Risk account processing method and device, electronic equipment and storage medium

Abstract

The invention discloses a risk account processing method and device, electronic equipment and a storage medium, and relates to the technical field of data processing, and the method comprises the steps: responding to a risk account to log in an automatic teller machine, and obtaining a face image of an operator of the risk account collected by the automatic teller machine; inputting the face image into a preset face expression recognition model for expression recognition to obtain an expression recognition result of the face image; if it is determined that the expression recognition result is the first emotional expression or the second emotional expression, it is determined that the transaction behavior of the risk account is abnormal, and early warning information is output. Compared with the prior art, the automatic teller machine is used for collecting the face image of the operator of the risk account for expression recognition, if the expression recognition result is the first emotional expression or the second emotional expression, it is indicated that the emotion of the operator of the risk account is abnormal, and the emotion of the operator is closely related to the safety of the transaction behavior; whether the transaction behavior is abnormal or not can be determined through the expression recognition result, so that the security of the risk account is improved.

Description

Technical Field

[0001] The present disclosure relates to the field of data processing technology, and in particular to a method and device for processing risky accounts, an electronic device, and a storage medium. Background Art

[0002] With the digital development of banking services, automated teller machines (ATMs) have become an important tool for people's daily financial transactions. However, criminals often use risky accounts to conduct illegal transactions on ATMs, which brings huge risks to banks and users. Risky accounts refer to accounts with credit risks, abnormal transactions, or security loopholes.

[0003] In the related technologies for processing risk accounts, usually only risk accounts with higher risk levels are restricted in some transaction functions, and risk accounts with lower risk levels are not processed at all, resulting in lower security for risk accounts with lower risk levels. Summary of the invention

[0004] The present disclosure provides a method and device for processing risky accounts, an electronic device and a storage medium, the main purpose of which is to solve the problem of low security of risky accounts with low risk levels.

[0005] According to a first aspect of the present disclosure, a method for processing a risk account is provided, which includes:

[0006] In response to a risk account logging into an ATM, acquiring a facial image of an operator of the risk account captured by the ATM;

[0007] Inputting the facial image into a preset facial expression recognition model to perform expression recognition, and obtaining an expression recognition result of the facial image;

[0008] If it is determined that the expression recognition result is the first emotional expression or the second emotional expression, it is determined that the transaction behavior of the risk account is abnormal, and a warning message is output.

[0009] Optionally, after inputting the facial image into a preset facial expression recognition model for expression recognition to obtain an expression recognition result of the facial image, the method further includes:

[0010] If it is determined that the expression recognition result is the third emotional expression or the fourth emotional expression, it is determined that the transaction behavior of the risk account is normal.

[0011] Optionally, the method further includes:

[0012] Acquire a first training emotional expression facial image set, a second training emotional expression facial image set, a third training emotional expression facial image set, and a fourth training emotional expression facial image set; the first training emotional expression facial image set includes a first training facial image and a corresponding first emotional expression label, the second training emotional expression facial image set includes a second training facial image and a corresponding second emotional expression label, the third training emotional expression facial image set includes a third training facial image and a corresponding third emotional expression label, and the fourth training emotional expression facial image set includes a fourth training facial image and a corresponding fourth emotional expression label;

[0013] The initial facial expression recognition model is trained using the first training emotional expression facial image set, the second training emotional expression facial image set, the third training emotional expression facial image set and the fourth training emotional expression facial image set to obtain the preset facial expression recognition model.

[0014] Optionally, the obtaining of the first training emotional expression facial image set, the second training emotional expression facial image set, the third training emotional expression facial image set, and the fourth training emotional expression facial image set comprises:

[0015] Acquire at least five fifth training emotional expression facial image sets; different fifth training emotional expression facial image sets in the at least five fifth training emotional expression facial image sets correspond to training emotional expression facial images of different emotional expression types, and each of the fifth training emotional expression facial image sets includes a fifth training emotional expression image and a corresponding fifth emotional expression label;

[0016] Based on the first similarity between each of the fifth emotional expression labels and the first emotional expression label, the fifth training emotional expression face image set of the fifth emotional expression labels whose first similarity is greater than a preset threshold is combined into the training first emotional expression face image set;

[0017] Based on the second similarity between each of the fifth emotional expression labels and the second emotional expression label, the fifth training emotional expression face image set of the fifth emotional expression labels whose second similarity is greater than the preset threshold is used to form the second training emotional expression face image set;

[0018] Based on the third similarity between each of the fifth emotional expression labels and the third emotional expression label, the fifth training emotional expression face image set of the fifth emotional expression labels whose third similarity is greater than the preset threshold is used to form the third training emotional expression face image set;

[0019] Based on the fourth similarity between each of the fifth emotional expression labels and the fourth emotional expression label, the fifth training emotional expression facial image set of the fourth emotional expression labels whose fourth similarity is greater than the preset threshold is combined into the training fourth emotional expression facial image set.

[0020] Optionally, after obtaining at least five fifth training emotional expression face image sets, the method further includes:

[0021] Performing facial masking processing on all fifth training emotional expression images in the at least five fifth training emotional expression face image sets to obtain masked fifth training emotional expression images;

[0022] The fifth training emotional expression image includes the masked fifth training emotional expression image.

[0023] Optionally, the preset facial expression recognition model includes two separable convolutional networks with the same step size; the separable convolutional networks are used to perform expression recognition on the facial image.

[0024] According to a second aspect of the present disclosure, a device for processing a risk account is provided, comprising:

[0025] A first acquisition unit is used for acquiring a face image of an operator of the risky account collected by the automatic teller machine in response to the risky account logging into the automatic teller machine;

[0026] A recognition unit, used for inputting the face image into a preset face expression recognition model to perform expression recognition, and obtaining an expression recognition result of the face image;

[0027] The determination unit is used to determine that the transaction behavior of the risk account is abnormal and output warning information when it is determined that the expression recognition result is the first emotional expression or the second emotional expression.

[0028] Optionally, the device further comprises:

[0029] The determination unit is also used to, after inputting the facial image into a preset facial expression recognition model for expression recognition and obtaining an expression recognition result of the facial image, determine that the transaction behavior of the risk account is normal when it is determined that the expression recognition result is a third emotional expression or a fourth emotional expression.

[0030] Optionally, the device further comprises:

[0031] A second acquisition unit is used to acquire a first emotional expression face image set for training, a second emotional expression face image set for training, a third emotional expression face image set for training, and a fourth emotional expression face image set for training; the first emotional expression face image set for training includes a first training face image and a corresponding first emotional expression label, the second emotional expression face image set for training includes a second training face image and a corresponding second emotional expression label, the third emotional expression face image set for training includes a third training face image and a corresponding third emotional expression label, and the fourth emotional expression face image set for training includes a fourth training face image and a corresponding fourth emotional expression label;

[0032] The training unit is used to train the initial facial expression recognition model using the first training emotional expression facial image set, the second training emotional expression facial image set, the third training emotional expression facial image set and the fourth training emotional expression facial image set to obtain the preset facial expression recognition model.

[0033] Optionally, the second acquiring unit includes:

[0034] An acquisition module is used to acquire at least five fifth training emotional expression facial image sets; different fifth training emotional expression facial image sets in the at least five fifth training emotional expression facial image sets correspond to training emotional expression facial images of different emotional expression types, and each of the fifth training emotional expression facial image sets includes a fifth training emotional expression image and a corresponding fifth emotional expression label;

[0035] A composition module, configured to, based on a first similarity between each of the fifth emotional expression labels and the first emotional expression label, respectively, compose the training first emotional expression facial image set from a fifth training emotional expression facial image set of the fifth emotional expression labels whose first similarity is greater than a preset threshold;

[0036] The composition module is also used to, based on the second similarity between each of the fifth emotional expression labels and the second emotional expression label, form the fifth training emotional expression face image set of the fifth emotional expression label whose second similarity is greater than the preset threshold into the second training emotional expression face image set;

[0037] The composition module is also used to, based on the third similarity between each of the fifth emotional expression labels and the third emotional expression label, form the fifth training emotional expression face image set of the fifth emotional expression labels whose third similarity is greater than the preset threshold into the third training emotional expression face image set;

[0038] The composition module is also used to, based on the fourth similarity between each of the fifth emotional expression labels and the fourth emotional expression label, form the fifth training emotional expression facial image set of the fourth emotional expression labels whose fourth similarity is greater than the preset threshold into the fourth training emotional expression facial image set.

[0039] Optionally, the device further comprises:

[0040] A processing unit is used for, after acquiring at least five fifth training emotional expression face image sets, performing facial masking processing on all fifth training emotional expression images in the at least five fifth training emotional expression face image sets to obtain masked fifth training emotional expression images;

[0041] The fifth training emotional expression image includes the masked fifth training emotional expression image.

[0042] Optionally, the preset facial expression recognition model includes two separable convolutional networks with the same step size; the separable convolutional networks are used to perform expression recognition on the facial image.

[0043] According to a third aspect of the present disclosure, there is provided an electronic device, including:

[0044] at least one processor; and

[0045] a memory communicatively connected to the at least one processor; wherein,

[0046] The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method described in the first aspect.

[0047] According to a fourth aspect of the present disclosure, a non-transitory computer-readable storage medium storing computer instructions is provided, wherein the computer instructions are used to enable the computer to execute the method described in the first aspect.

[0048] According to a fifth aspect of the present disclosure, a computer program product is provided, comprising a computer program, wherein when the computer program is executed by a processor, the computer program implements the method as described in the first aspect above.

[0049] The risk account processing method and device, electronic device and storage medium provided by the present disclosure, in response to the risk account logging into the ATM, obtains the facial image of the operator of the risk account collected by the ATM; inputs the facial image into a preset facial expression recognition model for expression recognition, and obtains the expression recognition result of the facial image; if it is determined that the expression recognition result is the first emotional expression or the second emotional expression, it is determined that the transaction behavior of the risk account is abnormal, and outputs warning information. Compared with the related art, the embodiment of the present disclosure uses the facial image of the operator of the risk account collected by the ATM to perform expression recognition. If the expression recognition result is the first emotional expression or the second emotional expression, it indicates that the emotion of the operator of the risk account is abnormal. Since the operator's emotion is closely related to the security of the transaction behavior, the expression recognition result can be used to determine whether the transaction behavior is abnormal, thereby improving the security of the risk account.

[0050] It should be understood that the content described in this section is not intended to identify the key or important features of the embodiments of the present application, nor is it intended to limit the scope of the present application. Other features of the present application will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] The accompanying drawings are used to better understand the present solution and do not constitute a limitation of the present disclosure.

[0052] Figure 1 A flowchart of a method for processing a risk account provided by an embodiment of the present disclosure;

[0053] Figure 2 A flowchart of a method for training a preset facial expression recognition model provided by an embodiment of the present disclosure;

[0054] Figure 3 A schematic diagram of the structure of a risk account processing device provided by an embodiment of the present disclosure;

[0055] Figure 4 A schematic diagram of the structure of another risk account processing device provided by an embodiment of the present disclosure;

[0056] Figure 5 A schematic block diagram of an exemplary electronic device provided for an embodiment of the present disclosure. DETAILED DESCRIPTION

[0057] The following is a description of exemplary embodiments of the present disclosure in conjunction with the accompanying drawings, including various details of the embodiments of the present disclosure to facilitate understanding, which should be considered as merely exemplary. Therefore, it should be recognized by those of ordinary skill in the art that various changes and modifications may be made to the embodiments described herein without departing from the scope and spirit of the present disclosure. Similarly, for the sake of clarity and conciseness, descriptions of well-known functions and structures are omitted in the following description.

[0058] The following describes the risk account processing method and device, electronic device and storage medium according to the embodiments of the present disclosure with reference to the accompanying drawings.

[0059] Figure 1 A flowchart of a method for processing risk accounts provided in an embodiment of the present disclosure.

[0060] like Figure 1 As shown, the method is applied in a server, and the method comprises the following steps:

[0061] Step 101, in response to a risky account logging into an ATM, obtaining a facial image of an operator of the risky account captured by the ATM.

[0062] A risk account refers to an account with credit risk, abnormal transactions or security loopholes. Risk accounts can be easily exploited by criminals to conduct illegal transactions on ATMs. Risk accounts are the target accounts for the entire risk account processing method. When a risk account logs into an ATM, the entire risk account processing method will be executed.

[0063] For example, the face image of the operator of the risky account is collected through the built-in camera of the ATM, and transmitted to the server through the network, so that the face expression recognition model deployed in the server can be used to recognize the expression of the collected face image. This can realize real-time monitoring of the operator of the risky account, timely discover the operator's abnormal behavior, and reduce financial risks.

[0064] Step 102: input the facial image into a preset facial expression recognition model to perform expression recognition to obtain an expression recognition result of the facial image.

[0065] The preset facial expression recognition model is a model that can be used to recognize facial expressions after training. It is trained by using a set of facial images containing different types of emotional expressions to ultimately obtain a model that can accurately determine the type of expression of the input facial image. The preset facial recognition model can be a model built based on machine learning or deep learning technology.

[0066] The processing process of the preset facial expression recognition model on the facial image can be achieved by the following method to enhance the quality of the collected facial image, such as adjusting the brightness, contrast, reducing noise, etc. Ensure that the facial image is correctly aligned before entering the preset facial expression recognition model. It is usually necessary to align the key feature points such as eyes and nose to the preset position. Use facial key point detection technology to locate facial feature points such as eyes, eyebrows, nose, mouth, etc. in the facial image. Based on these facial feature points, extract feature vectors that can represent facial expressions. Feature vectors include but are not limited to geometric features (such as the distance and angle between feature points) and texture features (such as color and texture information); convert the extracted feature vectors into an input format acceptable to the preset facial expression recognition model, and normalize the converted feature vectors to make them conform to the data distribution during the training of the preset facial expression recognition model. Input the normalized feature vectors into the preset facial expression recognition model; the preset facial expression recognition model performs inference based on the input feature vectors to predict the emotional expression category corresponding to the facial image. The preset facial expression recognition model outputs the probability distribution of each emotional expression category. According to the probability distribution, the emotional expression with the largest probability distribution is selected as the expression recognition result of the face image.

[0067] Emoticon recognition adds an extra layer of security, helping to prevent illegal transactions and protect account holders’ funds.

[0068] Step 103: If it is determined that the expression recognition result is the first emotional expression or the second emotional expression, it is determined that the transaction behavior of the risk account is abnormal, and a warning message is output.

[0069] The first emotional expression and the second emotional expression are pre-set specific emotional expression types. When the face image of the operator of the risk account is recognized by the preset facial expression recognition model, if the expression recognition result is one of the two, then the transaction behavior of the risk account will be determined to be abnormal. Among them, the first emotional expression can be happy, and the second emotional expression can be sad, but it should be clear that this statement is not intended to limit the first emotional expression to only happy and the second emotional expression to only sad. The first emotional expression and the second emotional expression can also be other expressions. The early warning information can be to notify bank staff to conduct further investigation, or automatically limit some transaction functions of the risk account, and notify the operator to go to the bank counter to handle related business in the form of voice or text to prevent possible illegal transactions.

[0070] Timely warning can reduce the economic losses of risk account owners. If criminals are using stolen accounts to conduct transactions, quick warnings can prevent funds from flowing out and protect the user's legitimate property rights. For example, a user's account is marked as a risky account due to password leakage. When the criminals try to withdraw money at an ATM, the abnormal expression is recognized, and the transaction is suspended after the warning, avoiding the loss of user funds.

[0071] The method for processing risk accounts provided by the present disclosure is, in response to a risk account logging into an ATM, obtaining a facial image of the operator of the risk account collected by the ATM; inputting the facial image into a preset facial expression recognition model for expression recognition to obtain an expression recognition result of the facial image; if it is determined that the expression recognition result is a first emotional expression or a second emotional expression, then determining that the transaction behavior of the risk account is abnormal, and outputting a warning message. Compared with the related art, the embodiment of the present disclosure uses an ATM to collect facial images of the operator of the risk account for expression recognition. If the expression recognition result is a first emotional expression or a second emotional expression, it indicates that the emotion of the operator of the risk account is abnormal. Since the operator's emotion is closely related to the security of the transaction behavior, the expression recognition result can be used to determine whether the transaction behavior is abnormal, thereby improving the security of the risk account.

[0072] In practical applications, after the face image is input into the preset face expression recognition model for expression recognition and the expression recognition result of the face image is obtained, the preset face expression recognition model may recognize other expressions in addition to the first and second emotional expressions. When other expressions are recognized, it can be implemented in the following manner but not limited to: if it is determined that the expression recognition result is the third emotional expression or the fourth emotional expression, then the transaction behavior of the risk account is determined to be normal. Among them, the third emotional expression and the fourth emotional expression are pre-set specific emotional expression types. When the face image of the risk account operator is recognized by the preset face expression recognition model, if the expression recognition result obtained is one of the two, then the transaction behavior of the risk account will be determined to be normal. Among them, the third emotional expression can be expressionless, and the fourth emotional expression can be surprise, but it should be clear that this statement is not intended to limit the third emotional expression to only expressionless and the fourth emotional expression to only surprise. The third emotional expression and the fourth emotional expression can also be other expressions.

[0073] By identifying emotional expressions that are normal for transaction behavior, it is possible to reduce instances where legitimate transactions are mistakenly flagged as abnormal, thus avoiding unnecessary alerts and subsequent investigation costs.

[0074] In practical applications, in order to improve the accuracy of facial expression recognition of facial images by the preset facial expression recognition model, it is necessary to train the preset facial expression recognition model, which can be achieved by but not limited to the following methods: Figure 2 As shown, Figure 2 A flowchart of a method for training a preset facial expression recognition model provided in an embodiment of the present disclosure includes:

[0075] Step 201, obtain a first emotional expression facial image set for training, a second emotional expression facial image set for training, a third emotional expression facial image set for training, and a fourth emotional expression facial image set for training; the first emotional expression facial image set for training includes a first training facial image and a corresponding first emotional expression label, the second emotional expression facial image set for training includes a second training facial image and a corresponding second emotional expression label, the third emotional expression facial image set for training includes a third training facial image and a corresponding third emotional expression label, and the fourth emotional expression facial image set for training includes a fourth training facial image and a corresponding fourth emotional expression label.

[0076] The first training face image, the second training face image, the third training face image, and the fourth training face image are actual face images with different emotional expression types. For example, the first training face image is a happy face image, the second training face image is a sad face image, the third training face image is an expressionless face image, and the fourth training face image is a surprised face image. The first emotional expression label, the second emotional expression label, the third emotional expression label, and the fourth emotional expression label are used to mark the identification information of the emotional expression type presented by the corresponding face image. For example, the first emotional expression label is happy, the second emotional expression label is sad, the third emotional expression label is expressionless, and the fourth emotional expression label is surprised.

[0077] The image sets and labels of different emotional expressions are independent of each other but together constitute complete training materials, so that the preset facial expression recognition model can comprehensively learn the differences and characteristics of multiple emotional expressions, avoid confusion between different emotional expression features, and thus enhance the ability of the preset facial expression recognition model to classify and recognize various emotional expressions.

[0078] Step 202, using the first training emotional expression facial image set, the second training emotional expression facial image set, the third training emotional expression facial image set and the fourth training emotional expression facial image set to train the initial facial expression recognition model to obtain the preset facial expression recognition model.

[0079] The initial facial expression recognition model refers to the basic model before the training process begins. It is the initial version of the algorithm or machine learning model used for expression recognition. The initial facial expression recognition model usually has a basic architecture and functions, but has not yet been trained and optimized for a specific task or dataset.

[0080] The first training face image and the first emotional expression label, the second training face image and the second emotional expression label, the third training face image and the third emotional expression label, and the fourth training face image and the fourth emotional expression label are respectively input into the initial facial expression recognition model. The initial facial expression recognition model performs expression recognition on the first training face image using the first emotional expression label, performs expression recognition on the second training face image using the second emotional expression label, performs expression recognition on the third training face image using the third emotional expression label, and performs expression recognition on the fourth training face image using the fourth emotional expression label to obtain a loss function. The loss function includes the weights of the initial facial expression recognition model. The weights are adjusted in a manner that makes the loss function approach a minimum value to obtain a preset facial expression recognition model.

[0081] As a refinement of step 201, when executing the acquisition of the first emotional expression facial image set for training, the second emotional expression facial image set for training, the third emotional expression facial image set for training and the fourth emotional expression facial image set for training, it can be achieved by but not limited to the following method: acquiring at least five fifth emotional expression facial image sets for training; different fifth emotional expression facial image sets in the at least five fifth emotional expression facial image sets correspond to training emotional expression facial images of different emotional expression types, and each of the fifth emotional expression facial image sets includes a fifth training emotional expression image and a corresponding fifth emotional expression label; based on the first similarity between each of the fifth emotional expression labels and the first emotional expression label, the fifth training emotional expression facial image sets of the fifth emotional expression labels whose first similarity is greater than a preset threshold are formed into the fifth training emotional expression facial image sets. A first training set of emotional expression facial images; based on the second similarity between each of the fifth emotional expression labels and the second emotional expression label, the fifth training set of emotional expression facial images of the fifth emotional expression label whose second similarity is greater than the preset threshold is used to form the second training set of emotional expression facial images; based on the third similarity between each of the fifth emotional expression labels and the third emotional expression label, the fifth training set of emotional expression facial images of the fifth emotional expression label whose third similarity is greater than the preset threshold is used to form the third training set of emotional expression facial images; based on the fourth similarity between each of the fifth emotional expression labels and the fourth emotional expression label, the fifth training set of emotional expression facial images of the fourth emotional expression label whose fourth similarity is greater than the preset threshold is used to form the fourth training set of emotional expression facial images.

[0082] The fifth training emotional expression image is a set of facial images for training a preset expression recognition model, showing different emotional expressions. The fifth training emotional expression image is used to expand or assist in constructing the first training facial image set, the second training facial image set, the third training facial image set, and the fourth training facial image set. The fifth emotional expression label is a label corresponding to the fifth training emotional expression image. The fifth emotional expression label is a type-differentiated emotional expression label of the first emotional expression label, the second emotional expression label, the third emotional expression label, and the fourth emotional expression label. For example, the fifth emotional expression label can be happy, excited, and excited. The first emotional expression label is happy, that is, happy, excited, and excited are types of differentiation of happy.

[0083] Similarity is an indicator to measure the degree of association between different emotional expression labels. By calculating the semantic similarity between different fifth emotional expression labels and the first emotional expression label, the second emotional expression label, the third emotional expression label and the fourth emotional expression label, and based on the size between the similarity and the preset threshold, it is determined which of the first training face image set, the second training face image set, the third training face image set and the fourth training face image set the fifth training emotional expression image corresponding to the different fifth emotional expression labels belongs to.

[0084] Through similarity screening, it can be ensured that the emotional expression images in the training set are highly consistent with the target emotional expression labels, thereby improving the recognition accuracy of the preset facial expression recognition model.

[0085] In actual applications, after obtaining at least five sets of fifth training emotional expression facial images, the operator of the risky account may wear a mask to cover his face when operating in front of the ATM. In order to improve the expression recognition rate of the preset facial expression recognition model for facial images with facial occlusion, it is necessary to perform facial occlusion on the fifth training emotional expression facial image. This can be achieved by but is not limited to the following method: all the fifth training emotional expression images in the at least five sets of fifth training emotional expression facial images are facially covered to obtain the covered fifth training emotional expression image; the fifth training emotional expression image includes the covered fifth training emotional expression image.

[0086] The deep learning facial key point detection model is used to identify the coordinate positions of key parts of the face, such as the eyes, nose, and mouth. Then, based on these key points, a rectangular, circular, or irregular polygonal area containing the main facial features is delineated as the part to be covered. A common method is to expand a certain range outward from the eyes, nose, and mouth as the center to form a rectangular area; the rectangular area is randomly blocked. Since the masked image weakens the individual identity-related features, the model will not overfit to the facial features of certain specific people during training, but will learn based on more general expression features. This enables the model to accurately identify the facial images of risk account operators of different individuals and different appearance features in actual applications by relying on the learned common expression features, which enhances the model's generalization ability in different scenarios and different object situations, and improves its stability and reliability.

[0087] In practical applications, the preset facial expression recognition model includes two separable convolutional networks with the same step size; the separable convolutional network is used to perform expression recognition on the facial image. The separable convolutional network decomposes this conventional convolution operation into two steps: depth-wise separable convolution and point-wise convolution. Depth-wise separable convolution first performs depth-wise separable convolution, which performs independent spatial convolution operations on each input channel. In other words, for each channel of the input image, a separate convolution kernel is used for convolution, which is equivalent to separating the processing of channel information and spatial information, and first processing the spatial features of each channel separately, reducing the amount of calculation. Through this step, the same number of feature maps as the number of input channels will be obtained, and each feature map corresponds to the result of an input channel after spatial convolution. Point-by-point convolution: After the depth-wise separable convolution, point-by-point convolution is performed. Point-by-point convolution will act on all feature maps output by the depth-wise separable convolution, which is used to fuse information between different channels and realize feature transformation in the channel dimension. For example, if the depthwise separable convolution outputs multiple feature maps, using a point-by-point convolution kernel for convolution can recombine and transform the features of multiple channels to obtain the final desired number of output feature maps and the corresponding feature representation, thereby completing the extraction and integration of the input image features.

[0088] By decomposing the convolution operation, the separable convolutional network significantly reduces the amount of computation compared to the traditional convolutional network. Especially when processing image data with a certain resolution and a large number of channels (such as considering multiple feature channels) such as face images, the saving of computing resources is more obvious, which allows the model to complete training and reasoning (the process of identifying and judging the expression of the input image) in a relatively short time, improving computing efficiency.

[0089] In one possible implementation of the disclosed embodiment, a network fine-grained pruning method is used to reduce redundant parameters of a preset facial expression recognition model, so that the preset facial expression recognition model retains the residual mechanism and separable convolution, while optimizing the correlation between different network layers, and improving the recognition ability of micro-expression information. In order to prevent network overfitting, random deactivation technology is used. Traditional neural networks need to learn each node. The introduction of random deactivation technology allows some nodes to be automatically and randomly deleted during the training process. Not only is the node temporarily deleted, but its associated incoming and outgoing connections are also deleted. In this way, the adjusted network structure is reduced to a certain extent, but each neural unit becomes more robust, reducing the complex relationship between neurons, and the training time is significantly reduced; further improving the expression recognition ability of the preset facial expression recognition model. Use global average pooling processing, and then use a normalized exponential function to complete the classification of facial expressions.

[0090] By using the face masking technology, the mask is added to the uncovered training emotional expression face images in the training emotional expression face image set.

[0091] The preset facial expression recognition model is divided into three parts as a whole: input layer, core layer, and output layer. In the input layer, two identical standardized convolutional networks with an interval of 1 are used, which not only retains the feature information of the input image, but also strengthens the depth of the network. The use of separable convolution can optimize the ability to extract expression information while reducing network parameters. After the separable convolution module is processed, global average pooling is used, and then the normalized exponential function is used to complete the classification of facial expressions. The preset facial expression recognition model uses a linear activation function between the depth separation convolution layer and the point-by-point convolution, and batch normalization layers and linear rectification activation functions are added between other convolution layers. This is done to enhance the nonlinear expression ability of the preset facial expression recognition model while preventing data divergence.

[0092] The preset facial expression recognition model is connected to the bank account risk management system, and combined with the bank's micro-expression risk determination model, a preventive measure for special facial expression recognition of risky accounts at bank ATMs is constructed. If it is determined to be an abnormal transaction, the transaction will be intercepted and the customer will be prompted to go to the branch counter to handle the fund business.

[0093] The preset facial expression recognition model is trained using 85% of the emotional expression face image set as a training set and 15% as a test set.

[0094] In order to better understand the difference between the preset facial expression recognition model and other facial expression recognition models, the embodiment of the present disclosure provides comparative data of the preset facial expression recognition model and other facial expression recognition models as shown in the following table:

[0095] Table 1

[0096]

[0097] It can be seen from Table 1 that the preset facial expression recognition model has fewer parameters and fewer training times than other facial expression recognition models, while the accuracy of expression recognition is comparable to that of other facial expression recognition models.

[0098] In summary, the embodiments of the present disclosure can achieve the following effects:

[0099] The disclosed embodiment utilizes an ATM to collect facial images of operators of risky accounts for expression recognition. If the expression recognition result is a first emotional expression or a second emotional expression, it indicates that the emotions of the operator of the risky account are abnormal. Since the emotions of the operator are closely related to the security of the transaction behavior, whether the transaction behavior is abnormal can be determined through the expression recognition result, thereby improving the security of the risky account.

[0100] Corresponding to the above-mentioned method for processing risk accounts, the present invention also proposes a device for processing risk accounts. Since the device embodiment of the present invention corresponds to the above-mentioned method embodiment, details not disclosed in the device embodiment can be referred to the above-mentioned method embodiment, and will not be repeated in the present invention.

[0101] Figure 3 A schematic diagram of a structure of a risk account processing device provided by an embodiment of the present disclosure, wherein the device is applied to a server, such as Figure 3 As shown, including:

[0102] A first acquisition unit 31 is used for acquiring a face image of an operator of the risky account collected by the ATM in response to the risky account logging into the ATM;

[0103] The recognition unit 32 is used to input the face image into a preset face expression recognition model to perform expression recognition and obtain an expression recognition result of the face image;

[0104] The determination unit 33 is used to determine that the transaction behavior of the risk account is abnormal and output warning information when it is determined that the expression recognition result is the first emotional expression or the second emotional expression.

[0105] The processing device of the risk account provided by the present disclosure obtains the facial image of the operator of the risk account collected by the ATM in response to the risk account logging into the ATM; inputs the facial image into a preset facial expression recognition model for expression recognition to obtain the expression recognition result of the facial image; if it is determined that the expression recognition result is the first emotional expression or the second emotional expression, it is determined that the transaction behavior of the risk account is abnormal, and outputs warning information. Compared with the related art, the embodiment of the present disclosure uses the facial image of the operator of the risk account collected by the ATM for expression recognition. If the expression recognition result is the first emotional expression or the second emotional expression, it indicates that the emotion of the operator of the risk account is abnormal. Since the emotion of the operator is closely related to the security of the transaction behavior, the expression recognition result can be used to determine whether the transaction behavior is abnormal, thereby improving the security of the risk account.

[0106] Furthermore, in a possible implementation of the embodiment of the present disclosure, as Figure 4 As shown, the device also includes:

[0107] The determination unit 33 is also used for, after inputting the facial image into a preset facial expression recognition model for expression recognition and obtaining the expression recognition result of the facial image, when it is determined that the expression recognition result is the third emotional expression or the fourth emotional expression, determining that the transaction behavior of the risk account is normal.

[0108] Furthermore, in a possible implementation of the embodiment of the present disclosure, as Figure 4 As shown, the device also includes:

[0109] The second acquisition unit 34 is used to acquire a first emotional expression face image set for training, a second emotional expression face image set for training, a third emotional expression face image set for training, and a fourth emotional expression face image set for training; the first emotional expression face image set for training includes a first training face image and a corresponding first emotional expression label, the second emotional expression face image set for training includes a second training face image and a corresponding second emotional expression label, the third emotional expression face image set for training includes a third training face image and a corresponding third emotional expression label, and the fourth emotional expression face image set for training includes a fourth training face image and a corresponding fourth emotional expression label;

[0110] The training unit 35 is used to train the initial facial expression recognition model using the first training emotional expression facial image set, the second training emotional expression facial image set, the third training emotional expression facial image set and the fourth training emotional expression facial image set to obtain the preset facial expression recognition model.

[0111] Furthermore, in a possible implementation of the embodiment of the present disclosure, as Figure 4 As shown, the second acquisition unit 34 includes:

[0112] An acquisition module 341 is used to acquire at least five fifth training emotional expression facial image sets; different fifth training emotional expression facial image sets in the at least five fifth training emotional expression facial image sets correspond to training emotional expression facial images of different emotional expression types, and each of the fifth training emotional expression facial image sets includes a fifth training emotional expression image and a corresponding fifth emotional expression label;

[0113] A composition module 342 is used to form the first emotional expression face image set for training with the fifth emotional expression face image set of the fifth emotional expression labels whose first similarity is greater than a preset threshold based on the first similarity between each of the fifth emotional expression labels and the first emotional expression label;

[0114] The composition module 342 is also used to, based on the second similarity between each of the fifth emotional expression labels and the second emotional expression label, form the fifth training emotional expression face image set of the fifth emotional expression label whose second similarity is greater than the preset threshold into the second training emotional expression face image set;

[0115] The composition module 342 is also used for, based on the third similarity between each of the fifth emotional expression labels and the third emotional expression label, composing the fifth training emotional expression face image set of the fifth emotional expression label whose third similarity is greater than the preset threshold into the third training emotional expression face image set;

[0116] The composition module 342 is also used to, based on the fourth similarity between each of the fifth emotional expression labels and the fourth emotional expression label, form the fifth training emotional expression facial image set of the fourth emotional expression labels whose fourth similarity is greater than the preset threshold into the fourth training emotional expression facial image set.

[0117] Furthermore, in a possible implementation of the embodiment of the present disclosure, as Figure 4 As shown, the device also includes:

[0118] A processing unit 36 ​​is used for, after obtaining at least five fifth training emotional expression face image sets, performing facial masking processing on all fifth training emotional expression images in the at least five fifth training emotional expression face image sets to obtain masked fifth training emotional expression images;

[0119] The fifth training emotional expression image includes the masked fifth training emotional expression image.

[0120] Furthermore, in a possible implementation of the embodiment of the present disclosure, the preset facial expression recognition model includes two separable convolutional networks with the same step size; the separable convolutional network is used to perform expression recognition on the facial image.

[0121] It should be noted that the above explanation of the method embodiment is also applicable to the device of the embodiment of the present disclosure, and the principle is the same, which is no longer limited in the embodiment of the present disclosure.

[0122] According to an embodiment of the present disclosure, the present disclosure also provides an electronic device, a readable storage medium and a computer program product.

[0123] Figure 5A schematic block diagram of an example electronic device 400 that can be used to implement an embodiment of the present disclosure is shown. 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 can also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices, 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 disclosure described and / or required herein.

[0124] like Figure 5 As shown, the device 400 includes a computing unit 401, which can perform various appropriate actions and processes according to a computer program stored in a ROM (Read-Only Memory) 402 or a computer program loaded from a storage unit 408 to a RAM (Random Access Memory) 403. In the RAM 403, various programs and data required for the operation of the device 400 can also be stored. The computing unit 401, the ROM 402, and the RAM 403 are connected to each other via a bus 404. An I / O (Input / Output) interface 405 is also connected to the bus 404.

[0125] A number of components in the device 400 are connected to the I / O interface 405, including: an input unit 406, such as a keyboard, a mouse, etc.; an output unit 407, such as various types of displays, speakers, etc.; a storage unit 408, such as a disk, an optical disk, etc.; and a communication unit 409, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 409 allows the device 400 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0126] The computing unit 401 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the computing unit 401 include, but are not limited to, a CPU (Central Processing Unit), a GPU (Graphic Processing Units), various dedicated AI (Artificial Intelligence) computing chips, various computing units running machine learning model algorithms, a DSP (Digital Signal Processor), and any appropriate processor, controller, microcontroller, etc. The computing unit 401 performs the various methods and processes described above, such as the processing method of the risk account. For example, in some embodiments, the processing method of the risk account may be implemented as a computer software program, which is tangibly contained in a machine-readable medium, such as a storage unit 408. In some embodiments, part or all of the computer program may be loaded and / or installed on the device 400 via the ROM 402 and / or the communication unit 409. When the computer program is loaded into the RAM 403 and executed by the computing unit 401, one or more steps of the method described above may be performed. Alternatively, in other embodiments, the computing unit 401 may be configured to execute the aforementioned risk account processing method in any other appropriate manner (eg, by means of firmware).

[0127] Various embodiments of the systems and techniques described above herein may be implemented in digital electronic circuit systems, integrated circuit systems, FPGAs (Field Programmable Gate Arrays), ASICs (Application-Specific Integrated Circuits), ASSPs (Application Specific Standard Products), SOCs (System On Chips), CPLDs (Complex Programmable Logic Devices), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include: being implemented in one or more computer programs that may be executed and / or interpreted on a programmable system including at least one programmable processor that may be a special purpose or general purpose programmable processor that may receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0128] The program code for implementing the method of the present disclosure may be written in any combination of one or more programming languages. These program codes may be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that the program code, when executed by the processor or controller, enables the functions / operations specified in the flow chart and / or block diagram to be implemented. The program code may be executed entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine as a stand-alone software package, or entirely on a remote machine or server.

[0129] In the context of the present disclosure, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, device, or equipment. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or device, or any suitable combination of the foregoing. More specific examples of machine-readable storage media may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a RAM, a ROM, an EPROM (Electrically Programmable Read-Only-Memory) or a flash memory, an optical fiber, a CD-ROM (Compact Dis sc Read-Only Memory), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0130] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer 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 a pointing device (e.g., a mouse or a trackball), through which the user can provide input to the computer. Other types of devices can also be used to provide interaction with the user; for example, the 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 acoustic input, voice input, or tactile input).

[0131] The systems and techniques described herein may be implemented in a computing system that includes backend components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes frontend components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such backend components, middleware components, or frontend components. The components of the system may be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: LAN (Local Area Network), WAN (Wide Area Network), the Internet, and blockchain networks.

[0132] A computer system may include a client and a server. The client and the server are generally remote from each other and usually interact through a communication network. The relationship between the client and the server is generated by computer programs running on the corresponding computers and having a client-server relationship with each other. The server may be a cloud server, also known as a cloud computing server or cloud host, which is a host product in the cloud computing service system to solve the defects of difficult management and weak business scalability in traditional physical hosts and VPS services ("Virtual Private Server", or "VPS" for short). The server may also be a server of a distributed system, or a server combined with a blockchain.

[0133] It should be noted that artificial intelligence is a discipline that studies how computers can simulate certain human thought processes and intelligent behaviors (such as learning, reasoning, thinking, planning, etc.), and includes both hardware-level and software-level technologies. Artificial intelligence hardware technologies generally include technologies such as sensors, dedicated artificial intelligence chips, cloud computing, distributed storage, and big data processing; artificial intelligence software technologies mainly include computer vision technology, speech recognition technology, natural language processing technology, as well as machine learning / deep learning, big data processing technology, knowledge graph technology, and other major directions.

[0134] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps recorded in this disclosure can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solutions disclosed in this disclosure can be achieved, and this document does not limit this.

[0135] The above specific implementations do not constitute a limitation on the protection scope of the present disclosure. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present disclosure shall be included in the protection scope of the present disclosure.

Claims

1. A method for handling risky accounts, characterized in that: include: In response to a risk account logging into an ATM, acquiring a facial image of an operator of the risk account captured by the ATM; Inputting the facial image into a preset facial expression recognition model to perform expression recognition, and obtaining an expression recognition result of the facial image; If it is determined that the expression recognition result is the first emotional expression or the second emotional expression, it is determined that the transaction behavior of the risk account is abnormal, and a warning message is output.

2. The method according to claim 1, characterized in that After inputting the face image into a preset face expression recognition model for expression recognition to obtain the expression recognition result of the face image, the method further includes: If it is determined that the expression recognition result is the third emotional expression or the fourth emotional expression, it is determined that the transaction behavior of the risk account is normal.

3. The method according to claim 2, characterized in that The method further comprises: Acquire a first training emotional expression facial image set, a second training emotional expression facial image set, a third training emotional expression facial image set, and a fourth training emotional expression facial image set; the first training emotional expression facial image set includes a first training facial image and a corresponding first emotional expression label, the second training emotional expression facial image set includes a second training facial image and a corresponding second emotional expression label, the third training emotional expression facial image set includes a third training facial image and a corresponding third emotional expression label, and the fourth training emotional expression facial image set includes a fourth training facial image and a corresponding fourth emotional expression label; The initial facial expression recognition model is trained using the first training emotional expression facial image set, the second training emotional expression facial image set, the third training emotional expression facial image set and the fourth training emotional expression facial image set to obtain the preset facial expression recognition model.

4. The method according to claim 3, characterized in that The obtaining of the first emotional expression face image set for training, the second emotional expression face image set for training, the third emotional expression face image set for training, and the fourth emotional expression face image set for training comprises: Acquire at least five fifth training emotional expression facial image sets; different fifth training emotional expression facial image sets in the at least five fifth training emotional expression facial image sets correspond to training emotional expression facial images of different emotional expression types, and each of the fifth training emotional expression facial image sets includes a fifth training emotional expression image and a corresponding fifth emotional expression label; Based on the first similarity between each of the fifth emotional expression labels and the first emotional expression label, the fifth training emotional expression face image set of the fifth emotional expression labels whose first similarity is greater than a preset threshold is combined into the training first emotional expression face image set; Based on the second similarity between each of the fifth emotional expression labels and the second emotional expression label, the fifth training emotional expression face image set of the fifth emotional expression labels whose second similarity is greater than the preset threshold is used to form the second training emotional expression face image set; Based on the third similarity between each of the fifth emotional expression labels and the third emotional expression label, the fifth training emotional expression face image set of the fifth emotional expression labels whose third similarity is greater than the preset threshold is used to form the third training emotional expression face image set; Based on the fourth similarity between each of the fifth emotional expression labels and the fourth emotional expression label, the fifth training emotional expression facial image set of the fourth emotional expression labels whose fourth similarity is greater than the preset threshold is combined into the training fourth emotional expression facial image set.

5. The method according to claim 4, characterized in that After obtaining at least five fifth training emotional expression face image sets, the method further includes: Performing facial masking processing on all fifth training emotional expression images in the at least five fifth training emotional expression face image sets to obtain masked fifth training emotional expression images; The fifth training emotional expression image includes the masked fifth training emotional expression image.

6. The method according to claim 1, characterized in that The preset facial expression recognition model includes two separable convolutional networks with the same step size; the separable convolutional networks are used to perform expression recognition on the facial image.

7. A device for processing risk accounts, characterized in that: include: A first acquisition unit is used for acquiring a face image of an operator of the risky account collected by the automatic teller machine in response to the risky account logging into the automatic teller machine; A recognition unit, used for inputting the face image into a preset face expression recognition model to perform expression recognition, and obtaining an expression recognition result of the face image; The determination unit is used to determine that the transaction behavior of the risk account is abnormal and output warning information when it is determined that the expression recognition result is the first emotional expression or the second emotional expression.

8. An electronic device, characterized in that: include: at least one processor; as well as a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method according to any one of claims 1 to 6.

9. A non-transitory computer-readable storage medium storing computer instructions, characterized in that: The computer instructions are used to cause the computer to execute the method according to any one of claims 1-6.

10. A computer program product, characterized in that The invention comprises a computer program which, when executed by a processor, implements the method according to any one of claims 1 to 6.

Images