A box single image recognition method, device and medium

Abstract

The present application relates to the field of image recognition, and more particularly to a box single image recognition method, device and medium, the method comprising: identifying a plurality of image slices from the box single image; for each image slice, determining reference slices of the image slice; the reference slices satisfy the adjacent position relationship with the image slice in the box single image; determining the entity label of the image slice according to the text information in each reference slice, the text information in the image slice and the position relationship between each reference slice and the image slice; the entity label is used to represent the semantic object to which the text information belongs; obtaining the recognition result of the box single image according to the entity label of each image slice. Through the method, the problem that the existing box single image recognition method does not consider the relationship between various entities in the box single image and has the probability of error in the identification of the audit can be solved.

Description

Technical Field

[0001] This invention relates to the field of image recognition, and more particularly to a method, apparatus and medium for recognizing packing list images. Background Technology

[0002] A packing list is a supplementary document to the invoice. By filling in physical information such as the origin, destination, and quantity of the packed products on the packing list, it clarifies the specific details of the packed products. This makes it easier for the seller to understand and grasp the quantity of the packed products, and also makes it easier for the buyer to check and verify the packed products.

[0003] Packing lists contain important information as well as some irrelevant information. By reviewing images of the packing lists, the important information in the images can be identified, facilitating subsequent processing. Traditionally, packing list image review is conducted by experienced staff, which is time-consuming and labor-intensive. Therefore, existing packing list image review methods mainly employ two approaches: (1) summarizing common packing list content based on regular expressions and reviewing the images by matching these expressions; (2) using deep learning models to review the images. However, the first approach struggles to summarize common content for different types of packing lists using regular expressions, while the second approach requires training with a large number of packing list images. Furthermore, neither approach considers the relationships between various entities in the packing list images, potentially leading to errors in the review process. Therefore, a packing list image recognition method is urgently needed to review the images based on the relationships between various entities in the images, thereby improving the accuracy of the review. Summary of the Invention

[0004] This invention provides a packing list image recognition method, device, and medium to solve the problem that existing packing list image recognition methods do not take into account the relationships between various entities in the packing list image, and therefore may result in errors during the review and recognition process.

[0005] In a first aspect, embodiments of the present invention provide a packing list image recognition method, comprising:

[0006] Multiple image slices were identified from the packing list image;

[0007] For each image slice, a reference slice is determined; the reference slices in the packing list image satisfy an adjacent position relationship with the image slice.

[0008] Based on the text information in each reference slice, the text information in the image slice, and the positional relationship between each reference slice and the image slice, the entity label of the image slice is determined; the entity label is used to characterize the semantic object to which the text information belongs;

[0009] The recognition result of the packing list image is obtained based on the entity labels of each image slice.

[0010] The above method first obtains multiple image slices from the packing list image, then determines reference slices adjacent to each image slice, and then determines the entity label corresponding to each image slice based on the text information of each image slice, the text information of the reference slices, and the positional relationship between the image slice and the reference slices. This yields the semantic object to which the text information of the image slice belongs. Finally, the recognition result corresponding to the packing list image can be obtained based on the entity labels of the multiple image slices in the packing list image. By determining the entity label corresponding to each image slice based on both the image slice and its adjacent reference slices, the relationship between each image slice and its surrounding reference slices is considered, thereby improving the accuracy of the review and recognition of each image slice and the overall accuracy of the packing list image recognition result.

[0011] Optionally, each reference slice has an adjacency relationship with the image slice in different directions, and each reference slice is the image slice that is closest to the image slice in the corresponding direction.

[0012] By using the above method, other image slices that are closest to the image slice in different directions are selected as reference slices, so that the selected reference slices have a high correlation with the image slices.

[0013] Optionally, the entity label of the image slice is determined based on the text information in each reference slice, the text information in the image slice, and the positional relationship between each reference slice and the image slice, including:

[0014] Determine the word sequence corresponding to the text information in each reference slice and the word sequence corresponding to the text information in the image slice;

[0015] Based on the positional relationship between each reference slice and the image slice, the word sequence corresponding to each reference slice and the word sequence corresponding to the image slice are used to construct the sentence sequence of the image slice;

[0016] The sentence sequence is used to obtain entity labels for the image slices through a classification and recognition model.

[0017] Using the above method, based on the positional relationship between the reference slice and the image slice, the word sequence of the reference slice and the word sequence of the image slice are combined to form a sentence sequence, thereby enabling the classification and recognition model to correctly obtain the entity labels of the image slice.

[0018] Optionally, the classification and recognition model is the FastText model and adopts the N-gram word segmentation and combination method.

[0019] Optionally, the classification and recognition model classifies target entity labels; the target entity labels are determined from the entity labels in each packing list image.

[0020] Optionally, multiple image slices are identified from the packing list image, including:

[0021] Text recognition technology was used to identify the text information and location information of each image slice from the packing list image.

[0022] Secondly, embodiments of the present invention provide a packing list image recognition device, comprising:

[0023] The transceiver unit is used to identify multiple image slices from the packing list image;

[0024] The processing unit is configured to, for each image slice, determine each reference slice of the image slice; the reference slices satisfy the adjacent position relationship with the image slice in the packing list image; determine the entity label of the image slice based on the text information in each reference slice, the text information in the image slice, and the positional relationship between each reference slice and the image slice; the entity label is used to characterize the semantic object to which the text information belongs; and obtain the recognition result of the packing list image based on the entity labels of each image slice.

[0025] Thirdly, this application also provides an apparatus. This apparatus can perform the above-described method design. The apparatus may be a chip or circuit capable of performing the functions corresponding to the above-described method, or a device including the chip or circuit.

[0026] In one possible implementation, the device includes: a memory for storing computer-executable program code; and a processor coupled to the memory. The program code stored in the memory includes instructions that, when executed by the processor, cause the device or a device equipped with the device to perform any of the methods described above.

[0027] The device may also include a communication interface, which may be a transceiver, or, if the device is a chip or circuit, the communication interface may be the chip's input / output interface, such as input / output pins.

[0028] In one possible design, the device includes corresponding functional units, each used to implement the steps in the above method. The functions can be implemented in hardware or by hardware executing corresponding software. The hardware or software includes one or more units corresponding to the functions described above.

[0029] Fourthly, this application provides a computer-readable storage medium storing a computer program that, when run on a device, executes the method described in any of the above possible designs.

[0030] Fifthly, this application provides a computer program product comprising: computer program code, which, when run on a computer, causes the computer to perform the method described in any one of the first aspects.

[0031] Furthermore, the technical effects of any of the implementation methods in the third to fifth aspects can be found in the technical effects of different implementation methods in the first aspect, and will not be repeated here. Attached Figure Description

[0032] Figure 1 A flowchart of a packing list image recognition method provided in an embodiment of the present invention;

[0033] Figure 2 This is a schematic diagram of a portion of an English packing list image provided in an embodiment of the present invention;

[0034] Figure 3 A flowchart illustrating a packing list image recognition method based on OCR and FastText models provided in this embodiment of the invention;

[0035] Figure 4 A schematic diagram of a packing list image to be identified, provided as an embodiment of the present invention;

[0036] Figure 5 A communication device 500 is provided for an embodiment of the present invention;

[0037] Figure 6 A communication device 600 is provided for an embodiment of the present invention. Detailed Implementation

[0038] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of this invention, and not all embodiments. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0039] The application scenarios described in the embodiments of this invention are for the purpose of more clearly illustrating the technical solutions of the embodiments of this invention, and do not constitute a limitation on the technical solutions provided by the embodiments of this invention. Those skilled in the art will understand that with the emergence of new application scenarios, the technical solutions provided by the embodiments of this invention are also applicable to similar technical problems. In the description of this invention, unless otherwise stated, "multiple" means two or more.

[0040] Currently, it is necessary to identify important image slices within packing list images. Existing methods for reviewing packing list images mainly rely on regular expression matching or training deep learning models for image review. However, both of these methods review each image slice individually, without considering the relationships between adjacent slices. This can lead to errors in the identification results, potentially misclassifying unimportant slices as important ones.

[0041] Based on this, this application provides a packing list image recognition method to solve the problem that existing packing list image recognition methods do not take into account the relationships between various entities in the packing list image, and there is a probability of errors in the review and recognition.

[0042] For example, the packing list image can be a Chinese packing list image, an English packing list image, or a packing list image of any other type of text; this application does not impose any specific limitations. This application uses an English packing list image as an example for specific illustration.

[0043] like Figure 1 As shown, the specific flow of the packing list image recognition method provided in this application is as follows:

[0044] Step 100: Identify multiple image slices from the packing list image.

[0045] Specifically, text recognition technology can be used to obtain multiple image slices from a packing list image, and the text recognition technology can also identify the text information and position information of each image slice. This application does not limit the specific text recognition technology used; however, it uses Optical Character Recognition (OCR) as an example for specific illustration.

[0046] Specifically, each image slice includes text information corresponding to that image slice, and the location information of each image slice is its relative position within the packing list image. The location information can be represented in coordinate form, or in other forms; this application does not limit its representation.

[0047] For example, such as Figure 2 The image shown is a partial illustration of an English packing list. Using OCR, five image slices can be obtained from this packing list image. Each image slice includes the following text information: Packing list, TOBEIJING,CHINA, TO SHANGHAI,CHINA, FROM NANJING,CHINA, and DATA:25 April,2023. Setting the bottom left corner as the origin, and using coordinates to represent the position information of each image slice, the position information for TO BEIJING,CHINA is (1,1), FROM NANJING,CHINA is (2,1), TO SHANGHAI,CHINA is (2,2), DATA:25 April,2023 is (3,1), and Packing list is (2,3).

[0048] Step 110: For each image slice, determine the reference slices for that image slice.

[0049] For example, after obtaining each image slice in step 100, corresponding reference slices can be determined for each image slice. The reference slices are adjacent to the image slices in the packing list image.

[0050] Specifically, each reference slice has an adjacency relationship with the image slice in different directions, and each reference slice is the image slice closest to the image slice in the corresponding direction. The image slice closest to the image slice in the corresponding direction can be determined by calculating the positional information of the image slice relative to other image slices. These different directions can include the four directions of east, south, west, and north relative to the image slice, or the four directions of southeast, southwest, northeast, and northwest relative to the image slice.

[0051] For example, such as Figure 2As shown, for the image slice TO SHANGHAI, CHINA, its reference slices are Packing list, TO BEIJING, CHINA, FROM NANJING, CHINA, and DATA: 25 April, 2023. For the image slice FROM NANJING, CHINA, its reference slices are TO BEIJING, CHINA, TO SHANGHAI, CHINA, and DATA: 25 April, 2023. While Packing list is also located north of FROM NANJING, CHINA, it is not the closest image slice to FROM NANJING, CHINA in the north direction relative to TO SHANGHAI, CHINA.

[0052] Step 120: Determine the entity label of the image slice based on the text information in each reference slice, the text information in the image slice, and the positional relationship between each reference slice and the image slice.

[0053] Specifically, entity tags are used to represent the semantic object to which the text information belongs;

[0054] For example, after each reference slice of the image slice is determined in step 110, the entity label of the image slice can be determined based on the text information in each reference slice, the text information in the image slice, and the positional relationship between each reference slice and the image slice.

[0055] Specifically, first, the word sequences corresponding to the text information of each reference slice are determined based on the text information of each reference slice, and the word sequences corresponding to the text information of the image slice are determined based on the text information of the image slice. Then, based on the positional relationship between each reference slice and the image slice, the word sequences corresponding to each reference slice and the word sequences corresponding to the image slice are concatenated to form the sentence sequence of the image slice; the concatenation must be performed in a pre-set preset order. Finally, the sentence sequence is input into the classification and recognition model to obtain the entity labels of the image slice.

[0056] This application does not limit the specific classification and recognition model. It uses the FastText model as an example for detailed explanation, employing the N-gram word segmentation and combination method of the FastText model for recognition. Furthermore, due to certain limitations on the input of the FastText model, the sentence sequence needs to be converted into a sentence vector sequence before being input into the FastText model for classification. This application does not limit the method used to convert the sentence sequence into a sentence vector sequence.

[0057] Specifically, the classification and recognition model classifies target entity labels, which are determined from the entity labels in each packing list image based on empirical values. The specific content of the target entity labels is shown in Table 1 below. The target entity label "other" indicates that the image slice is non-important content (i.e., other content), while all other target entity labels indicate that the image slice is important content.

[0058] Table 1

[0059]

[0060]

[0061]

[0062] For example, the classification and recognition model needs to be trained before the sentence sequence is input into it.

[0063] Specifically, the stitching order is preset. For example, when there are four corresponding reference slices in the four directions of east, south, west, and north of the image slice, the stitching can be performed based on the usual reading order, in the following order: reference slice in the north direction, reference slice in the west direction, image slice, reference slice in the east direction, and reference slice in the south direction. Alternatively, the stitching order can be any other order, which is not limited in this application. For example, when there are three corresponding reference slices in the three directions of east, south, and north of the image slice, the stitching can be performed in the following order: image slice first, followed by reference slices in the three directions.

[0064] Specifically, after setting the splicing order, staff with years of experience in order review need to label the spliced ​​sentence sequence. Based on the target entity labels in Table 1, the entity labels corresponding to the spliced ​​sentence sequence are determined. Then, the labeled sentence sequence is used as the training set to train the classification and recognition model.

[0065] After training the classification and recognition model, the image slices to be recognized are spliced ​​together in a pre-set order to obtain a sentence sequence, which is then input into the classification and recognition model to obtain the entity label corresponding to the image slice. In this way, the entity labels corresponding to multiple image slices in the packing list image can be obtained.

[0066] Step 130: Obtain the recognition result of the packing list image based on the entity labels of each image slice.

[0067] For example, after determining the entity labels corresponding to multiple image slices of the packing list image in step 120, the recognition result of the packing list image can be determined based on these entity labels. Specifically, the recognition result of the packing list image is to identify the important image slices in the packing list image.

[0068] For example, such as Figure 2 As shown, settings Figure 2 In the image, the entity label for FROM NANJING,CHINA is FromPlace, the entity label for TO BEIJING,CHINA is Via Place, the entity label for TO SHANGHAI,CHINA is ToPlace, the entity label for DATA:25 April,2023 is Packing List Date, and the entity label for Packing list is Other. Therefore, except for Packing list, which is a non-important image slice, all others are important image slices. Thus, these four important image slices are filled into the corresponding parts of the target entity labels in Table 1 as the recognition result of the packing list image. Furthermore, the presentation format of the packing list image recognition result can also include other formats, which are not limited in this application. For example, important image slices and entity labels in the packing list image can be presented in image form.

[0069] Using the above method, the corresponding reference slice is determined based on the positional information of each image slice in the packing list image. Then, based on the positional relationship between the image slice and its corresponding reference slice, the text information of the image slice and the reference slice are concatenated to obtain the sentence sequence of the image slice. Finally, based on the sentence sequence of the image slice and a pre-trained classification model, the entity label corresponding to the image slice is obtained, and the recognition result of the packing list image is obtained based on the entity labels corresponding to each image slice in the packing list image. This ensures that each image slice in the packing list image to be recognized takes into account its adjacent reference slices and the relationships between various entities in the packing list image, thereby improving the success rate of packing list image verification and recognition.

[0070] For example, Figure 3 This is a flowchart illustrating a packing list image recognition method proposed in this application, which uses OCR technology and the FastText classification model. Figure 4 The specific process for identifying the packing list image is as follows:

[0071] Step 301: Obtain the text information and location information of the image slices through OCR.

[0072] Specifically, in identification Figure 4 When the packing list image is shown, the text information of the image slices obtained by OCR are as follows: No.145, Packing list, FROM NANJING, CHINA, TO SHANGHAI, CHINA, Quantity: 10000, and DATA: 11 May, 2023.

[0073] With the lower left corner as the origin, the location information of image slice No.145 is (1,1), the location information of image slice Packing list is (1,2), the location information of image slice FROM NANJING,CHINA is (2,1), the location information of image slice TO SHANGHAI,CHINA is (2,2), the location information of image slice Quantity:10000 is (3,1), and the location information of image slice DATA:11May,2023 is (3,2).

[0074] Step 302: Determine the reference slice for each image slice.

[0075] Specifically, after obtaining the text and location information of each image slice in step 301, the reference slice for each image slice can be determined. Specifically, the image slice closest to each image slice in the four cardinal directions (north, south, east, and west) is used as the corresponding reference slice. Therefore, for image slice No. 145, its reference slices are Packing list and FROM NANJING, CHINA; for image slice Packing list, its reference slices are No. 145 and TOSHANGHAI, CHINA; and so on, the reference slices for each of these six image slices can be obtained.

[0076] Step 303: Assemble each image slice and its corresponding reference slice in a preset order to obtain a sentence sequence.

[0077] Specifically, referring to step 120, the preset order is set as image slices first, followed by reference slices in the east, west, south, and north directions. For image slice No. 145, its corresponding sentence sequence is No. 145 FROM NANJING, CHINA Packing list; for image slice Packing list, its corresponding sentence sequence is Packing list TOSHANGHAI, CHINA No. 145. And so on, the sentence sequences corresponding to the remaining four image slices can be determined.

[0078] Step 304: Input the sentence sequence of the image slice into the FastText model to determine the corresponding entity label of the image slice.

[0079] Specifically, the FastText model is pre-trained as per step 120. Then, the sentence sequences of these 6 image slices are converted into corresponding sentence vector sequences and input into the FastText model to obtain the corresponding entity labels for these 6 image slices.

[0080] Specifically, based on Table 1, the entity label for image slice No. 145 is Lot Number, the entity label for image slice Packinglist is Other, the entity label for image slice FROM NANJING,CHINA is From Place, the entity label for image slice TO SHANGHAI,CHINA is To Place, the entity label for image slice Quantity:10000 is Container Counts, and the entity label for image slice DATA:11May,2023 is Packing List Date.

[0081] Step 305: Obtain the recognition result of the packing list image based on the entity labels of the image slices.

[0082] Specifically, after determining the entity labels of the six image slices in step 304, the recognition result of the packing list image can be obtained.

[0083] Specifically, since the entity label of the packing list is "Other", the recognition result of the packing list image does not include this image slice. The recognition result of the packing list image is shown in Table 2 below.

[0084] Table 2

[0085]

[0086] The acquisition, storage, use, and processing of data in this application all comply with the relevant provisions of national laws and regulations.

[0087] The division of units in the embodiments of this invention is illustrative and represents only one logical functional division. In actual implementation, other division methods may be used. Furthermore, the functional units in the various embodiments of this invention can be integrated into a single processor, exist as separate physical units, or be integrated into a single unit. The integrated units described above can be implemented in hardware or as software functional units.

[0088] This invention also provides a communication device 500, see [link to documentation]. Figure 5 As shown, it includes: a processing module 510 and a transceiver module 520.

[0089] The transceiver module 520 may include a receiving unit and a transmitting unit. The processing module 510 is used to control and manage the operation of the communication device 500. The transceiver module 520 is used to support communication between the communication device 500 and other devices. Optionally, the communication device 500 may also include a storage unit for storing the program code and data of the communication device 500.

[0090] Optionally, each module in the communication device 500 can be implemented by software.

[0091] Optionally, the processing module 510 may be a processor or controller, such as a general-purpose central processing unit (CPU), a general-purpose processor, a digital signal processing unit (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. It can implement or execute various exemplary logic blocks, modules, and circuits described in conjunction with the embodiments of this application. The processor may also be a combination that implements computing functions, such as a combination of one or more microprocessors, a combination of a DSP and a microprocessor, etc. The transceiver module 520 may be a communication interface, a transceiver, or a transceiver circuit, etc., wherein the communication interface is a general term, and in a specific implementation, the communication interface may include multiple interfaces, and the storage unit may be a memory.

[0092] Processing module 510 calls transceiver module 520 to execute:

[0093] The transceiver module 520 is used to identify multiple image slices from the packing list image;

[0094] The processing module 510 is configured to determine, for each image slice, reference slices of the image slice; the reference slices satisfy the adjacent position relationship with the image slice in the packing list image; determine the entity label of the image slice based on the text information in each reference slice, the text information in the image slice, and the positional relationship between each reference slice and the image slice; the entity label is used to characterize the semantic object to which the text information belongs; and obtain the recognition result of the packing list image based on the entity label of each image slice.

[0095] This invention also provides another communication device 600, see [link to previous document]. Figure 6 As shown, it includes:

[0096] Communication interface 601, memory 602 and processor 603;

[0097] The communication device 600 communicates with other devices through the communication interface 601, such as sending and receiving messages; the memory 602 is used to store program instructions; and the processor 603 is used to call the program instructions stored in the memory 602 and execute them according to the obtained program.

[0098] Processor 603 executes program instructions stored in communication interface 601 and memory 602:

[0099] Multiple image slices are identified from the packing list image; for each image slice, reference slices are determined; the reference slices in the packing list image satisfy an adjacent position relationship with the image slice; based on the text information in each reference slice, the text information in the image slice, and the positional relationship between each reference slice and the image slice, entity labels of the image slices are determined; the entity labels are used to characterize the semantic object to which the text information belongs; based on the entity labels of each image slice, the recognition result of the packing list image is obtained.

[0100] In this embodiment of the invention, the specific connection medium between the communication interface 601, the memory 602 and the processor 603 is not limited, such as a bus. A bus can be divided into an address bus, a data bus, a control bus, etc.

[0101] In this embodiment of the invention, the processor may be a general-purpose processor, a digital signal processor, an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components, capable of implementing or executing the methods, steps, and logic block diagrams disclosed in this embodiment of the invention. The general-purpose processor may be a microprocessor or any conventional processor. The steps of the methods disclosed in this embodiment of the invention can be directly manifested as being executed by a hardware processor, or executed by a combination of hardware and software modules within the processor.

[0102] In embodiments of the present invention, the memory can be non-volatile memory, such as a hard disk drive (HDD) or a solid-state drive (SSD), or it can be volatile memory, such as random-access memory (RAM). The memory can also be any other medium capable of carrying or storing desired program code having an instruction or data structure form and accessible by a computer, but is not limited thereto. The memory in embodiments of the present invention can also be a circuit or any other device capable of implementing a storage function for storing program instructions and / or data.

[0103] This invention also provides a computer-readable storage medium including program code. When the program code is run on a computer, the program code is used to cause the computer to perform the steps of the method provided in the above embodiments of this invention.

[0104] This invention also provides a computer program product comprising: computer program code, which, when run on a computer, causes the computer to perform the steps of the method provided in this invention.

[0105] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0106] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0107] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0108] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0109] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention.

[0110] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A method for recognizing packing list images, characterized in that, The method includes: Multiple image slices were identified from the packing list image; For each image slice, reference slices are determined; the reference slices in the packing list image satisfy the adjacent position relationship with the image slice; each reference slice has an adjacent relationship with the image slice in different directions and each reference slice is the image slice closest to the image slice in the corresponding direction; Based on the text information in each reference slice, the text information in the image slice, and the positional relationship between each reference slice and the image slice, the entity label of the image slice is determined; the entity label is used to characterize the semantic object to which the text information belongs. Based on the entity labels of each image slice, the identification result of the packing list image is obtained, and the identification result is used to characterize whether the image slice is an important image slice.

2. The method as described in claim 1, characterized in that, Based on the text information in each reference slice, the text information in the image slice, and the positional relationship between each reference slice and the image slice, the entity label of the image slice is determined, including: Determine the word sequence corresponding to the text information in each reference slice and the word sequence corresponding to the text information in the image slice; Based on the positional relationship between each reference slice and the image slice, the word sequence corresponding to each reference slice and the word sequence corresponding to the image slice are used to construct the sentence sequence of the image slice; The sentence sequence is used to obtain entity labels for the image slices through a classification and recognition model.

3. The method as described in claim 2, characterized in that, The classification and recognition model is the FastText model and uses N-gram word segmentation and combination.

4. The method as described in claim 2, characterized in that, The classification and recognition model is used to classify target entity labels; the target entity labels are determined from the entity labels in each packing list image.

5. The method according to any one of claims 1-4, characterized in that, Multiple image slices were identified from the packing list image, including: Text recognition technology was used to identify the text information and location information of each image slice from the packing list image.

6. A packing list image recognition device, characterized in that, The device includes: The transceiver unit is used to identify multiple image slices from the packing list image; The processing unit is configured to determine, for each image slice, reference slices of the image slice; the reference slices in the packing list image satisfy the adjacent position relationship with the image slice; each reference slice has an adjacent relationship with the image slice in different directions and each reference slice is the image slice closest to the image slice in the corresponding direction; Based on the text information in each reference slice, the text information in the image slice, and the positional relationship between each reference slice and the image slice, the entity label of the image slice is determined; the entity label is used to characterize the semantic object to which the text information belongs; based on the entity label of each image slice, the recognition result of the packing list image is obtained, and the recognition result is used to characterize whether the image slice is an important image slice.

7. A computing device, characterized in that, The device includes a processor and an interface circuit, the interface circuit being used to receive signals from other devices outside the device and transmit them to the processor or to send signals from the processor to other devices outside the device, the processor being used to implement the method as described in any one of claims 1 to 5 through logic circuits or execution code instructions.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions that, when executed on a computer, cause the computer to perform the method of any one of claims 1 to 5.

9. A computer program product, characterized in that, The computer program product includes: computer program code, which, when run on a computer, causes the computer to perform the method as described in any one of claims 1 to 5.

Images