Cross-border trade remittance processing method and device, electronic equipment and storage medium

Abstract

The invention discloses a cross-border trade remittance processing method and device, electronic equipment and a storage medium, and relates to the field of artificial intelligence and financial science and technology. The method comprises the following steps: in response to a remittance verification request of cross-border trade, interacting with a user to obtain a first remittance element, a first customs declaration number and a customs declaration image in a remittance interaction interface; extracting a second customs declaration number from the customs declaration image by adopting a title area positioning strategy; if the first customs declaration number is consistent with the second customs declaration number, inputting a first trade element according to the first customs declaration number; if the matching between the customs declaration image and the customs declaration template fails, performing optical character recognition on the customs declaration image based on multi-scale semantic segmentation and mixed feature optical character recognition to obtain image customs declaration data; and verifying the first remittance element, the first trade element and the image customs declaration data, and processing the remittance verification request according to a verification result. According to the technical scheme, the efficiency and accuracy of remittance verification of cross-border trade are improved.

Description

Methods, devices, electronic equipment, and storage media for processing cross-border trade remittances Technical Field

[0001] This application relates to the field of computer technology, and more particularly to the fields of artificial intelligence and financial technology, specifically to a method, apparatus, electronic device, and storage medium for processing cross-border trade remittances. Background Technology

[0002] With the rapid development of the global economy and the growth of international trade, cross-border trade has become an important component of international trade, and the demand for verifying customs declaration information in inbound remittances for cross-border goods trade is increasing. In goods trade, the customs declaration is one of the most important documents, containing a wealth of information such as goods name, quantity, price, freight, and insurance. Verifying the information in the customs declaration is a crucial step in ensuring the smooth operation of goods trade. When processing inbound remittances for cross-border export trade, staff must strictly verify the consistency between the inbound remittance information and the export customs declaration information; this is a vital step in anti-money laundering, foreign exchange compliance, and trade authenticity verification.

[0003] Currently, the traditional method of verifying customs declaration information mainly relies on manual operation. Differences in customs declaration formats between different countries lead to low processing efficiency. The information on the customs declaration is often blurry or handwritten, resulting in low recognition accuracy and a high risk of errors. Moreover, it requires a lot of manpower and time. Summary of the Invention

[0004] This application provides a method, apparatus, electronic device, and storage medium for processing cross-border trade remittances, so as to improve the efficiency and accuracy of remittance verification for cross-border trade.

[0005] In a first aspect, embodiments of this application provide a method for processing cross-border trade remittances, including:

[0006] In response to a remittance verification request in cross-border trade, the system interacts with the user to obtain the first remittance element and the first customs declaration data from the remittance interface; the first customs declaration data includes the first customs declaration number and the customs declaration image.

[0007] A title region positioning strategy is used to extract the second customs declaration number from the customs declaration image;

[0008] If the first customs declaration number and the second customs declaration number are the same, then the preset interface is called, and the first trade element is entered in the remittance interaction interface according to the first customs declaration number;

[0009] The customs declaration image is matched with customs declaration templates in the customs declaration template library. If the match fails, optical character recognition is performed on the customs declaration image based on multi-scale semantic segmentation and hybrid feature optical character recognition to obtain image customs declaration data; the image customs declaration data includes second trade data.

[0010] The first remittance element, the first trade element, and the image customs declaration data are verified, and the remittance verification request is processed based on the verification results.

[0011] Secondly, embodiments of this application also provide a processing apparatus for cross-border trade remittances, including:

[0012] The user interaction module is used to respond to remittance verification requests in cross-border trade and interact with users to obtain the first remittance element and the first customs declaration data from the remittance interaction interface; the first customs declaration data includes the first customs declaration number and the customs declaration image.

[0013] The customs declaration number extraction module is used to extract the second customs declaration number from the customs declaration image using a title area positioning strategy.

[0014] The interface call module is used to call a preset interface if the first customs declaration number and the second customs declaration number are the same, and input the first trade element in the remittance interaction interface according to the first customs declaration number;

[0015] The image extraction module is used to match the customs declaration image with customs declaration templates in the customs declaration template library. If the matching fails, optical character recognition is performed on the customs declaration image based on multi-scale semantic segmentation and hybrid feature optical character recognition to obtain image customs declaration data; the image customs declaration data includes second trade data.

[0016] The request processing module is used to verify the first remittance element, the first trade element, and the image customs declaration data, and to process the remittance verification request based on the verification results.

[0017] Thirdly, embodiments of this application also provide an electronic device, which includes:

[0018] One or more processors;

[0019] Storage device for storing one or more programs;

[0020] When one or more programs are executed by one or more processors, the one or more processors implement any of the cross-border trade remittance processing methods provided in the embodiments of this application.

[0021] Fourthly, embodiments of this application also provide a computer program product, which includes a computer program that, when executed by a processor, implements any of the cross-border trade remittance processing methods provided in embodiments of this application.

[0022] This application, in response to remittance verification requests, obtains the first remittance element, the first customs declaration number, and the customs declaration image submitted by the user from the remittance interaction interface; extracts the second customs declaration number from the customs declaration image using a title region positioning strategy; if the first and second customs declaration numbers match, a preset interface is automatically invoked to backfill the first trade element into the interaction interface, thereby reducing repetitive operations caused by manual entry errors. If the customs declaration image fails to match any templates in the template library, multi-scale semantic segmentation and hybrid feature optical character recognition are combined to extract customs declaration data from the image, and cross-verification is performed on the first remittance element, the first trade element, and the image customs declaration data. The remittance verification request is then processed based on the verification results, improving the efficiency and accuracy of remittance verification for cross-border trade.

[0023] Therefore, the technical solution of this application solves the problem and achieves the desired effect. Attached Figure Description

[0024] Figure 1 is a flowchart of a cross-border trade remittance processing method provided according to Embodiment 1 of this application;

[0025] Figure 2 is a flowchart of another method for processing cross-border trade remittances according to Embodiment 2 of this application;

[0026] Figure 3 is a flowchart of another method for processing cross-border trade remittances according to Embodiment 3 of this application;

[0027] Figure 4 is a schematic diagram of a cross-border trade remittance processing device provided according to Embodiment 4 of this application;

[0028] Figure 5 is a schematic diagram of the structure of an electronic device that implements the cross-border trade remittance processing method of the present application embodiment. Detailed Implementation

[0029] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort should fall within the scope of protection of the present application.

[0030] It should be noted that the terms "first" and "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0031] Example 1

[0032] Figure 1 is a flowchart of a cross-border trade remittance processing method according to Embodiment 1 of this application. This embodiment is applicable to the processing of cross-border remittances received by corporate customers and can be executed by a cross-border trade remittance processing device. This device can be implemented in hardware and / or software and can be configured in a computer device. As shown in Figure 1, the method includes:

[0033] S101. In response to a remittance verification request for cross-border trade, interact with the user to obtain the first remittance element and the first customs declaration data from the remittance interaction interface; the first customs declaration data includes the first customs declaration number and the customs declaration image;

[0034] S102. Using a title area positioning strategy, extract the second customs declaration number from the customs declaration image;

[0035] S103. If the first customs declaration number and the second customs declaration number are the same, then call the preset interface and input the first trade element in the remittance interaction interface according to the first customs declaration number.

[0036] S104. Match the customs declaration image with the customs declaration templates in the customs declaration template library. If the matching fails, perform optical character recognition on the customs declaration image based on multi-scale semantic segmentation and hybrid feature optical character recognition to obtain image customs declaration data; the image customs declaration data includes second trade data.

[0037] S105. Verify the first remittance element, the first trade element, and the image customs declaration data, and process the remittance verification request based on the verification results.

[0038] Upon receiving inbound remittances related to cross-border trade, the bank's operating system generates a remittance verification request and sends a cross-border remittance notification to the corporate client's representative. After receiving the notification, the corporate client's representative (user) can interact with the user through the remittance interface of the corporate online banking system to obtain the structured data submitted by the user. This structured data may include the first remittance element and the first customs declaration data. The first customs data includes the first customs declaration number and a customs declaration image. The customs declaration image can be uploaded as an attachment, including but not limited to DOC, TXT, PDF, PNG, and JPEG formats, and supports multi-source data input such as JPEG, PDF, and fax. The first remittance element may include at least one of the following: overseas remitter, remittance account number, amount, payment currency; domestic payee, payee account number, payee currency, and receipt date. The remittance interface may also include an automatically generated transaction serial number.

[0039] To improve user interaction efficiency and the accuracy of structured data in the remittance interface, a title region positioning strategy is adopted. This involves locating the title position in the customs declaration image and extracting the second customs declaration number based on that position. The system compares the user-inputted first customs declaration number with the extracted second customs declaration number. If they match, a pre-defined interface is invoked to send a query request carrying the first customs declaration number to the customs server. The customs server then queries the first trade element associated with the first customs declaration number and inputs it into the corresponding trade element field in the remittance interface. If they do not match, a customs declaration number error alert is generated, prompting the user to update the first customs declaration number or the customs declaration image. The first trade element includes at least one of the following: contract number, invoice number, commodity name, commodity quantity, and unit price.

[0040] To improve the recognition efficiency and accuracy of customs declaration images, the images are matched with customs declaration templates in a template library. For example, feature extraction can be performed on the customs declaration image to obtain customs declaration features, and the similarity between these features and the pre-extracted features of the customs declaration templates can be calculated. If the similarity between the customs declaration image and each template image is less than a similarity threshold, the matching is considered a failure. Multi-scale semantic segmentation is then performed on the customs declaration image to obtain multiple regions. Combined with the region type, layout, and other mixed features, optical character recognition is performed to obtain image customs declaration data. Image customs declaration data may include second trade data. This second trade data may include at least one of the following: commodity name, commodity number, and commodity price. The image customs declaration data also includes at least one of the following: trade terms price, transportation information, shipping and receiving information, and transport documents. The trade terms price includes CIF price and FOB price. The transportation information includes the mode of transport or transportation time. The shipping and receiving information includes the consignee and the shipper. The transport documents include invoices or packing lists.

[0041] Furthermore, robotic process automation (RPA) technology is used to compare and verify the first remittance element, the first trade element, and the image customs declaration data extracted from the customs declaration image in the remittance interaction interface, obtaining verification results. For example, the accuracy of commodity information, price information, transportation information, shipping and receiving information, and transport documents can be verified. If the verification passes, the bank's operating system will display "Customer has declared, automatic review passed" under the payment processing menu, and the transaction will proceed to automatic accounting. If the verification fails, the bank's operating system will display "Customer has declared, automatic review failed," and subsequent processing will require manual intervention. If the customs declaration image matches any customs declaration template, the text content of the corresponding element is extracted from the customs declaration image using the predefined element area location in the customs declaration template.

[0042] The technical solution of this embodiment, in response to a remittance verification request, obtains the first remittance element, the first customs declaration number, and the customs declaration image submitted by the user from the remittance interaction interface; extracts the second customs declaration number from the customs declaration image using a title region positioning strategy; if the first customs declaration number matches the second customs declaration number, a preset interface is automatically invoked to backfill the first trade element into the interaction interface, thereby reducing repetitive operations caused by manual entry errors. If the customs declaration image fails to match any templates in the template library, multi-scale semantic segmentation and hybrid feature optical character recognition are combined to extract customs declaration data from the image, which can improve the data extraction accuracy in complex scenarios. Furthermore, cross-verification is performed on the first remittance element, the first trade element, and the image customs declaration data, and the remittance verification request is processed based on the verification results, improving the efficiency and accuracy of remittance verification for cross-border trade.

[0043] Example 2

[0044] Figure 2 is a flowchart of another cross-border trade remittance processing method according to Embodiment 2 of this application. The technical solution of this embodiment further refines the title area positioning strategy based on the above technical solution. Referring to Figure 2, a cross-border trade remittance processing method includes:

[0045] S201. In response to a remittance verification request for cross-border trade, interact with the user to obtain the first remittance element and the first customs declaration data from the remittance interaction interface; the first customs declaration data includes the first customs declaration number and the customs declaration image;

[0046] S202. Using a title area positioning strategy, extract the second customs declaration number from the customs declaration image;

[0047] S203. If the first customs declaration number and the second customs declaration number are the same, then call the preset interface and input the first trade element in the remittance interaction interface according to the first customs declaration number.

[0048] S204. Match the customs declaration image with the customs declaration templates in the customs declaration template library. If the matching fails, input the customs declaration image into the multi-scale fusion semantic segmentation model to generate a multi-scale feature map.

[0049] S205. Cluster the multi-scale feature map to generate a heat map, and segment the customs declaration image into multiple regions to be identified according to the weight distribution of the heat map.

[0050] S206. Classify the region to be identified to obtain the region type to which the region to be identified belongs;

[0051] S207. Using a hybrid feature optical character recognition model, optical character recognition is performed on the region to be recognized according to the corresponding region type to obtain image customs declaration data; the image customs declaration data includes second trade data.

[0052] S208. Verify the first remittance element, the first trade element, and the image customs declaration data, and process the remittance verification request based on the verification results.

[0053] If the customs declaration image fails to match any of the customs declaration templates in the template library, indicating an unknown format, automatic segmentation can be performed using an unsupervised method. For example, the customs declaration image can be input into the encoder of a U-shaped network to obtain initial multi-scale features. These initial features are then input into the decoder and the first deformer encoder of the U-shaped network, respectively, to obtain the local feature map output by the U-shaped network decoder and the global feature sequence containing global dependencies output by the first deformer encoder. The local feature map and the global feature sequence are then fused to obtain a multi-scale feature map incorporating both local and global information. An unsupervised clustering algorithm is used to cluster the multi-scale feature map, generating a block heatmap. Based on the weight distribution of the heatmap, the customs declaration image is segmented into multiple regions to be identified. For each region to be identified, local features such as texture, edges, and color distribution, as well as contextual features with adjacent regions, can be extracted. The local and contextual features of the region to be identified are then input into a pre-trained classification type to obtain the region type to which the region belongs. Region types include table areas, handwritten areas, or stamp areas, etc.

[0054] For example, the region type includes at least one of a table region, a handwritten region, or a seal region; the hybrid feature optical character recognition model includes a feature extraction network and a deformer; the feature extraction network used in the table region has fewer layers than the feature extraction network used in the handwritten region; the seal region uses a dilated convolutional network. The hybrid feature optical character recognition model can employ a hybrid structure including a feature extraction network and a deformer network. The feature extraction network used in the table region has fewer layers than the feature extraction network used in the handwritten region; the seal region uses a dilated convolutional network. Different region types correspond to different feature extraction network structures. The feature extraction network used in the table region has fewer layers than the feature extraction network used in the handwritten region, thus balancing the processing efficiency and recognition accuracy of the handwritten region; the seal region corresponds to a dilated convolutional network, which expands the receptive field through dilated convolution to capture the overall shape of the seal pattern, circular text, and artistic font features, thereby improving the accuracy of seal recognition.

[0055] For example, based on the region type of the region to be identified, the image of the region to be identified is input into the corresponding feature extraction network for feature extraction, resulting in a local feature map. This local feature map is then converted into a sequence and concatenated with the corresponding region type before being input into the second deformer encoder. The output sequences of the local feature map and the second deformer encoder are then fused through cross-attention to obtain a fused feature. This fused feature is then input into the decoder of the second deformer to obtain the image customs declaration data corresponding to the region to be identified. By performing optical character recognition on the region to be identified based on the corresponding region type, especially for cursive characters in handwritten areas, the self-attention mechanism can capture the contextual information of cursive characters, further improving the accuracy of optical character recognition for customs declaration images with unknown formats.

[0056] The technical solution of this embodiment, for customs declaration images of unknown format, accurately segments the customs declaration image into multiple regions to be identified by combining multi-scale fusion semantic segmentation and unsupervised clustering algorithms; according to the region type to which the region to be identified belongs, such as table area, handwritten area, or stamp area, a suitable feature extraction network is dynamically selected to obtain local feature maps, thereby improving the accuracy of local feature maps and further improving the accuracy of remittance verification for cross-border trade.

[0057] In one optional implementation, the step of using a title region positioning strategy to extract the second customs declaration number from the customs declaration image includes: determining a title region from the top of the customs declaration image by combining preset title keywords; obtaining candidate text by performing optical character recognition based on the title region; and filtering the second customs declaration number from the candidate text according to a preset customs declaration number format.

[0058] The title keywords can be things like "customs declaration" or "declaration form." For example, based on the title keywords and font format (e.g., size, bold), the title area can be located quickly and accurately from the customs declaration image. Then, based on the title area's location, optical character recognition (OCR) is performed on a region of 50-200 pixels below the title to obtain candidate text. A second customs declaration number is then selected from the candidate text using a preset format. For example, regular expressions are used to match the customs declaration number format (e.g., an 18-digit combination of numbers and letters) to quickly filter out other text. By performing OCR only on the area near the title, computational load is reduced; furthermore, customs declaration number format matching improves the efficiency and accuracy of customs declaration number recognition.

[0059] If the second customs declaration number is not identified near the title area, the preset interface can still be called to automatically retrieve the first trade element based on the first customs declaration number; alternatively, the first trade element entered by the user in the remittance interaction interface can be obtained; and cross-validation can be performed on the first trade element and the first customs declaration number. If the first customs declaration number matches the commodity name and specifications in the first trade element, the cross-validation passes; otherwise, the cross-validation fails. For example, if both the first customs unit and the first trade element correspond to a dress, the cross-validation passes; otherwise, the cross-validation fails.

[0060] In one optional implementation, the step of obtaining candidate text by optical character recognition based on the title region includes: determining the height and font format of the title region, and expanding the position of the title region according to the height and font format to obtain a title extension region; and performing optical character recognition on the title extension region to obtain the candidate text.

[0061] For example, the height of the title area, as well as font size, bolding, and other font formatting parameters, can be determined. An expansion coefficient matching the title area's height, font size, bolding, and other title attributes can be determined, and the title area is dynamically expanded using this expansion coefficient to obtain an expanded title area. The expansion coefficient value is greater than 1. Optical character recognition (OCR) is then performed on the expanded title area to obtain candidate text. By dynamically expanding the title area in conjunction with title attributes, title expansion areas of different sizes can be obtained, thus balancing the efficiency and success rate of customs declaration number recognition.

[0062] Example 3

[0063] Figure 3 is a flowchart of another method for processing cross-border trade remittances according to Embodiment 3 of this application. The technical solution of this embodiment further refines the generation of the customs declaration template based on the above-mentioned technical solution. Referring to Figure 3, a method for processing cross-border trade remittances includes:

[0064] S301. In response to a remittance verification request for cross-border trade, interact with the user to obtain the first remittance element and the first customs declaration data from the remittance interaction interface; the first customs declaration data includes the first customs declaration number and the customs declaration image;

[0065] S302. Using a title area positioning strategy, extract the second customs declaration number from the customs declaration image;

[0066] S303. If the first customs declaration number and the second customs declaration number are the same, then call the preset interface and input the first trade element in the remittance interaction interface according to the first customs declaration number.

[0067] S304. Match the customs declaration image with the customs declaration templates in the customs declaration template library. If the matching fails, perform optical character recognition on the customs declaration image based on multi-scale semantic segmentation and hybrid feature optical character recognition to obtain image customs declaration data; the image customs declaration data includes second trade data.

[0068] S305. Verify the first remittance element, the first trade element, and the image customs declaration data, and process the remittance verification request based on the verification results.

[0069] S306. Obtain the M customs declaration images that fail to match any customs declaration template, and the image layout and structural features of the M customs declaration images.

[0070] S307. Cluster the M customs declaration images according to the image layout structure features to obtain multiple categories;

[0071] S308. For each category, input the image layout structure features and noise features of the customs declaration images in that category into the generator to generate the corresponding customs declaration template image.

[0072] S309. Input the different customs declaration template images corresponding to the category into the discriminator to obtain the similarity between the different customs declaration template images, and update the generator according to the similarity to generate a customs declaration template generation model.

[0073] The layout structure features can include field position, field name, field style, etc.; the field style can be font type, color histogram, or line thickness, etc. For M customs declaration images of unknown format, the layout structure features can be extracted from each of the unknown format customs declaration images to obtain image layout structure features; the image layout structure features are then used to cluster the M customs declaration images to obtain N categories.

[0074] For each category, multiple customs declaration images within that category are acquired. The image layout and structural features of these images, along with random noise features, are input into a generator to obtain the corresponding customs declaration template image. Different customs declaration template images are then input into a discriminator; for example, the i-th and j-th customs declaration template images within the same category are input to the discriminator to obtain the similarity between the i-th and j-th template images. The generator is updated based on this similarity until training is complete, and the trained generator serves as the customs declaration template generation model. Here, M, N, i, and j are all positive integers. It should be noted that if the number of customs declaration images in a category is small, geometric transformations such as rotation and perspective can be used for sample augmentation. By clustering customs declaration images of unknown formats, categories are obtained. For each category, adversarial learning and intra-class similarity constraints are used to construct a customs declaration template generation model, which strengthens intra-class consistency and improves the quality of the model.

[0075] In one optional implementation, after generating the customs declaration template generation model, the method further includes: determining the central layout structure features of the category based on the image layout structure features of each customs declaration image in the category; and inputting the central layout structure features and noise features into the customs declaration template generation model to obtain the customs declaration template image of the category.

[0076] For each category, the mean of the image layout structure features of each customs declaration image in that category can be statistically analyzed to obtain the central layout structure features of that category. These central layout structure features and noise features are then input into the customs declaration template generation model to obtain the customs declaration template image for that category, which is then added to the customs declaration template library. This process automates the generation of customs declaration templates for unknown customs declaration images, thereby improving the recognition efficiency of subsequent customs declaration images.

[0077] The technical solution of this embodiment performs cluster analysis on customs declaration images of unknown format to classify them into categories, and constructs a customs declaration template generation model for each category using adversarial learning combined with intra-class similarity constraints. This effectively strengthens intra-class layout consistency and improves template generation quality. Furthermore, by extracting the central layout features from the mean of the layout structure features of similar images and fusing noise features as model input, customs declaration template images of that category can be automatically generated. This facilitates rapid template adaptation for customs declarations of unknown format and improves the efficiency and accuracy of subsequent optical character recognition.

[0078] For example, multi-level verification can be performed on the first remittance element, the first trade element, and image customs declaration data. The first level verifies the compliance of basic fields such as amount, date, and currency, preventing basic errors. For the amount format, it must be a number (including decimal points, e.g., 25000.00), positive, and free of illegal symbols (e.g., $ needs to be separated into the currency field); regular expressions can be used to automatically remove thousands separators (e.g., 25,000.00 → 25000.00); the payment interface is called to verify the legality of the amount value. For the date format, it needs to conform to international standards (e.g., YYYY-MM-DD or DD / MM / YYYY) and be logically valid (e.g., shipment date ≤ letter of credit validity period). For example, multi-format parsing templates can be preset to automatically convert dates to a unified format; and date logic is validated, for example, if the shipment date is later than the current date, the date is marked as abnormal. For currency formats, it can conform to the ISO 4217 standard (such as USD, CNY) and match the transaction country (e.g., Chinese customers need to use CNY instead of RMB). If a Level 1 anomaly is detected, it can automatically revert to correction, log the error, and notify the user.

[0079] The secondary verification is used to check the dynamic fluctuation range (±5%) between the declared amount and the foreign exchange receipt amount, mitigating business logic risks such as money laundering or false declarations. The foreign exchange receipt amount must be within a preset range (e.g., 95% to 105%) of the declared amount, and exchange rate fluctuations must be dynamically adjusted based on the central bank's midpoint rate on the day of customs declaration. For example, the real-time exchange rate can be obtained by calling the exchange rate API. The floating receipt amount can be calculated as the product of the received foreign exchange amount and the real-time exchange rate; the fluctuation range can be calculated based on the floating receipt amount and the declared amount. Alternatively, secondary verification can be performed using the following formula: Allowable fluctuation range = Declared amount × (1 ± 0.05 × Exchange rate volatility / Base exchange rate). If a secondary anomaly is detected, the process is transferred to manual review and the account manager is notified.

[0080] The three verification steps are used to link with external systems to verify the validity of customs codes, license status, etc. The customs (HS) code must match the product name and specifications (e.g., 6204.43 corresponds to "women's dress"), and be the latest version from the General Administration of Customs. If a level-three anomaly is detected, remittances can be frozen and a compliance investigation process can be triggered.

[0081] Example 4

[0082] Figure 4 is a schematic diagram of a cross-border trade remittance processing device according to Embodiment 4 of this application. This embodiment is applicable to processing cross-border remittances received by corporate clients. The cross-border trade remittance processing device can be implemented in hardware and / or software, and can be configured in a computer device. Referring to Figure 4, the specific structure of the cross-border trade remittance processing device 400 is as follows:

[0083] User interaction module 410 is used to respond to remittance verification requests in cross-border trade and interact with users to obtain the first remittance element and the first customs declaration data in the remittance interaction interface; the first customs declaration data includes the first customs declaration number and the customs declaration image.

[0084] The customs declaration number extraction module 420 is used to extract the second customs declaration number from the customs declaration image using a title area positioning strategy.

[0085] The interface calling module 430 is used to call a preset interface if the first customs declaration number and the second customs declaration number are the same, and input the first trade element in the remittance interaction interface according to the first customs declaration number;

[0086] The image extraction module 440 is used to match the customs declaration image with customs declaration templates in the customs declaration template library. If the matching fails, optical character recognition is performed on the customs declaration image based on multi-scale semantic segmentation and hybrid feature optical character recognition to obtain image customs declaration data. The image customs declaration data includes second trade data.

[0087] The request processing module 450 is used to verify the first remittance element, the first trade element, and the image customs declaration data, and to process the remittance verification request based on the verification results.

[0088] In one alternative embodiment, the image extraction module 440 includes:

[0089] The feature extraction unit is used to input the customs declaration image into the multi-scale fusion semantic segmentation model to generate a multi-scale feature map;

[0090] The region segmentation unit is used to cluster the multi-scale feature map to generate a heat map, and to segment the customs declaration image into multiple regions to be identified according to the weight distribution of the heat map.

[0091] A region classification unit is used to classify the region to be identified to obtain the region type to which the region to be identified belongs;

[0092] The image extraction unit is used to perform optical character recognition on the region to be recognized according to the corresponding region type using a hybrid feature optical character recognition model to obtain image customs declaration data.

[0093] In one optional implementation, the region type includes at least one of a table region, a handwritten region, or a stamp region; the hybrid feature optical character recognition model includes a feature extraction network and a deformer; the feature extraction network used in the table region has fewer network layers than the feature extraction network used in the handwritten region; and the stamp region uses a dilated convolutional network.

[0094] In one optional embodiment, the cross-border trade remittance processing device 400 further includes a generative model building module, the generative model building module comprising:

[0095] The image layout feature unit is used to obtain the M customs declaration images that fail to match each customs declaration template, as well as the image layout structure features in the M customs declaration images.

[0096] An image clustering unit is used to cluster M customs declaration images according to the image layout and structural features to obtain multiple categories;

[0097] The template generation unit is used to input the image layout structure features and noise features of the customs declaration images in each category into the generator to generate the corresponding customs declaration template image.

[0098] The generator update unit is used to input different customs declaration template images corresponding to the category into the discriminator, obtain the similarity between different customs declaration template images, and update the generator according to the similarity to generate a customs declaration template generation model.

[0099] In one optional embodiment, the cross-border trade remittance processing device 400 further includes a template generation module, the template generation module comprising:

[0100] The central layout feature unit is used to determine the central layout structure feature of the category based on the image layout structure features of each customs declaration image in the category.

[0101] The template generation unit is used to input the central layout structural features and noise features into the customs declaration template generation model to obtain the customs declaration template image of this category.

[0102] In one optional embodiment, the customs declaration number extraction module 420 includes:

[0103] The title positioning unit is used to determine the title area from the top of the customs declaration image by combining preset title keywords;

[0104] A candidate text unit is used to obtain candidate text based on optical character recognition of the title region;

[0105] The customs declaration number unit is used to filter the second customs declaration number from the candidate text according to a preset customs declaration number format.

[0106] In one optional implementation, the customs declaration number unit is specifically used for:

[0107] Determine the height and font format of the title area, and expand the position of the title area according to the height and font format to obtain the title expansion area;

[0108] The candidate text is obtained by performing optical character recognition on the extended title area.

[0109] The cross-border trade remittance processing apparatus provided in this application embodiment can execute the cross-border trade remittance processing method provided in any embodiment of this application, and has the corresponding functional modules and beneficial effects for executing the cross-border trade remittance processing method.

[0110] According to embodiments of the present invention, the present invention also provides an electronic device, a readable storage medium, and a computer program product.

[0111] Example 5

[0112] Figure 5 is a schematic diagram of the structure of an electronic device 510 implementing the cross-border trade remittance processing method according to an embodiment of this application. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device may also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices (e.g., helmets, glasses, watches, etc.), and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present application described and / or claimed herein.

[0113] As shown in Figure 5, the electronic device 510 includes at least one processor 511 and a memory, such as a read-only memory (ROM) 512 or a random access memory (RAM) 513, communicatively connected to the at least one processor 511. The memory stores computer programs executable by the at least one processor. The processor 511 can perform various appropriate actions and processes based on the computer program stored in the ROM 512 or loaded from storage unit 518 into the RAM 513. The RAM 513 can also store various programs and data required for the operation of the electronic device 510. The processor 511, ROM 512, and RAM 513 are interconnected via a bus 514. An input / output (I / O) interface 515 is also connected to the bus 514.

[0114] Multiple components in electronic device 510 are connected to I / O interface 515, including: input unit 516, such as keyboard, mouse, etc.; output unit 517, such as various types of displays, speakers, etc.; storage unit 518, such as disk, optical disk, etc.; and communication unit 519, such as network card, modem, wireless transceiver, etc. Communication unit 519 allows electronic device 510 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0115] Processor 511 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of processor 511 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. Processor 511 performs the various methods and processes described above, such as the processing methods for cross-border trade remittances.

[0116] In some embodiments, the method for processing cross-border trade remittances may be implemented as a computer program tangibly contained in a computer-readable storage medium, such as storage unit 518. In some embodiments, part or all of the computer program may be loaded and / or installed on electronic device 510 via ROM 512 and / or communication unit 519. When the computer program is loaded into RAM 513 and executed by processor 511, one or more steps of the cross-border trade remittance processing method described above may be performed. Alternatively, in other embodiments, processor 511 may be configured as the cross-border trade remittance processing method by any other suitable means (e.g., by means of firmware).

[0117] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-a-chip (SoCs), payload-programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0118] Computer programs used to implement the methods of this application may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable device, such that when executed by the processor, the computer programs cause the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The computer programs may be executed entirely on a machine, partially on a machine, or as a standalone software package, partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0119] In the context of this application, a computer-readable storage medium can be a tangible medium that may contain or store a computer program for use by or in conjunction with an instruction execution system, apparatus, or device. A computer-readable storage medium can be, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. Alternatively, a computer-readable storage medium can be a machine-readable signal medium. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0120] To provide interaction with a user, the systems and techniques described herein can be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the electronic device. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0121] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as data servers), or computing systems that include middleware components (e.g., application servers), or computing systems that include frontend components (e.g., user computers with graphical user interfaces or web browsers through which users can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., communication networks). Examples of communication networks include local area networks (LANs), wide area networks (WANs), blockchain networks, and the Internet.

[0122] A computing system can include clients and servers. Clients and servers are generally located far apart and typically interact through communication networks. The client-server relationship is created by computer programs running on the respective computers and having a client-server relationship with each other. The server can be a cloud server, also known as a cloud computing server or cloud host, which is a hosting product within the cloud computing service system to address the shortcomings of traditional physical hosts and VPS services, such as high management difficulty and weak business scalability.

[0123] It should be understood that the various forms of processes shown above can be used to rearrange, add, or delete steps. For example, the steps described in this application can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution of this application can be achieved, and this is not limited herein.

[0124] The specific embodiments described above do not constitute a limitation on the scope of protection of this application. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A method for processing cross-border trade remittances, characterized in that, include: In response to remittance verification requests in cross-border trade, the system interacts with users to obtain the first remittance element and the first customs declaration data from the remittance interface. The first customs declaration data includes a first customs declaration number and a customs declaration image; a title region positioning strategy is used to extract a second customs declaration number from the customs declaration image; if the first customs declaration number and the second customs declaration number are consistent, a preset interface is called, and a first trade element is input in the remittance interaction interface according to the first customs declaration number; the customs declaration image is matched with customs declaration templates in the customs declaration template library; if the matching fails, optical character recognition is performed on the customs declaration image based on multi-scale semantic segmentation and hybrid feature optical character recognition to obtain image customs declaration data; the image customs declaration data includes second trade data; the first remittance element, the first trade element, and the image customs declaration data are verified, and the remittance verification request is processed according to the verification results.

2. The method according to claim 1, characterized in that, A dynamic layout analysis and hybrid feature optical character recognition (HCR) method is used to perform optical character recognition on the customs declaration image to obtain image customs declaration data. This includes: inputting the customs declaration image into a multi-scale fusion semantic segmentation model to generate a multi-scale feature map; clustering the multi-scale feature map to generate a heatmap, and segmenting the customs declaration image into multiple regions to be identified based on the weight distribution of the heatmap; classifying the regions to be identified to obtain their respective region types; and using a hybrid feature optical character recognition model to perform optical character recognition on the regions to be identified according to their corresponding region types to obtain image customs declaration data.

3. The method according to claim 2, wherein, The region type includes at least one of a table region, a handwritten region, or a seal region; the hybrid feature optical character recognition model includes a feature extraction network and a deformer; the feature extraction network used in the table region has fewer network layers than the feature extraction network used in the handwritten region; the seal region uses a dilated convolutional network.

4. The method according to claim 1, characterized in that, The method further includes: acquiring M customs declaration images that fail to match any customs declaration template, and image layout structure features in the M customs declaration images; clustering the M customs declaration images according to the image layout structure features to obtain multiple categories; for each category, inputting the image layout structure features and noise features of the customs declaration images in that category into a generator to generate a corresponding customs declaration template image; inputting different customs declaration template images corresponding to that category into a discriminator to obtain the similarity between different customs declaration template images, and updating the generator according to the similarity to generate a customs declaration template generation model.

5. The method according to claim 4, characterized in that, After generating the customs declaration template generation model, the method further includes: determining the central layout structure features of the category based on the image layout structure features of each customs declaration image in the category; inputting the central layout structure features and noise features into the customs declaration template generation model to obtain the customs declaration template image of the category.

6. The method according to claim 1, characterized in that, The method of using a title region positioning strategy to extract the second customs declaration number from the customs declaration image includes: determining the title region from the top of the customs declaration image by combining preset title keywords; obtaining candidate text by performing optical character recognition based on the title region; and filtering the second customs declaration number from the candidate text according to a preset customs declaration number format.

7. The method according to claim 6, characterized in that, The step of obtaining candidate text by performing optical character recognition based on the title region includes: determining the height and font format of the title region, and expanding the position of the title region according to the height and font format to obtain a title extension region; and performing optical character recognition on the title extension region to obtain the candidate text.

8. A processing device for cross-border trade remittances, characterized in that, include: The user interaction module is used to respond to remittance verification requests in cross-border trade and interact with users to obtain the first remittance element and the first customs declaration data from the remittance interaction interface. The first customs declaration data includes a first customs declaration number and a customs declaration image; the customs declaration number extraction module is used to extract a second customs declaration number from the customs declaration image using a title area positioning strategy; The interface call module is used to call a preset interface if the first customs declaration number and the second customs declaration number are the same, and input the first trade element in the remittance interaction interface according to the first customs declaration number; The image extraction module is used to match the customs declaration image with customs declaration templates in the customs declaration template library. If the matching fails, optical character recognition is performed on the customs declaration image based on multi-scale semantic segmentation and hybrid feature optical character recognition to obtain image customs declaration data; the image customs declaration data includes second trade data. The request processing module is used to verify the first remittance element, the first trade element, and the image customs declaration data, and to process the remittance verification request based on the verification results.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the cross-border trade remittance processing method as described in any one of claims 1-7.

10. A computer program product, characterized in that, The computer program product includes a computer program that, when executed by a processor, implements the method for processing cross-border trade remittances according to any one of claims 1-7.

Images