the coding rules and coding information matched with the code system identifier, the target prediction sequence is identified, and the recognition result is obtained. This application recognizes the barcode image through the pre-built target neural network model, and outputs the predicted sequence and the corresponding code system identification; after decoding the predicted sequence, the target predicted sequence with coding information is obtained; finally, the coded information and code system are used Identify the corresponding coding rules, identify the target prediction sequence to obtain the recognition result, and realize the recognition of barcode images of any code system; without manual switching and adjustment according to the code system of the barcode image, the barcode image can be directly recognized, improving Improve the efficiency of barcode recognition.

In some embodiments, step 21, inputting the barcode image into the pre-built target neural network model to obtain the predicted sequence output by the target neural network model includes: if the size of the barcode image is not equal to the preset size, adjusting the barcode image to the preset size After setting the size, input the pre-built target neural network model to obtain the predicted sequence output by the target neural network model.

Specifically, the preset size can be set according to the size of the sample barcode image when the target neural network model is trained. In this embodiment, the input barcode image is consistent with the sample barcode image, so that the target neural network model can predict and recognize the barcode image on a consistent basis, improving the efficiency and accuracy of the target neural network model in recognizing barcode images.

In some embodiments, as shown in FIG. 3, the above step 22 is to decode the predicted sequence to obtain the target predicted sequence containing coding information, including:

Step 31, obtaining position information corresponding to at least one maximum value in the prediction sequence, and generating a decoding sequence according to the position information;

Step 32, after removing duplicate items and preset placeholders in the decoding sequence, a target prediction sequence including coding information is obtained.

Specifically, the preset placeholder is set according to the CTC algorithm used to construct the loss function in the target neural network model, so the preset placeholder needs to be deleted after decoding.

Obtaining the position of each maximum value in the target prediction sequence, and generating a decoding sequence according to the position of the maximum value can be obtained by calculating through the Argmax algorithm. argmax is a function, which is a function for finding parameters (sets) of a function; that is, argmax(f(x)) is the variable point x (or set of x) corresponding to the maximum value of f(x).

The decoding sequence can be obtained through the beam search algorithm; beam search (Beam Search) is a heuristic graph search algorithm, which can affect the algorithm delay and final effect through the preset empirical parameter K (the empirical parameter K can be used according to the actual scene and experience to set). The decoded sequence can also be processed in a naive decoding manner, that is, the first line in the decoded sequence is directly obtained as the decoded decoded sequence, that is, the sequence corresponding to the position of the Top-1 maximum value.

In this embodiment, the predicted sequence output by the model can be gradually adjusted to the target predicted sequence close to the barcode encoding format by generating the decoded sequence by position, obtaining the decoded decoded sequence, and removing duplicates in the decoded decoded sequence, improving the The efficiency of barcode recognition.

In some embodiments, as shown in Figure 4, the above-mentioned step 23, according to the encoding rules and encoding information matched with the code system identification, identifies the target prediction sequence, and obtains the identification result, including:

Step 41: Analyze the target prediction sequence according to the encoding rules and encoding information matched with the code system identifier, and obtain the analysis result;

Step 42, constructing a subsequence of the target prediction sequence according to the parsing result;

Step 43, verifying the subsequence of the target prediction sequence according to the encoding rule, and determining the recognition result according to the verification result.

Specifically, the process of identifying the target prediction sequence mainly includes parsing and verification; if the result of the parsing is in the state of passing the verification, it can be confirmed that the target prediction sequence has been identified; otherwise, if the result of the parsing is in the state of failing the verification, it can be confirmed that The target prediction sequence has not been identified.

The process of parsing needs to determine the specific parsing method through coding rules, which are used to identify different components in the sequence. The encoding rules can also verify the analysis results, and according to the verification results, it can be judged whether the analysis results can be used as credible identification results. For example, when a terminator appears when parsing to the end of the sequence, a new sequence is constructed according to the elements before the terminator as a subsequence of the target prediction sequence, and the subsequence is verified.

In this embodiment, the verification of the target predicted sequence and the analysis result is realized by constructing a corresponding subsequence through the analysis result, which improves the efficiency and accuracy of barcode recognition.

In some embodiments, as shown in FIG. 5, the above step 21, before inputting the barcode image into the pre-built target neural network model to obtain the predicted sequence output by the target neural network model, further includes:

Step 51, acquiring sample barcode images and corresponding sample coding information;

Step 52, inputting the sample barcode image into the initial neural network model to obtain the prediction result of the sample barcode image at a preset height;

Step 53: Train the initial neural network model according to the prediction result and sample encoding information to obtain the target neural network model.

Wherein, the sample barcode image refers to an image that contains clear barcode information and is used for training the model; the sample encoding information refers to encoding information corresponding to the sample barcode image one-to-one. The corresponding sample code information can be obtained from the sample barcode image, and the corresponding sample barcode image can also be obtained according to the sample code information.

Specifically, for example, if you want to construct a sample barcode image of the CODE128 standard and contain the information "C391245678", you need to encode the contained information through the CODE128 encoding rule to obtain the encoded sequence [104,35,19,99,91, 24,56,78,89,106]; the first bit is the start character, and it indicates that the code follows the CODE128-B substandard to encode the subsequent information; the encoded content includes "35,19,99,91,24, 56,78", 35 encodes the "C" in the information, 19 encodes the "3" in the information, "99" indicates that the subsequent information is encoded in accordance with the CODE128-C substandard; "91" encodes the "91" in the information ", "24" encodes the "24" in the information, and so on, until the penultimate digit "89" is the check digit, and the last digit "106" is the stop digit. After all the sample coding information is identified, a sample barcode image corresponding to the sample coding information can be generated according to the coding rules of the CODE128 standard.

If you have obtained sample barcode images and need to obtain their corresponding sample code information, you can use the existing decoding software to directly identify the sample barcode images on the basis of knowing the code rules used by the sample barcode images to obtain the corresponding sample code information .

The height of the output can be fixed by the maximum pooling layer in the initial neural network model; for example, in the case of h-height sample barcode image input, the initial neural network model output height is recorded as h_model. If the value of h_model is not 1, the output can be max-pooled through the maximum pooling layer with height h_model, width 1, step size 1, and padding 0, so that the initial neural network model output is a height of 1 The two-dimensional matrix of is used as the prediction result of the sample barcode image. Similarly, the prediction result of the sample barcode image can also be controlled at a preset height by means of specific parameters such as pooling, convolution, resize (size adjustment function), interpolation, and the like.

In this embodiment, the initial neural network model including the maximum pooling layer keeps the prediction result of the output sample barcode image uniform in size, which improves the training efficiency of the initial neural network model and the prediction accuracy of the target neural network model.

In some embodiments, the above step 51, acquiring the sample barcode image and the corresponding sample coding information includes: randomly generating a plurality of original coding information according to different coding rules; generating the original coded image according to the corresponding coding rules from the original coding information; The original coded image is used as a sample barcode image, and the original coded information is used as corresponding sample coded information.

Specifically, the initial neural network model needs to use the information pair composed of the sample code information and the sample barcode image in the training process, so one of the sample barcode image and the sample code information can be obtained, and the other can be generated through related technologies. information to complete the accumulation of training data; in addition to using the randomly generated original coding information to generate the corresponding sample barcode image, the sample barcode image can also be decoded according to the sample barcode image that has been obtained to obtain the corresponding original coding information as a sample Encoding information, that is, an information pair of sample barcode image-sample coding information is obtained.

In this embodiment, corresponding sample barcode images are generated from a plurality of randomly generated original coding information, and the accumulation of sample barcode images and sample coding information is realized; the accumulation of samples can improve the effect of initial neural network model training.

In some embodiments, the above step 53 is to train the initial neural network model according to the prediction result and sample encoding information to obtain the target neural network model, including: constructing a loss function based on the degree of difference between the prediction result and the sample encoding information; The initial neural network model is trained according to the loss function until the trained initial neural network model meets the preset training conditions, and the trained initial neural network model is used as the target neural network model.

Specifically, the loss function can be constructed using the CTC (Connectionist temporal classification, a temporal classification algorithm) algorithm; the loss function value is obtained by calculating the difference between the prediction results and the sample encoding information through the CTC loss function, and the loss function value is reversed according to the loss function value. Propagate to the initial neural network model, and adjust each weight parameter; until the initial neural network model meets the preset training conditions, such as training for a certain number of times, or the loss function value is lower than a preset threshold, etc.

In this embodiment, by constructing a loss function, the process of model training is continuously carried out towards reducing the degree of difference between prediction results and sample coding information, thereby improving the effect of initial neural network model training.

In some embodiments, before inputting the sample barcode image into the initial neural network model to obtain the prediction result of the sample barcode image at a preset height, it also includes: performing data enhancement processing on the sample barcode image, and adjusting the height of the sample barcode image to the preset height to obtain the processed sample barcode image;

Input the sample barcode image into the initial neural network model to obtain the prediction result of the sample barcode image at the preset height, including: input the processed sample barcode image into the initial neural network model to obtain the prediction result of the sample barcode image at the preset height .

In some embodiments, yet another barcode identification method is provided, including:

1. Obtain the sample barcode image and sample coding information: the coding rules that the sample barcode image and sample coding information can follow are EAN series, CODE code series, UPC code series, and ITF code series, and the sample coding information usually includes the start symbol, content, check digit and terminator.

2. Train the initial neural network model: standardize all sample barcode images to a uniform height h. Perform data enhancement processing on all sample barcode images, including: blurring, adjusting image brightness, cropping images, adding noise, etc. The purpose is to reduce the quality of relatively regular images; obtain the changed sample barcode image. Select a neural network model, and calculate the output height value h_model of the model in the case of h-height image input; if the value of h_model is not 1, then pass the model output through a layer with a height of h_model and a width of 1 and a step size of 1, padding (fill Recharge) is the MaxPooling (maximum pooling) layer of 0.

Perform a 1*1 convolution operation on the prediction results output by the initial neural network model, and the number of output channels C is the total number of codes that may appear in the barcode. For example, for Code128, the output dimension is 106+1, where +1 is required by the placeholder "Blank" designed in the CTC algorithm.

Calculate the error L between the sample encoding information and the prediction result through the CTC algorithm, update the model parameters of the initial neural network model through back propagation, and iterate the process multiple times until the error L converges (the typical value of L convergence is 0-20) , get the trained initial neural network model and use it as the pre-built target neural network model.

3. Use the pre-built target neural network model to identify the barcode image: standardize the height of the barcode image to be recognized to h and send it to the target neural network model M to obtain the predicted sequence output by the target neural network model; the predicted sequence is A two-dimensional matrix with the number of channels C and whose width varies with the width of the input image.

Decode the predicted sequence: perform Argmax operation on the predicted sequence, extract the Top_N maximum value position at each position i, and obtain the sequence T.

Naive decoding method: take out the first row of the sequence T, which is the position of the maximum value of Top 1 in the predicted sequence. BeamSearch decoding method: Select the empirical parameter K (K affects the algorithm delay and final effect; usually weighed according to the actual scene). Taking K as the optimal K sequence for alternative decoding, the sequence T is searched by BeamSearch.

Merge consecutive and repeated signals in T; remove the "Blank" placeholder in the merged sequence. Taking CODO128 as an example, if number 106 is specified as a placeholder during the training phase, 106 needs to be removed in the above example. The transformed sequence becomes [1,2,2,7,5,3]. Repeated characters in the sequence do not need to be merged at this time. The sequence after the three steps of decoding, merging and elimination is called the target prediction sequence. The target prediction sequence includes original coding information such as start character, content, check digit, and terminator.

4. According to the encoding rules and encoding information, identify the target prediction sequence and obtain the recognition result: use the encoding rule to decode the target prediction sequence; ,89,106] as an example, according to the start character 104, it is judged that the code is a CODE128-B sub-standard, and the subsequent value is analyzed through the CODE128-B sub-standard, and when the end character 106 appears at the end, it is backward 1 bit from the terminator, Construct a subsequence, and pass the subsequence through the CODE128 verification method, that is, verify [104, 35, 19, 99, 91, 24, 56, 78, 89]. If it passes the verification, the target prediction sequence is considered valid. Complete barcode recognition.

The above barcode recognition method recognizes the barcode image through the pre-built target neural network model, and outputs the predicted sequence and the corresponding code system identification; after decoding the predicted sequence, the target predicted sequence with coding information is obtained; finally, using the coding information And the coding rules corresponding to the code system logo, identify the target prediction sequence to get the recognition result, and realize the recognition of barcode images of any code system; without manual switching and adjustment according to the code system of the barcode image in advance, the barcode image can be directly processed Recognition, improve the efficiency of barcode recognition.

It should be understood that although the various steps in the flow charts of FIGS. 2-5 are shown sequentially as indicated by the arrows, these steps are not necessarily executed sequentially in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and these steps can be executed in other orders. Moreover, at least some of the steps in Figures 2-5 may include multiple steps or stages, these steps or stages are not necessarily executed at the same moment, but may be executed at different moments, the execution of these steps or stages The sequence is not necessarily performed sequentially, but may be performed alternately or alternately with other steps or at least a part of steps or stages in other steps.

In some embodiments, as shown in Figure 6, a barcode recognition device is provided, including:

The sequence output module 61 is used to input the barcode image into the pre-built target neural network model to obtain the predicted sequence output by the target neural network model; wherein the predicted sequence carries a code system identification;

A sequence decoding module 62, configured to decode the predicted sequence to obtain a target predicted sequence containing coding information;

The sequence identification module 63 is configured to identify the target predicted sequence according to the coding rules and coding information matched with the coding system identifier, and obtain the recognition result.

In some embodiments, the sequence output module 61 is also used to adjust the barcode image to the preset size and then input the pre-built target neural network model to obtain the target neural network model output if the size of the barcode image is not equal to the preset size. Prediction sequence.

In some embodiments, the sequence decoding module 62 is further configured to obtain position information corresponding to at least one maximum value in the predicted sequence, and generate a decoding sequence according to the position information; after removing duplicates and preset placeholders in the decoding sequence, Obtain the target prediction sequence containing the coding information.

In some embodiments, the sequence identification module 63 is further configured to analyze the target prediction sequence according to the coding rules and coding information matched with the code system identifier to obtain the analysis result; according to the analysis result, construct a subsequence of the target prediction sequence; The subsequence of the target prediction sequence is verified according to the coding rules, and the recognition result is determined according to the verification result.

In some embodiments, the barcode recognition device also includes a model training module, which is used to obtain sample barcode images and corresponding sample coding information; input the sample barcode images into the initial neural network model to obtain the prediction results of the sample barcode images at a preset height ; Train the initial neural network model according to the prediction result and sample encoding information to obtain the target neural network model.

In some embodiments, the model training module is also used to randomly generate a plurality of original encoding information according to different encoding rules; generate the original encoding image according to the corresponding encoding rules; use the original encoding image as a sample barcode image, and convert the original The encoding information serves as the corresponding sample encoding information.

In some embodiments, the model training module is also used to construct a loss function based on the difference between the prediction results and the sample encoding information; train the initial neural network model according to the loss function until the trained initial neural network model meets the preset When training conditions, the trained initial neural network model is used as the target neural network model.

In some embodiments, the model training module is also used to perform data enhancement processing on the sample barcode image, and adjust the height of the sample barcode image to a preset height to obtain the processed sample barcode image; the processed sample barcode image Input the initial neural network model to obtain the prediction result of the sample barcode image at the preset height.

In some embodiments, the symbolic identity is determined by the number of activations of corresponding neurons in the target neural network model.

For specific limitations on the barcode recognition device, please refer to the above-mentioned limitations on the barcode recognition method, which will not be repeated here. Each module in the above-mentioned barcode recognition device can be fully or partially realized by software, hardware and a combination thereof. The above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, and can also be stored in the memory of the computer device in the form of software, so that the processor can invoke and execute the corresponding operations of the above-mentioned modules.

In some embodiments, a computer device is provided. The computer device may be a server, and its internal structure may be as shown in FIG. 7 . The computer device includes a processor, memory and a network interface connected by a system bus. Wherein, the processor of the computer device is used to provide calculation and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, computer readable instructions and a database. The internal memory provides an environment for the execution of the operating system and computer readable instructions in the non-volatile storage medium. The database of the computer device is used to store barcode identification data. The network interface of the computer device is used to communicate with an external terminal via a network connection. When the computer-readable instructions are executed by the processor, a barcode recognition method is realized.

Those skilled in the art can understand that the structure shown in Figure 7 is only a block diagram of a part of the structure related to the solution of this application, and does not constitute a limitation to the computer equipment on which the solution of this application is applied. The specific computer equipment can be More or fewer components than shown in the figures may be included, or some components may be combined, or have a different arrangement of components.

In some embodiments, a computer device is provided, including a memory and a processor, where computer-readable instructions are stored in the memory, and the processor implements the following steps when executing the computer-readable instructions:

In some embodiments, the following steps are also implemented when the processor executes the computer-readable instructions: if the size of the barcode image is not equal to the preset size, then adjust the barcode image to the preset size and input the pre-built target neural network model to obtain the target The sequence of predictions output by the neural network model.

In some embodiments, when the processor executes the computer-readable instructions, the following steps are also implemented: obtaining the position information corresponding to at least one maximum value in the prediction sequence, generating a decoding sequence according to the position information; eliminating duplicates and presets in the decoding sequence After the placeholder, the target prediction sequence containing the coding information is obtained.

In some embodiments, when the processor executes the computer-readable instructions, the following steps are also implemented: according to the encoding rules and encoding information matched with the code system identifier, the target prediction sequence is analyzed to obtain the analysis result; according to the analysis result, the target prediction sequence is constructed The subsequence of the sequence; the subsequence of the target prediction sequence is verified according to the coding rules, and the recognition result is determined according to the verification result.

In some embodiments, when the processor executes the computer-readable instructions, the following steps are also implemented: acquiring the sample barcode image and corresponding sample coding information; inputting the sample barcode image into the initial neural network model to obtain the sample barcode image at a preset height Prediction results: train the initial neural network model according to the prediction results and sample encoding information to obtain the target neural network model.

In some embodiments, when the processor executes the computer-readable instructions, the following steps are also implemented: randomly generating a plurality of original encoding information according to different encoding rules; generating an original encoded image from the original encoding information according to corresponding encoding rules; using the original encoded image as The sample barcode image uses the original coded information as the corresponding sample coded information.

In some embodiments, when the processor executes the computer-readable instructions, the following steps are also implemented: constructing a loss function based on the prediction results and the degree of difference between the coded information of the samples; training the initial neural network model according to the loss function until the initial neural network model after training When the neural network model meets the preset training conditions, the trained initial neural network model is used as the target neural network model.

In some embodiments, when the processor executes the computer-readable instructions, the following steps are also implemented: performing data enhancement processing on the sample barcode image, and adjusting the height of the sample barcode image to a preset height to obtain the processed sample barcode image; The processed sample barcode image is input to the initial neural network model to obtain the prediction result of the sample barcode image at a preset height.

In some embodiments, when the processor executes the computer-readable instructions, the following steps are further implemented: determining the code system identification by the number of times corresponding neurons in the target neural network model are activated.

In some embodiments, a computer-readable storage medium is provided, on which computer-readable instructions are stored, and when the computer-readable instructions are executed by a processor, the following steps are implemented:

In some embodiments, when the computer readable instructions are executed by the processor, the following steps are also implemented: if the size of the barcode image is not equal to the preset size, then adjust the barcode image to the preset size and input the pre-built target neural network model to obtain The sequence of predictions output by the target neural network model.

In some embodiments, when the computer-readable instructions are executed by the processor, the following steps are further implemented: obtaining the position information corresponding to at least one maximum value in the prediction sequence, generating a decoding sequence according to the position information; eliminating duplicates and predictions in the decoding sequence After the placeholder is set, the target prediction sequence containing the coding information is obtained.

In some embodiments, when the computer-readable instructions are executed by the processor, the following steps are further implemented: analyzing the target prediction sequence according to the encoding rules and encoding information matched with the code system identifier to obtain the analysis result; according to the analysis result, constructing the target The subsequence of the predicted sequence; the subsequence of the target predicted sequence is verified according to the coding rules, and the recognition result is determined according to the verification result.

In some embodiments, when the computer readable instructions are executed by the processor, the following steps are also implemented: obtaining the sample barcode image and corresponding sample coding information; inputting the sample barcode image into the initial neural network model to obtain the sample barcode image at a preset height The prediction results; according to the prediction results and sample coding information, the initial neural network model is trained to obtain the target neural network model.

In some embodiments, when the computer-readable instructions are executed by the processor, the following steps are further implemented: randomly generating a plurality of original encoding information according to different encoding rules; generating the original encoding image from the original encoding information according to the corresponding encoding rule; converting the original encoding image As a sample barcode image, the original coding information is used as corresponding sample coding information.

In some embodiments, when the computer-readable instructions are executed by the processor, the following steps are also implemented: constructing a loss function based on the prediction result and the degree of difference between the coded information of the samples; training the initial neural network model according to the loss function until the trained When the initial neural network model meets the preset training conditions, the trained initial neural network model is used as the target neural network model.

In some embodiments, when the computer readable instructions are executed by the processor, the following steps are further implemented: performing data enhancement processing on the sample barcode image, and adjusting the height of the sample barcode image to a preset height to obtain the processed sample barcode image; The processed sample barcode image is input into the initial neural network model to obtain the prediction result of the sample barcode image at a preset height.

In some embodiments, when the computer-readable instructions are executed by the processor, the following step is further implemented: determining the code system identification by the number of activation times of corresponding neurons in the target neural network model.

Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above-mentioned embodiments can be implemented by instructing related hardware through computer-readable instructions. The above-mentioned computer-readable instructions can be stored in a non-volatile computer-readable When being read from the storage medium, the computer-readable instructions may include the processes of the embodiments of the above-mentioned methods when executed. Wherein, any references to memory, storage, database or other media used in the various embodiments provided in the present application may include at least one of non-volatile memory and volatile memory. Non-volatile memory may include read-only memory (Read-Only Memory, ROM), magnetic tape, floppy disk, flash memory or optical memory, etc. Volatile memory can include Random Access Memory (RAM) or external cache memory. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM).

The technical features of the above embodiments can be combined arbitrarily. To make the description concise, all possible combinations of the technical features in the above embodiments are not described. However, as long as there is no contradiction in the combination of these technical features, they should be It is considered to be within the range described in this specification.

The above embodiments only express several implementation modes of the present application, and the description thereof is relatively specific and detailed, but it should not be construed as limiting the scope of the patent for the invention. It should be noted that those skilled in the art can make several modifications and improvements without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the scope of protection of the patent application should be based on the appended claims.

Claims

A barcode recognition method is characterized in that, comprising:

Inputting the barcode image into the pre-built target neural network model to obtain a predicted sequence output by the target neural network model; wherein, the predicted sequence carries a code system identification;

Decoding the predicted sequence to obtain a target predicted sequence including coding information;

The target prediction sequence is identified according to the encoding rule matched with the coding system identifier and the encoding information, and an identification result is obtained.
The method according to claim 1, wherein the input of the barcode image into the pre-built target neural network model to obtain the predicted sequence output by the target neural network model comprises:

If the size of the barcode image is not equal to the preset size, the barcode image is adjusted to the preset size and then input into a pre-built target neural network model to obtain a prediction sequence output by the target neural network model.
The method according to claim 1, wherein decoding the predicted sequence to obtain a target predicted sequence containing coding information includes:

Obtaining position information corresponding to at least one maximum value in the predicted sequence, and generating a decoding sequence according to the position information;

After eliminating repeated items and preset placeholders in the decoding sequence, a target prediction sequence including coding information is obtained.
The method according to claim 1, wherein the identifying the target prediction sequence according to the encoding rule matched with the coding system identifier and the encoding information, and obtaining an identification result includes:

Analyzing the target prediction sequence according to the encoding rule matched with the coding system identifier and the encoding information, to obtain an analysis result;

Constructing a subsequence of the target prediction sequence according to the parsing result;

Checking the subsequence of the target prediction sequence according to the encoding rule, and determining the recognition result according to the checking result.
The method according to claim 1, wherein the barcode image is input into the pre-built target neural network model, and before obtaining the predicted sequence output by the target neural network model, it also includes:

Obtain the sample barcode image and the corresponding sample coding information;

Inputting the sample barcode image into an initial neural network model to obtain a prediction result of the sample barcode image at a preset height;

The initial neural network model is trained according to the prediction result and the sample encoding information to obtain the target neural network model.
The method according to claim 5, wherein said acquiring sample barcode images and corresponding sample coding information comprises:

Randomly generate multiple original coding information according to different coding rules;

Generating an original encoded image from the original encoded information according to a corresponding encoding rule;

The original coded image is used as a sample barcode image, and the original coded information is used as corresponding sample coded information.
The method according to claim 5, wherein the training of the initial neural network model according to the prediction result and the sample encoding information to obtain the target neural network model includes:

Constructing a loss function based on the prediction results and the degree of difference between the sample coding information;

The initial neural network model is trained according to the loss function until the trained initial neural network model satisfies a preset training condition, and the trained initial neural network model is used as the target neural network model.
The method according to claim 5, wherein, before said inputting the sample barcode image into the initial neural network model to obtain the prediction result of the sample barcode image at a preset height, further comprising:

performing data enhancement processing on the sample barcode image, and adjusting the height of the sample barcode image to the preset height to obtain the processed sample barcode image;

Said inputting the sample barcode image into the initial neural network model to obtain the prediction result of the sample barcode image at a preset height, including:

Inputting the processed sample barcode image into an initial neural network model to obtain a prediction result of the sample barcode image at a preset height.
The method according to claim 1, wherein the coding system identification is determined by the number of times corresponding neurons in the target neural network model are activated.
A barcode recognition device is characterized in that it comprises:

The sequence output module is used to input the barcode image into the pre-built target neural network model to obtain the predicted sequence output by the target neural network model; wherein the predicted sequence carries a code system identification;

A sequence decoding module, configured to decode the predicted sequence to obtain a target predicted sequence containing coding information;

A sequence identification module, configured to identify the target predicted sequence according to the encoding rule matched with the coding system identifier and the encoding information, and obtain an identification result.
The device according to claim 10, wherein the sequence output module is further configured to adjust the barcode image to the preset size and then input the preset A target neural network model is constructed to obtain a predicted sequence output by the target neural network model.
The device according to claim 10, wherein the sequence decoding module is further configured to obtain position information corresponding to at least one maximum value in the predicted sequence, and generate a decoding sequence according to the position information; After decoding the repeated items and preset placeholders in the sequence, the target prediction sequence including the coding information is obtained.
The device according to claim 10, wherein the sequence identification module is further configured to analyze the target prediction sequence according to the encoding rule matched with the code system identifier and the encoding information, to obtain the analyzed Result; according to the parsing result, construct the subsequence of the target prediction sequence; verify the subsequence of the target prediction sequence according to the encoding rule, and determine the recognition result according to the verification result.
The device according to claim 10, further comprising a model training module for obtaining sample barcode images and corresponding sample encoding information; inputting the sample barcode images into an initial neural network model to obtain the A prediction result of the sample barcode image at a preset height; training the initial neural network model according to the prediction result and the sample encoding information to obtain the target neural network model.
The device according to claim 14, wherein the model training module is further configured to randomly generate a plurality of original encoding information according to different encoding rules; generate an original encoded image according to the corresponding encoding rules from the original encoding information; The original coded image is used as a sample barcode image, and the original coded information is used as corresponding sample coded information.
The device according to claim 14, wherein the model training module is further configured to construct a loss function based on the prediction result and the degree of difference between the sample coding information; The initial neural network model, until the trained initial neural network model meets the preset training conditions, the trained initial neural network model is used as the target neural network model.
The device according to claim 10, wherein the code system identification is determined by the number of times corresponding neurons in the target neural network model are activated.
A computer device, comprising a memory and a processor, the memory stores computer-readable instructions, wherein the processor implements the method described in any one of claims 1 to 9 when executing the computer-readable instructions method steps.
A computer-readable storage medium, on which computer-readable instructions are stored, wherein, when the computer-readable instructions are executed by a processor, the steps of the method described in any one of claims 1 to 9 are implemented.
A computer program product, comprising computer-readable instructions, wherein the steps of the method according to any one of claims 1 to 9 are implemented when the computer-readable instructions are executed by a processor.