Image-based transaction method, apparatus, electronic device, and computer program product

Abstract

The application discloses an image-based transaction method and device, electronic equipment and computer program product. It relates to the field of financial technology, and the method comprises the following steps: receiving an image recognition instruction initiated by a target user through a client; under the authorization of the target user, obtaining the user image of the target user based on the image recognition instruction; inputting the user image into an image recognition model to obtain the identity recognition result of the target user, wherein the image recognition model comprises a feature extraction module and a classification module, the image recognition model is used for feature extraction of the user image through the feature extraction module, and identity recognition is performed through the classification module to obtain the identity recognition result; and determining a transaction instruction according to the identity recognition result. Through the application, the technical problem of low recognition accuracy when recognizing a face image in the related art, and low transaction efficiency when executing a transaction is solved.

Description

Technical Field

[0001] This application relates to the field of financial technology, and more specifically, to an image-based trading method, device, electronic device, and computer program product. Background Technology

[0002] With the rapid development of mobile internet and artificial intelligence technologies, the fintech sector has undergone profound changes. Biometric identification technology, due to its convenience, contactless nature, and efficiency, has been widely applied in various payment scenarios. Among them, facial recognition payment, as a payment method, allows users to quickly complete identity verification and payment using only facial features, improving payment efficiency, reducing queuing time, and is particularly user-friendly for the elderly who are unfamiliar with operating smart devices, effectively lowering the barrier to digital payments.

[0003] However, despite the convenience and speed of these systems, several technical shortcomings and deficiencies have been exposed in complex real-world application environments. First, in payment scenarios, ambient lighting is often uneven, with some areas dimly lit or exposed to direct sunlight. Existing recognition algorithms are prone to feature extraction failures due to low image contrast or overexposure when dealing with such uneven lighting, resulting in a high failure rate and severely impacting the customer's payment experience. Second, current facial recognition models need improvement in accuracy when targeting specific groups such as twins and the elderly. For example, the highly similar facial features among relatives like twins can easily lead to misidentification.

[0004] Finally, some users habitually wear masks, hats, or high-necked clothing during daily outings, obscuring key facial features such as the eyes, mouth, and nose. The relevant systems have a low tolerance for occluded information, making it impossible to extract sufficient valid features for matching, thus causing payment process interruptions.

[0005] There is currently no effective solution to the technical problem of low recognition accuracy in facial image recognition, which leads to low transaction efficiency. Summary of the Invention

[0006] The main objective of this application is to provide an image-based transaction method, apparatus, electronic device, and computer program product to solve the technical problem of low recognition accuracy and low transaction efficiency when performing transactions in related technologies.

[0007] To achieve the above objectives, according to one aspect of this application, an image-based transaction method is provided. The method includes: receiving an image recognition instruction initiated by a target user through a client; obtaining a user image of the target user based on the image recognition instruction, with the target user's authorization; inputting the user image into an image recognition model to process and obtain an identity recognition result for the target user, wherein the image recognition model includes a feature extraction module and a classification module, the image recognition model being used to extract features from the user image through the feature extraction module and to perform identity recognition through the classification module to obtain an identity recognition result; and determining a transaction instruction based on the identity recognition result.

[0008] Optionally, acquiring the user image of the target user based on the image recognition instruction includes: invoking the acquisition device to acquire the initial user image of the target user based on the image recognition instruction; performing grayscale conversion on the initial user image to obtain the converted initial user image; performing tone processing on the converted initial user image to obtain the processed initial user image; determining the first feature location information of the target feature in the processed initial user image; and cropping the processed initial user image based on the first feature location information to obtain the user image.

[0009] Optionally, before inputting the user image into the image recognition model, the method further includes: performing image transformation on the user image to obtain a transformed user image, wherein the image transformation method includes at least: image rotation, image translation, and image flipping; determining the second feature position information of the target feature in the transformed user image; cropping the transformed user image based on the second feature position information to obtain a target image; and inputting the target image into the image recognition model to process and obtain an identity recognition result.

[0010] Optionally, inputting the target image into the image recognition model and processing it to obtain the identity recognition result includes: inputting the target image into the feature extraction module, whereby the feature extraction module performs feature extraction processing on the target image to obtain processed image features; and inputting the processed image features into the classification module, whereby the classification module converts the processed image features to obtain the identity recognition result.

[0011] Optionally, the feature extraction module includes a feature extraction submodule and a feature fusion submodule. The feature extraction module performs feature extraction processing on the target image to obtain processed image features, including: inputting the target image into the feature extraction submodule, where the feature extraction submodule performs convolution processing on the target image to obtain first convolutional image features; performing attention enhancement processing on the first convolutional image features to obtain enhanced image features; performing sampling processing on the enhanced image features to obtain sampled image features; inputting the sampled image features into the feature fusion submodule, where M dilated convolution branches in the feature fusion submodule perform dilated convolution processing on the sampled image features to obtain M second convolutional image features, where M is a positive integer; and performing feature fusion on the M second convolutional image features to obtain processed image features.

[0012] Optionally, the classification module transforms the processed image features to obtain the identity recognition result, including: pooling the processed image features to obtain an image feature vector; transforming the dimension of the image feature vector to obtain a transformed feature vector, and regularizing the transformed feature vector to obtain a processed feature vector; mapping the processed feature vector to a preset category space to obtain a mapping result, where the preset category space refers to the image information of users stored in a preset institution, and the mapping result is used to indicate the degree of matching between the processed feature vector and the image information of users stored in the preset institution; converting the mapping result into a mapping probability distribution, where the mapping probability distribution refers to the similarity probability distribution between the processed feature vector and the image information of users stored in the preset institution; extracting the highest probability value from the mapping probability distribution, and generating the identity recognition result based on the user information of the user corresponding to the highest probability value.

[0013] Optionally, determining the transaction instruction based on the identity verification result includes: generating a transaction instruction that allows the transaction if the identity verification result indicates that the target user's identity verification has passed; and generating a transaction instruction that disallows the transaction if the identity verification result indicates that the target user's identity verification has failed.

[0014] To achieve the above objectives, according to another aspect of this application, an image-based transaction apparatus is provided. The apparatus includes: a receiving unit for receiving an image recognition instruction initiated by a target user through a client; an acquisition unit for acquiring a user image of the target user based on the image recognition instruction, with the target user's authorization; a first input unit for inputting the user image into an image recognition model and processing it to obtain an identity recognition result for the target user, wherein the image recognition model includes a feature extraction module and a classification module, the image recognition model being used to extract features from the user image through the feature extraction module and to perform identity recognition through the classification module to obtain an identity recognition result; and a first determining unit for determining a transaction instruction based on the identity recognition result.

[0015] According to another aspect of the present invention, a computer-readable storage medium is also provided, the computer-readable storage medium including a stored executable program, wherein, when the executable program is executed, it controls the device where the computer-readable storage medium is located to execute any of the above-described image-based transaction methods.

[0016] According to another aspect of the present invention, an electronic device is also provided, including one or more processors and a memory, the memory storing an executable program, and the processor for running the program, wherein when the one or more programs are executed by the one or more processors, the one or more processors cause the one or more processors to implement any of the above-described image-based transaction methods.

[0017] According to another aspect of the present invention, a computer program product is also provided, the computer program product including a computer program, wherein the computer program, when executed by a processor, implements any of the above-described image-based transaction methods.

[0018] In this embodiment, an image-based transaction method is adopted. This involves receiving an image recognition command initiated by a target user through a client; with the target user's authorization, acquiring the target user's image based on the image recognition command; inputting the user image into an image recognition model for processing to obtain the target user's identity recognition result. The image recognition model includes a feature extraction module and a classification module. The image recognition model extracts features from the user image through the feature extraction module and performs identity recognition through the classification module to obtain the identity recognition result. A transaction command is determined based on the identity recognition result. This method solves the technical problem of low recognition accuracy and low transaction efficiency when recognizing facial images in related technologies. By inputting the user image into the image recognition model, processing it to obtain the target user's identity recognition result, and determining the transaction command based on the identity recognition result, the technical effect of improving the recognition accuracy of facial images is achieved. Attached Figure Description

[0019] The accompanying drawings, which form part of this application, are used to provide a further understanding of this application. The illustrative embodiments and descriptions of this application are used to explain this application and do not constitute an undue limitation of this application. In the drawings:

[0020] Figure 1 This is a hardware structure block diagram of a computer terminal (or mobile device) for implementing an image-based trading method;

[0021] Figure 2 This is a flowchart of an image-based transaction method provided according to an embodiment of this application;

[0022] Figure 3 This is a schematic diagram of an optional image-based transaction method provided according to an embodiment of this application;

[0023] Figure 4 This is a schematic diagram of a user image recognition method provided according to an embodiment of this application;

[0024] Figure 5 This is a schematic diagram of an image detection method provided according to an embodiment of this application;

[0025] Figure 6 This is a schematic diagram of an image preprocessing method provided according to an embodiment of this application;

[0026] Figure 7 This is a schematic diagram of an image-based transaction device provided according to an embodiment of this application;

[0027] Figure 8 This is a structural block diagram of an electronic device according to an embodiment of this application. Detailed Implementation

[0028] 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.

[0029] It should be noted that the terms "first," "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.

[0030] It should be noted that all information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for display, data used for analysis, etc.) involved in this application are information and data authorized by the user or fully authorized by all parties. For example, this system has interfaces with relevant users or organizations to provide users with corresponding operation data for them to choose to agree to or refuse automated decision-making results. Before obtaining relevant information, a request for obtaining the information needs to be sent to the aforementioned user or organization through the interface, and the relevant information is obtained after receiving consent from the aforementioned user or organization; if the user chooses to refuse, the expert decision-making process is initiated. Users can view the purpose of data use in real time through authorization decoding and have the right to withdraw authorization or delete data at any time. After the authorization is withdrawn, the system will terminate the relevant data processing within 24 hours.

[0031] It should be noted that the information collected in this application is information and data authorized by the user or fully authorized by all parties, and the collection, storage, use, processing, transmission, provision, disclosure and application of the relevant data all comply with the relevant laws, regulations and standards of the relevant regions, take necessary confidentiality measures, do not violate public order and good morals, and provide corresponding operation access points for users to choose to authorize use or refuse use.

[0032] Example 1

[0033] According to an embodiment of this application, an embodiment of a method for image-based transactions is also provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0034] The method embodiment provided in Embodiment 1 of this application can be executed on a mobile terminal, computer terminal, or similar computing device. Figure 1This is a hardware structure block diagram of a computer terminal (or mobile device) for implementing an image-based trading method, such as... Figure 1 As shown, computer terminal 10 (or mobile device) may include one or more ( Figure 1 The processor 102 (which may include, but is not limited to, a microprocessor MCU (Microcontroller Unit) or a programmable gate array (FPGA)) is shown as 102a, 102b, ..., 102n. It also includes a memory 104 for storing data and a transmission device 106 for communication functions. In addition, it may include: a display, an input / output interface, a Universal Serial Bus (USB) port (which may be included as one of the ports of a BUS bus), a network interface, a keyboard, a cursor control device, a power supply, and / or a camera. Those skilled in the art will understand that... Figure 1 The structure shown is for illustrative purposes only and does not limit the structure of the aforementioned electronic device. For example, computer terminal 10 may also include... Figure 1 The more or fewer components shown, or having the same Figure 1 The different configurations shown.

[0035] It should be noted that the aforementioned one or more processors 102 and / or other data processing circuits are generally referred to herein as "data processing circuits". These data processing circuits may be embodied, in whole or in part, in software, hardware, firmware, or any other combination thereof. Furthermore, the data processing circuits may be a single, independent processing module, or may be integrated, in whole or in part, into any other element within the computer terminal 10 (or mobile device). As involved in the embodiments of this application, the data processing circuits serve as a processor control mechanism (e.g., selection of a variable resistor termination path connected to an interface).

[0036] The memory 104 can be used to store software programs and modules of application software, such as the program instructions / data storage device corresponding to the image-based transaction method in this embodiment. The processor 102 executes various functional applications and data processing by running the software programs and modules stored in the memory 104, thereby realizing the aforementioned image-based transaction method. The memory 104 may include high-speed random access memory, and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory 104 may further include memory remotely located relative to the processor 102, and these remote memories can be connected to the computer terminal 10 via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0037] The transmission device 106 is used to receive or send data via a network. Specific examples of the network described above may include a wireless network provided by the communication provider of the computer terminal 10. In one example, the transmission device 106 includes a network interface controller (NIC) and a network interface, which can be connected to other network devices via a base station to communicate with the Internet. In another example, the transmission device 106 may be a radio frequency (RF) module, used for wireless communication with the Internet.

[0038] The display can be, for example, a touchscreen liquid crystal display (LCD), which allows the user to interact with the user interface of the computer terminal 10 (or mobile device).

[0039] Under the aforementioned operating environment, this application provides the following: Figure 2 The image-based trading method shown. Figure 2 This is a flowchart of an image-based transaction method provided according to an embodiment of this application, such as... Figure 2 As shown, the method includes the following steps:

[0040] Step S201: Receive the image recognition command initiated by the target user through the client.

[0041] Specifically, the target user refers to the user who is about to execute a financial transaction, and the image recognition instruction is the instruction generated after the target user completes the facial information registration through the client and triggers the facial recognition operation. The client refers to the hardware device deployed in the transaction scenario, which has the ability to acquire images, communicate, and interact with the backend of financial institutions.

[0042] Step S202: With the authorization of the target user, obtain the user image of the target user based on the image recognition instruction.

[0043] Specifically, upon receiving an image recognition instruction, if the user is aware of and consents to the image acquisition, i.e., authorizes the acquisition, the user image of the target user can be obtained based on the image recognition instruction. The user image refers to an image captured by a camera that contains the facial area of ​​the target user.

[0044] Step S203: Input the user image into the image recognition model and process it to obtain the identity recognition result of the target user. The image recognition model includes a feature extraction module and a classification module. The image recognition model is used to extract features from the user image through the feature extraction module and to perform identity recognition through the classification module to obtain the identity recognition result.

[0045] Specifically, the image recognition model can refer to the EAMNet (ECA Attention And Multi-scale Feature Fusion Based VGG16 Network, a VGG16 network (Visual Geometry Group 16-layer Network) based on ECA (Efficient Channel Attention) and multi-scale feature fusion, which consists of a feature extraction module and a classification module. The feature extraction module is the part of the model responsible for extracting high-dimensional semantic features from pixel values. It includes the VGG_ECA module (VGG Module with ECA Attention, i.e., the feature extraction submodule) and the EMFM module (ECA Attention Multi-scale Fusion Module, i.e., the feature fusion submodule). The VGG_ECA module, composed of Cblock1-Cblock5, introduces the ECA attention mechanism to enhance important feature channels, uses the Mish activation function to optimize gradients, and adds Dropout after pooling to prevent overfitting. The EMFM module uses dilated convolutional layers with different dilation rates to extract multi-scale features with different receptive fields in parallel, and combines them with the ECA mechanism for feature fusion. The classification module is the part of the model that maps high-dimensional features to specific identity probabilities, including a global average pooling layer, a Dropout layer, a fully connected layer, and a Softmax classifier.

[0046] After obtaining the user's image, it can be input into the image recognition model mentioned above to obtain the identity recognition result. The identity recognition result refers to the user with the highest confidence level and its corresponding probability value in the probability distribution output by the model.

[0047] Step S204: Determine the transaction instruction based on the identity verification result.

[0048] Specifically, after obtaining the identity verification result, which includes a confidence score, a corresponding transaction instruction can be determined based on this result. The transaction instruction is the electronic instruction ultimately sent to the financial institution to execute account changes, thus transforming the technical aspect of facial recognition into financial authorization for payment. Only when the identity verification result is credible (confidence score above a threshold) and matches the corresponding user is a transaction instruction that allows the transaction generated; otherwise, a transaction instruction that disallows the transaction is generated, the transaction is rejected, and a retry is prompted.

[0049] It should be noted that after generating a transaction instruction that allows the transaction, the payment interface of the financial institution's backend can be called. After the backend verifies the validity of the instruction, the corresponding amount is deducted and a payment signal is returned to the client.

[0050] The image-based transaction method provided in this application receives an image recognition command initiated by a target user through a client; with the target user's authorization, it acquires the target user's image based on the image recognition command; it inputs the user image into an image recognition model to obtain the target user's identity recognition result. The image recognition model includes a feature extraction module and a classification module. The image recognition model extracts features from the user image through the feature extraction module and performs identity recognition through the classification module to obtain the identity recognition result; and it determines a transaction command based on the identity recognition result. This method solves the technical problem of low recognition accuracy and low transaction efficiency when recognizing facial images in related technologies. By inputting the user image into the image recognition model, processing it to obtain the target user's identity recognition result, and determining the transaction command based on the identity recognition result, it achieves the technical effect of improving the recognition accuracy when recognizing facial images.

[0051] Optionally, in the image-based transaction method provided in this application embodiment, obtaining the user image of the target user based on the image recognition instruction includes: calling the acquisition device to acquire the initial user image of the target user based on the image recognition instruction; performing grayscale conversion on the initial user image to obtain the converted initial user image; performing tone processing on the converted initial user image to obtain the processed initial user image; determining the first feature location information of the target feature in the processed initial user image; and cropping the processed initial user image based on the first feature location information to obtain the user image.

[0052] Specifically, when acquiring user images, the initial user image of the target user can first be acquired using a capture device. This capture device refers to a hardware device with image capture capabilities deployed in the transaction scenario, such as a checkout terminal or camera. The initial user image refers to a raw video frame or still photograph directly captured by the capture device without any preprocessing. Then, the initial user image is converted to grayscale to obtain the converted initial user image. Grayscale conversion involves converting a color image into a grayscale image that contains only brightness information and no color information, eliminating interference from color information in the color image, reducing computational complexity, and preparing for subsequent elimination of the effects of uneven lighting.

[0053] Furthermore, the converted initial user image undergoes tone processing to obtain a processed initial user image. This involves histogram equalization, which adjusts the image's grayscale distribution, expands the dynamic range, enhances contrast, and addresses uneven brightness in complex environments (such as areas that are too dark or too bright), improving the clarity of facial features and thus enhancing the accuracy of subsequent recognition. Next, the first feature location information of the target features in the processed initial user image is determined. The target features can refer to the face region in the image, and the first feature location information refers to the coordinates of the face bounding box calculated by a face detection algorithm (such as a Haar-like feature-based detector). This can include the coordinates of the top-left and bottom-right corners or the center point and its width and height. Finally, the processed initial user image is cropped based on the first feature location information. Specifically, the face region is extracted from the background according to the face bounding box coordinates in the first feature location information, thus obtaining the user image.

[0054] This embodiment effectively solves the problems of low image contrast, large background interference, and inconsistent size in complex lighting environments by a series of preprocessing steps such as grayscale conversion, tone processing (histogram equalization), feature localization, and cropping. This improves the data quality of the input image recognition model, thereby enhancing the recognition accuracy and robustness in complex scenes.

[0055] Optionally, in the image-based transaction method provided in this application embodiment, before inputting the user image into the image recognition model, the method further includes: performing image transformation on the user image to obtain a transformed user image, wherein the image transformation method includes at least: image rotation, image translation, and image flipping; determining the second feature position information of the target feature in the transformed user image; cropping the transformed user image based on the second feature position information to obtain a target image; and inputting the target image into the image recognition model to process and obtain an identity recognition result.

[0056] Specifically, after obtaining the user image, it is necessary to perform image transformation on the user image, that is, to modify the image data geometrically or spatially. For example, rotating it clockwise or counterclockwise by a certain angle (such as 45 degrees), moving it in the horizontal or vertical direction (translation), and performing mirroring operations along the axis of symmetry (horizontal flip or vertical flip) to obtain the transformed user image and prevent the model from overfitting.

[0057] Because the image has been rotated, translated, or flipped, the relative position of the face in the image has changed. It is necessary to accurately locate the transformed face region so that subsequent cropping can accurately capture the effective face portion and eliminate non-face background interference introduced by the transformation. At this time, it is necessary to determine the second feature position information of the target feature in the transformed user image. Here, the target feature refers to the face region in the transformed image, and the second feature position information refers to the face bounding box coordinates (such as the coordinates of the upper left corner, lower right corner, or center point and width and height) that are re-detected and calculated for the new image after image transformation, which are different from the first feature position information of the initial image.

[0058] Furthermore, the transformed user image is cropped based on the second feature location information, that is, redundant background pixels generated during the transformation operation are removed, and the format and size of the input data are standardized to obtain the target image. Finally, the target image is input into the image recognition model for processing to obtain the identity recognition result. For example, to enhance the robustness of the model, the user image is rotated 45 degrees clockwise to generate a tilted face image (the transformed user image). Then, the face region in the tilted image is re-detected to obtain new bounding box coordinates (second feature location information). Subsequently, the face portion in the tilted image is cropped according to these coordinates and scaled to a standard size to obtain the target image. Finally, the target image is input into the image recognition model, which extracts features and outputs the user's identity recognition result.

[0059] This embodiment effectively solves the recognition problems caused by changes in user facial posture, angle shifts, and partial occlusion by introducing data augmentation techniques such as image rotation, translation, and flipping, combined with repositioning and cropping steps. It significantly improves the image recognition model's adaptability to complex environments and special groups, thereby increasing the recognition success rate and stability under non-ideal conditions.

[0060] Optionally, in the image-based transaction method provided in this application embodiment, inputting the target image into an image recognition model and processing it to obtain an identity recognition result includes: inputting the target image into a feature extraction module, whereby the feature extraction module performs feature extraction processing on the target image to obtain processed image features; and inputting the processed image features into a classification module, whereby the classification module converts the processed image features to obtain an identity recognition result.

[0061] Specifically, after the image recognition model receives the target image, it can first input the target image into the feature extraction module, which will then perform feature extraction processing on the target image to obtain the processed image features. The processed image features refer to the high-dimensional semantic vector generated after the target image has undergone convolution, pooling, attention weighting, and multi-scale fusion.

[0062] Furthermore, the processed image features are input into the classification module, which transforms the high-dimensional abstract feature vectors to a specific user identity space. Fully connected layers perform feature dimensionality reduction and non-linear mapping, and the Softmax function outputs a probability distribution to determine the most probable user identity, thus obtaining the identity recognition result. For example, after the image recognition model receives the target image, it inputs the image into the feature extraction module, which sequentially passes it through five convolutional blocks to extract basic facial features. Then, dilated convolutions with dilation rates of 2, 3, and 4 are used in parallel to extract features at different scales and fuse them. Finally, global average pooling is used to obtain the processed image features. Next, the processed image features are input into the classification module, passing through a Dropout layer to prevent overfitting, and then mapped to the user dimension through a fully connected layer. Finally, the Softmax layer calculates the probability of each user, and the user with the highest probability is output as the identity recognition result.

[0063] This embodiment enhances robustness through a feature extraction module and then transforms high-dimensional features into specific user identity probabilities through a classification module, achieving efficient and accurate mapping from the original image to precise identity authentication, effectively improving recognition accuracy and security in complex scenarios.

[0064] Optionally, in the image-based transaction method provided in this application embodiment, the feature extraction module includes a feature extraction submodule and a feature fusion submodule. The feature extraction module performs feature extraction processing on the target image to obtain processed image features, including: inputting the target image into the feature extraction submodule, where the feature extraction submodule performs convolution processing on the target image to obtain first convolutional image features; performing attention enhancement processing on the first convolutional image features by the feature extraction submodule to obtain enhanced image features; performing sampling processing on the enhanced image features by the feature extraction submodule to obtain sampled image features; inputting the sampled image features into the feature fusion submodule, where M dilated convolution branches in the feature fusion submodule perform dilated convolution processing on the sampled image features to obtain M second convolutional image features, where M is a positive integer; and performing feature fusion on the M second convolutional image features by the feature fusion submodule to obtain processed image features.

[0065] It should be noted that the feature extraction submodule refers to the VGG_ECA module. This module consists of multiple convolutional blocks (Cblock1-Cblock5), including convolutional layers, BN (Batch Normalization) layers, Mish activation functions, ECA attention mechanisms, max pooling layers, and Dropout layers. The feature fusion submodule includes multiple parallel dilated convolutional branches, ordinary convolutional branches, and feature fusion structures.

[0066] After the feature extraction module receives the target image, it first performs convolution processing on the target image to obtain the first convolutional image features. The first convolutional image features refer to the intermediate feature map output after the target image has passed through the initial convolutional layer (such as Cblock1) in the feature extraction module. Then, the feature extraction module performs attention enhancement processing on the first convolutional image features. This involves compressing spatial information using global average pooling, calculating channel weights using adaptive one-dimensional convolution, and then multiplying these weights element-wise with the feature map. This enhances the focus on effective features (such as key facial structures) and suppresses background noise, outputting the enhanced image features. The attention enhancement processing refers to ECA attention mechanism processing. The enhanced image features are the feature map weighted by the ECA mechanism, where the responses of important channels (such as key facial structures) are enhanced, while less important channels are suppressed.

[0067] It should be noted that the attention enhancement process can be performed as follows: First, a global average pooling (GAP) operation is performed on the input image features X after the first convolution, converting the feature map size from [H, W, C] to [1, 1, C]. This compresses the three-dimensional information of each channel into a scalar, which is beneficial for capturing the contextual information between channels. Here, H and W are the height and width of the feature map, and C is the number of channels in the feature map. The GAP operation on the input feature map X is shown in the following equation: Then, based on the number of channels C of the feature map X', an adaptive one-dimensional convolution with a kernel size of K is calculated: Where γ=2 and b=1. Then, an adaptive one-dimensional convolution with kernel size K is used to convolve the globally average pooled features, and the weights MC for each channel are obtained using the Sigmoid activation function: Where σ represents the Sigmoid activation function, and C1DK represents an adaptive one-dimensional convolution with a kernel size of K. Finally, the weights MC(X') are multiplied element-wise with the input image features X after the first convolution to obtain the final output enhanced image features. : ,in, This is represented as element-wise multiplication.

[0068] Furthermore, the feature extraction submodule samples the enhanced image features to obtain sampled image features. This sampling process involves max-pooling, or regularization via a dropout layer. The sampled image features are those that, after max-pooling downsampling, have reduced spatial resolution but increased receptive field and more abstract semantic information. The sampled image features are then input into the feature fusion submodule, where multiple dilated convolution branches perform dilated convolution processing on the sampled image features to obtain the corresponding second-convolutional image features. These second-convolutional image features are multi-scale feature maps with different receptive fields extracted through dilated convolutions with different dilation rates.

[0069] It should be noted that dilated convolution branches refer to multiple parallel dilated convolutional layers, for example, M=3, corresponding to dilated convolutional layers with dilation rates of 2, 3, and 4 respectively; however, it can also include ordinary convolutional branches. Dilated convolution processing refers to using convolutional kernels with dilation rates to perform convolution operations, allowing the kernels to cover a larger receptive field without increasing the number of parameters or reducing resolution. The formula for calculating the receptive field of a dilated convolutional layer is as follows: ; ; Where Ksize represents the kernel of the dilated convolutional layer, d represents the dilation rate of the dilated convolutional layer, K1 represents the kernel size of the ordinary convolution with the same receptive field as the dilated convolutional layer, RFn+1 represents the receptive field size of the current layer of the dilated convolution, RFn represents the receptive field size of the layer before the dilated convolution, and Sn represents the product of the strides of all previous layers.

[0070] As shown in the formula above, the receptive field of a dilated convolutional layer is related to its kernel, stride, and dilation rate. Therefore, the feature fusion submodule can consist of three dilated convolutional layers, one regular convolutional layer, one ECA attention mechanism, and two max-pooling layers. The dilation rates of the three dilated convolutional layers are set to 2, 3, and 4, respectively, as are the paddings, and the stride is set to 1 for each layer, to gradually increase the receptive field of these dilated convolutional layers while maintaining the same input and output feature map size. A BN layer and a Mish activation function are then added after each dilated convolutional layer and the regular convolutional layer to accelerate model convergence.

[0071] Furthermore, the feature fusion submodule performs feature fusion on the image features after the second convolution, uses the ECA attention mechanism for further weighting, and performs fusion by adding or concatenating elements to obtain the processed image features after multi-scale fusion.

[0072] This embodiment effectively solves the problem of insufficient feature extraction in traditional convolutional neural networks under complex lighting, facial occlusion and pose change scenarios by performing channel-level feature enhancement and using multi-branch dilated convolution for multi-scale spatial feature fusion, and significantly improves the recognition accuracy and robustness in non-ideal environments.

[0073] Optionally, in the image-based transaction method provided in this application embodiment, the classification module transforms the processed image features to obtain the identity recognition result, including: pooling the processed image features to obtain an image feature vector; transforming the image feature vector to obtain a transformed feature vector, and regularizing the transformed feature vector to obtain a processed feature vector; mapping the processed feature vector to a preset category space to obtain a mapping result, wherein the preset category space refers to the image information of users stored in a preset institution, and the mapping result is used to indicate the degree of matching between the processed feature vector and the image information of users stored in the preset institution; converting the mapping result into a mapping probability distribution, wherein the mapping probability distribution refers to the similarity probability distribution between the processed feature vector and the image information of users stored in the preset institution; extracting the highest probability value from the mapping probability distribution, and generating the identity recognition result based on the user information of the user corresponding to the highest probability value.

[0074] Specifically, after obtaining the processed image features, the classification module can first perform pooling on the processed image features, compressing them into a 1×1×512 feature vector, which greatly reduces the number of parameters, thus obtaining the image feature vector. Pooling refers to the global averaging of the two-dimensional or three-dimensional feature map (such as 3×3×512) on each channel, compressing the spatial dimension. The image feature vector is the one-dimensional vector (such as 1×1×512) obtained after global average pooling. It retains the global semantic information in the original feature map, but removes the spatial location information.

[0075] Furthermore, the image feature vector is dimensionally transformed to obtain a transformed feature vector, and then regularized to obtain a processed feature vector. In other words, the multidimensional tensor is flattened into a one-dimensional array, and then passed through a Dropout layer and a fully connected layer to obtain the processed feature vector, thereby improving the model's generalization ability on unknown data.

[0076] The processed feature vectors are then mapped to a predefined category space. This is achieved by projecting the processed feature vectors into a space with the same dimension as the predefined category space using a fully connected layer. The similarity score between the input features and the features of each predefined user is calculated to obtain the mapping result. The predefined category space refers to a pre-entered set of facial feature databases of all authorized users, with each user corresponding to a specific category in this space. The mapping result is then normalized using a Softmax classifier to obtain a mapping probability distribution. This distribution calculates the probability of each registered user using the Softmax function, resulting in a set of raw scores. Each score represents the probability that the current input image matches a specific user in the database. These scores are then converted into a probability distribution with a sum of 1, where each probability value represents the probability that the current image belongs to the corresponding user. Finally, the highest probability value is extracted from the mapping probability distribution, and the user information corresponding to the highest probability value is used to generate the identity recognition result.

[0077] It should be noted that the above model can dynamically adjust the learning rate during training using a cosine annealing decay strategy, allowing the model to converge to the global optimum and thus achieve higher accuracy during the inference phase. The cosine annealing decay strategy for adjusting the learning rate is obtained using the following formula: ηt represents the learning rate at the t-th iteration, ηmax represents the maximum learning rate, which is usually the initial learning rate, ηmin represents the minimum learning rate to prevent the learning rate from decreasing indefinitely, Tcur represents the number of iterations in the current cycle, and Tmax represents the total number of iterations in a complete cycle.

[0078] This embodiment achieves automated determination of user identity from abstract features through steps such as global average pooling, regularization, fully connected mapping, and Softmax probability transformation, effectively reducing the number of model parameters and improving the accuracy and robustness of classification.

[0079] Optionally, in the image-based transaction method provided in this application embodiment, determining the transaction instruction based on the identity recognition result includes: generating a transaction instruction that allows the transaction if the identity recognition result indicates that the target user's identity verification has passed; and generating a transaction instruction that disallows the transaction if the identity recognition result indicates that the target user's identity verification has failed.

[0080] Specifically, the identity recognition result refers to the user's identity and corresponding confidence probability output by the image recognition model. If the identity recognition result indicates that the target user's identity verification is successful, that is, the currently identified target user's identity is consistent with the user identity bound to the current payment order, and the recognition confidence is higher than the preset security threshold (e.g., 90%), the current operator is confirmed as an authorized user, and a transaction instruction allowing the transaction can be generated. Conversely, if the identity recognition result indicates that the target user's identity verification is unsuccessful, that is, the currently identified user's identity is inconsistent with the user bound to the order, or the recognition confidence is lower than the preset threshold (e.g., lower than 80%), and the user cannot be confirmed as an authorized user, or abnormal situations such as multiple faces or no face are detected, a transaction instruction disallowing the transaction is generated.

[0081] This embodiment achieves secure linkage between identity authentication and payment authorization by setting clear standards for whether the verification passes or fails, effectively preventing transaction risks caused by identification errors or identity theft.

[0082] This application also provides an image-based transaction method. Figure 3 This is a schematic diagram of an optional image-based transaction method provided according to an embodiment of this application, such as... Figure 3 As shown, the method includes:

[0083] When a target user needs to conduct a financial transaction for an order, image capture can be performed first, that is, facial information can be entered through the client to obtain the user's image. Then, after the order is confirmed, user image recognition is performed. Figure 4 This is a schematic diagram of a user image recognition method provided according to an embodiment of this application, such as... Figure 4 As shown, after obtaining the user's image, image detection is performed on it. Figure 5 This is a schematic diagram of the image detection method provided according to the embodiments of this application, such as... Figure 5 As shown, a face detector can be loaded, and the image can be sequentially processed with grayscale conversion, histogram equalization, and cropping. Then, image preprocessing and feature extraction are performed on the detected image to obtain an image for identity verification. The user's identity is then verified based on the recognition result. If verification is successful, order payment is processed and a receipt is issued; otherwise, user image recognition needs to be performed again.

[0084] It should be noted that, Figure 6 This is a schematic diagram of an image preprocessing method provided according to an embodiment of this application, such as... Figure 6As shown, image preprocessing can be data augmentation. After obtaining the augmented image, faces are detected, and rectangular bounding boxes are drawn on the image for subsequent recognition. Finally, the display result is generated. Data augmentation can include 45-degree clockwise rotation, 45-degree counterclockwise rotation, translation, horizontal flipping, and vertical flipping.

[0085] This embodiment improves the accuracy of facial recognition by inputting user images into an image recognition model, processing the images to obtain the target user's identity recognition result, and determining transaction instructions based on the identity recognition result.

[0086] It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases the steps shown or described may be executed in a different order than that shown here.

[0087] Example 2

[0088] This application also provides an image-based trading device. It should be noted that the image-based trading device of this application can be used to execute the image-based trading method provided in this application. The image-based trading device provided in this application will be described below.

[0089] According to an embodiment of this application, an apparatus for implementing the above-described image-based transaction method is also provided. Figure 7 This is a schematic diagram of an image-based transaction device provided according to an embodiment of this application, such as... Figure 7 As shown, the device includes: a receiving unit 70, an acquiring unit 71, a first input unit 72, and a first determining unit 73.

[0090] The receiving unit 70 is used to receive image recognition instructions initiated by the target user through the client;

[0091] The acquisition unit 71 is used to acquire the user image of the target user based on an image recognition instruction, provided that the target user authorizes it.

[0092] The first input unit 72 is used to input the user image into the image recognition model and process it to obtain the identity recognition result of the target user. The image recognition model includes a feature extraction module and a classification module. The image recognition model is used to extract features from the user image through the feature extraction module and perform identity recognition through the classification module to obtain the identity recognition result.

[0093] The first determining unit 73 is used to determine the transaction instruction based on the identity recognition result.

[0094] The image-based transaction device provided in this application embodiment receives an image recognition command initiated by a target user through a client via a receiving unit 70; an acquisition unit 71, with the authorization of the target user, acquires the user image of the target user based on the image recognition command; a first input unit 72 inputs the user image into an image recognition model, processes it to obtain the target user's identity recognition result, wherein the image recognition model includes a feature extraction module and a classification module, the image recognition model is used to extract features from the user image through the feature extraction module and to perform identity recognition through the classification module to obtain the identity recognition result; a first determination unit 73 determines the transaction command based on the identity recognition result, thereby solving the technical problem of low recognition accuracy and low transaction efficiency when recognizing face images in related technologies. By inputting the user image into the image recognition model, processing it to obtain the target user's identity recognition result, and determining the transaction command based on the identity recognition result, the technical effect of improving the recognition accuracy when recognizing face images is achieved.

[0095] Optionally, in the image-based transaction device provided in this application embodiment, the acquisition unit 71 includes: an acquisition module, used to call an acquisition device to acquire an initial user image of the target user based on an image recognition instruction; a first conversion module, used to perform grayscale conversion on the initial user image to obtain a converted initial user image; a first processing module, used to perform tone processing on the converted initial user image to obtain a processed initial user image; a determination module, used to determine the first feature location information of the target feature in the processed initial user image; and a cropping module, used to crop the processed initial user image based on the first feature location information to obtain a user image.

[0096] Optionally, in the image-based transaction apparatus provided in this application embodiment, the apparatus further includes: a transformation unit, used to transform the user image before inputting the user image into the image recognition model to obtain a transformed user image, wherein the image transformation method includes at least: image rotation, image translation, and image flipping; a second determination unit, used to determine the second feature position information of the target feature in the transformed user image; a cropping unit, used to crop the transformed user image based on the second feature position information to obtain a target image; and a second input unit, used to input the target image into the image recognition model and process it to obtain an identity recognition result.

[0097] Optionally, in the image-based transaction device provided in this application embodiment, the device includes: a first input module for inputting a target image into a feature extraction module, wherein the feature extraction module performs feature extraction processing on the target image to obtain processed image features; and a second input module for inputting the processed image features into a classification module, wherein the classification module converts the processed image features to obtain an identity recognition result.

[0098] Optionally, in the image-based transaction apparatus provided in this application embodiment, the apparatus includes: a third input module, used to input a target image into a feature extraction submodule, where the feature extraction submodule performs convolution processing on the target image to obtain first convolutional image features; a second processing module, used to perform attention enhancement processing on the first convolutional image features by the feature extraction submodule to obtain enhanced image features; a third processing module, used to perform sampling processing on the enhanced image features by the feature extraction submodule to obtain sampled image features; a fourth input module, used to input the sampled image features into a feature fusion submodule, where M dilated convolution branches in the feature fusion submodule perform dilated convolution processing on the sampled image features to obtain M second convolutional image features, where M is a positive integer; and a fusion module, used to perform feature fusion on the M second convolutional image features by the feature fusion submodule to obtain processed image features.

[0099] Optionally, in the image-based transaction device provided in this application embodiment, the device includes: a fourth processing module, used to perform pooling processing on the processed image features by the classification module to obtain an image feature vector; a second conversion module, used to perform dimensionality conversion on the image feature vector to obtain a converted feature vector, and to perform regularization processing on the converted feature vector to obtain a processed feature vector; a mapping module, used to map the processed feature vector to a preset category space to obtain a mapping result, wherein the preset category space refers to the image information of users stored in a preset institution, and the mapping result is used to indicate the degree of matching between the processed feature vector and the image information of users stored in the preset institution; a third conversion module, used to convert the mapping result into a mapping probability distribution, wherein the mapping probability distribution refers to the similarity probability distribution between the processed feature vector and the image information of users stored in the preset institution; and an extraction module, used to extract the highest probability value from the mapping probability distribution, and generate an identity recognition result based on the user information of the user corresponding to the highest probability value.

[0100] Optionally, in the image-based transaction device provided in this application embodiment, the first determining unit 73 includes: a first generating module, configured to generate a transaction instruction that allows the transaction when the identity recognition result indicates that the target user's identity verification has passed; and a second generating module, configured to generate a transaction instruction that disallows the transaction when the identity recognition result indicates that the target user's identity verification has failed.

[0101] It should be noted that the receiving unit 70, acquiring unit 71, first input unit 72, and first determining unit 73 mentioned above correspond to steps S201 to S204 in Embodiment 1. The instances and application scenarios implemented by the above units and corresponding steps are the same, but are not limited to the content disclosed in Embodiment 1. It should be noted that the above modules or units can be hardware or software components stored in memory (e.g., memory 104) and processed by one or more processors (e.g., processors 102a, 102b, ..., 102n). The above units can also be part of a device and run in the computer terminal 10 provided in Embodiment 1.

[0102] Example 3

[0103] Embodiments of this application may provide a computer terminal, which may be any computer terminal device in a group of computer terminals. Optionally, in this embodiment, the aforementioned computer terminal may also be replaced with a mobile terminal or an electronic device, etc.

[0104] Optionally, in this embodiment, the computer terminal may be located in at least one of a plurality of network devices in a computer network.

[0105] In this embodiment, the computer terminal described above can execute the program code for the following steps in the image-based transaction method: receiving an image recognition instruction initiated by the target user through a client; obtaining the target user's user image based on the image recognition instruction, with the target user's authorization; inputting the user image into an image recognition model and processing it to obtain the target user's identity recognition result, wherein the image recognition model includes a feature extraction module and a classification module, the image recognition model is used to extract features from the user image through the feature extraction module and perform identity recognition through the classification module to obtain the identity recognition result; and determining the transaction instruction based on the identity recognition result.

[0106] Optionally, the aforementioned computer terminal may execute program code for the following steps in the image-based transaction method: invoking an acquisition device to acquire an initial user image of the target user based on an image recognition instruction; performing grayscale conversion on the initial user image to obtain a converted initial user image; performing tone processing on the converted initial user image to obtain a processed initial user image; determining the first feature location information of the target feature in the processed initial user image; and cropping the processed initial user image based on the first feature location information to obtain a user image.

[0107] Optionally, the aforementioned computer terminal may execute program code for the following steps in the image-based transaction method: performing image transformation on the user image to obtain a transformed user image, wherein the image transformation methods include at least: image rotation, image translation, and image flipping; determining the second feature position information of the target feature in the transformed user image; cropping the transformed user image based on the second feature position information to obtain a target image; and inputting the target image into an image recognition model to process and obtain an identity recognition result.

[0108] Optionally, the aforementioned computer terminal can execute program code for the following steps in the image-based transaction method: inputting the target image into the feature extraction module, whereby the feature extraction module performs feature extraction processing on the target image to obtain processed image features; inputting the processed image features into the classification module, whereby the classification module converts the processed image features to obtain the identity recognition result.

[0109] Optionally, the aforementioned computer terminal can execute program code for the following steps in the image-based transaction method: inputting the target image into the feature extraction submodule, whereby the feature extraction submodule performs convolution processing on the target image to obtain image features after the first convolution; performing attention enhancement processing on the image features after the first convolution by the feature extraction submodule to obtain enhanced image features; performing sampling processing on the enhanced image features by the feature extraction submodule to obtain sampled image features; inputting the sampled image features into the feature fusion submodule, whereby the M dilated convolution branches in the feature fusion submodule perform dilated convolution processing on the sampled image features to obtain M image features after the second convolution, where M is a positive integer; and performing feature fusion on the M image features after the second convolution by the feature fusion submodule to obtain processed image features.

[0110] Optionally, the aforementioned computer terminal can execute the program code for the following steps in the image-based transaction method: The classification module performs pooling processing on the processed image features to obtain an image feature vector; the image feature vector undergoes dimensionality transformation to obtain a transformed feature vector, and the transformed feature vector is regularized to obtain a processed feature vector; the processed feature vector is mapped to a preset category space to obtain a mapping result, where the preset category space refers to the image information of users stored in a preset institution, and the mapping result is used to indicate the degree of matching between the processed feature vector and the image information of users stored in the preset institution; the mapping result is converted into a mapping probability distribution, where the mapping probability distribution refers to the similarity probability distribution between the processed feature vector and the image information of users stored in the preset institution; the highest probability value is extracted from the mapping probability distribution, and an identity recognition result is generated based on the user information of the user corresponding to the highest probability value.

[0111] Optionally, the aforementioned computer terminal may execute program code for the following steps in the image-based transaction method: generating a transaction instruction that allows the transaction if the identity verification result indicates that the target user's identity has been verified; and generating a transaction instruction that disallows the transaction if the identity verification result indicates that the target user's identity has not been verified.

[0112] Optionally, Figure 8 This is a structural block diagram of an electronic device according to an embodiment of this application. Figure 8 As shown, the electronic device may include: one or more ( Figure 8 (Only one is shown) processor 802, memory 804, memory controller, and peripheral interface, wherein the peripheral interface is connected to the radio frequency module, audio module and display.

[0113] The memory can be used to store software programs and modules, such as the program instructions / modules corresponding to the image-based transaction method and apparatus in this application embodiment. The processor executes various functional applications and data processing by running the software programs and modules stored in the memory, thereby realizing the aforementioned image-based transaction method. The memory may include high-speed random access memory, and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory may further include memory remotely located relative to the processor, and these remote memories can be connected to the terminal via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0114] The processor can access the information and application programs stored in the memory via the transmission device to execute the steps described above in the image-based transaction method.

[0115] Those skilled in the art will understand that Figure 8 The structure shown is for illustrative purposes only. Electronic devices can also be smartphones, tablets, handheld computers, mobile internet devices (MIDs), PADs, and other terminal devices. Figure 8 This does not limit the structure of the aforementioned electronic device. For example, electronic devices may also include components that are more... Figure 8 The more or fewer components shown (such as network interfaces, display devices, etc.), or having the same Figure 8 The different configurations shown.

[0116] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be implemented by a program instructing the hardware related to the terminal device. The program can be stored in a computer-readable storage medium, which may include: flash drive, read-only memory (ROM), random access memory (RAM), disk or optical disk, etc.

[0117] Example 4

[0118] Embodiments of this application also provide a storage medium. Optionally, in this embodiment, the storage medium can be used to store the program code executed by the image-based transaction method provided in Embodiment 1.

[0119] Optionally, in this embodiment, the storage medium may be located in any computer terminal in a group of computer terminals in a computer network, or in any mobile terminal in a group of mobile terminals.

[0120] Optionally, in this embodiment, the storage medium is configured to store program code for performing the following steps: receiving an image recognition instruction initiated by the target user through a client; obtaining the target user's user image based on the image recognition instruction, provided the target user authorizes the process; inputting the user image into an image recognition model and processing it to obtain the target user's identity recognition result, wherein the image recognition model includes a feature extraction module and a classification module, the image recognition model being used to extract features from the user image through the feature extraction module and to perform identity recognition through the classification module to obtain the identity recognition result; and determining a transaction instruction based on the identity recognition result.

[0121] This application also provides a computer program product that, when executed on a data processing device, is suitable for performing image-based transaction method steps.

[0122] The sequence numbers of the embodiments in this application are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0123] In the above embodiments of this application, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0124] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. The device embodiments described above are merely illustrative; for example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection of units or modules may be electrical or other forms.

[0125] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0126] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0127] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as a USB flash drive, read-only memory (ROM), random access memory (RAM), portable hard drive, magnetic disk, or optical disk.

[0128] The above description is only a preferred embodiment of this application. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of this application, and these improvements and modifications should also be considered within the scope of protection of this application.

Claims

1. An image-based transaction method, characterized in that, include: Receive image recognition commands initiated by the target user through the client; With the authorization of the target user, the user image of the target user is obtained based on the image recognition instruction; The user image is input into an image recognition model and processed to obtain the identity recognition result of the target user. The image recognition model includes a feature extraction module and a classification module. The image recognition model is used to extract features from the user image through the feature extraction module and to perform identity recognition through the classification module to obtain the identity recognition result. The transaction instruction is determined based on the identity verification result.

2. The method according to claim 1, characterized in that, Obtaining the user image of the target user based on the image recognition instruction includes: Based on the image recognition command, the acquisition device is invoked to acquire the initial user image of the target user; The initial user image is converted to grayscale to obtain the converted initial user image; The converted initial user image is subjected to tone processing to obtain the processed initial user image; Determine the first feature location information of the target feature in the processed initial user image; The processed initial user image is cropped based on the first feature location information to obtain the user image.

3. The method according to claim 1, characterized in that, Before inputting the user image into the image recognition model, the method further includes: The user image is transformed to obtain a transformed user image, wherein the image transformation methods include at least: image rotation, image translation, and image flipping; Determine the second feature location information of the target feature in the transformed user image; The transformed user image is cropped based on the second feature location information to obtain the target image; The target image is input into the image recognition model, and the identity recognition result is obtained through processing.

4. The method according to claim 3, characterized in that, The target image is input into the image recognition model, and the identity recognition result is obtained by processing it, including: The target image is input into the feature extraction module, which performs feature extraction processing on the target image to obtain the processed image features. The processed image features are input into the classification module, which then transforms the processed image features to obtain the identity recognition result.

5. The method according to claim 4, characterized in that, The feature extraction module includes a feature extraction submodule and a feature fusion submodule. The feature extraction module performs feature extraction processing on the target image to obtain processed image features, including: The target image is input into the feature extraction submodule, which performs convolution processing on the target image to obtain the image features after the first convolution. The feature extraction submodule performs attention enhancement processing on the image features after the first convolution to obtain enhanced image features; The feature extraction submodule performs sampling processing on the enhanced image features to obtain sampled image features; The sampled image features are input into the feature fusion submodule, and the sampled image features are subjected to dilated convolution processing by M dilated convolution branches in the feature fusion submodule to obtain M second convolution image features, where M is a positive integer; The feature fusion submodule performs feature fusion on the M second convolutional image features to obtain the processed image features.

6. The method according to claim 4, characterized in that, The classification module transforms the processed image features to obtain the identity recognition result, including: The classification module performs pooling on the processed image features to obtain an image feature vector; The image feature vector is dimensionally transformed to obtain a transformed feature vector, and the transformed feature vector is then regularized to obtain a processed feature vector. The processed feature vector is mapped to a preset category space to obtain a mapping result, wherein the preset category space refers to the image information of the user stored in the preset institution, and the mapping result is used to indicate the degree of matching between the processed feature vector and the image information of the user stored in the preset institution; The mapping result is converted into a mapping probability distribution, wherein the mapping probability distribution refers to the similarity probability distribution between the processed feature vector and the user's image information stored in the preset mechanism; The highest probability value is extracted from the mapping probability distribution, and the identity recognition result is generated based on the user information of the user corresponding to the highest probability value.

7. The method according to claim 1, characterized in that, Determining transaction instructions based on the identity verification results includes: If the identity verification result indicates that the target user's identity has been verified, a transaction instruction allowing the transaction is generated; If the identity verification result indicates that the target user's identity verification has failed, a transaction instruction that disallows the transaction is generated.

8. An image-based transaction device, characterized in that, include: The receiving unit is used to receive image recognition commands initiated by the target user through the client. The acquisition unit is configured to acquire a user image of the target user based on the image recognition instruction, provided that the target user authorizes the acquisition. The first input unit is used to input the user image into an image recognition model and process it to obtain the identity recognition result of the target user. The image recognition model includes a feature extraction module and a classification module. The image recognition model is used to extract features from the user image through the feature extraction module and perform identity recognition through the classification module to obtain the identity recognition result. The first determining unit is used to determine the transaction instruction based on the identity recognition result.

9. An electronic device, characterized in that, include: Memory, which stores executable programs; A processor for running the program, wherein the program, when running, performs the image-based transaction method according to any one of claims 1 to 7.

10. A computer program product comprising computer instructions, characterized in that, When the computer instructions are executed by the processor, they implement the steps of the image-based transaction method according to any one of claims 1 to 7.

Images