Method, device, equipment and storage medium for identifying abnormal loan behavior of enterprises

Abstract

The present application relates to a method, device, computer equipment, computer-readable storage medium and computer program product for identifying abnormal loan behavior of an enterprise, the method comprising: a credit institution server responds to a loan application of a pre-loan enterprise, performs a loan requirement check on an initial scene image, and extracts multiple key loan information of the enterprise through a key information recognition model when the initial scene image meets the loan requirements; and performs enterprise information matching on multiple key loan information of the enterprise, and determines that the loan behavior of the pre-loan enterprise is normal when the enterprise information matching is successful. The present application uses a server to audit the scene images used by the pre-loan enterprise for loan review, replacing manual review, and can achieve the technical effect of improving review efficiency; at the same time, it can improve the accuracy of judging the loan behavior of the pre-loan enterprise.

Description

Technical Field

[0001] The present application relates to the field of big data processing technology, and in particular to a method, device, computer equipment, computer-readable storage medium and computer program product for identifying abnormal loan behavior of an enterprise. Background Art

[0002] Due to the characteristics of online finance, numerous fraud incidents are often encountered in the process of business development. For example, in the corporate loan scenario in the financial industry, there are complex and changeable loan fraud incidents. Therefore, how to effectively solve the problem of bank anti-fraud loan incidents is particularly important.

[0003] Among the related technologies, most of the anti-loan fraud solutions based on artificial intelligence are based on face recognition and voice recognition, and some are added in emotional intelligence recognition technology. However, these technologies are mainly used to identify personal loan fraud, and the identification of corporate loan behavior still relies mainly on manual work, which leads to low efficiency and low accuracy in identifying corporate loan behavior. Summary of the invention

[0004] The present application provides a method, apparatus, computer device, computer-readable storage medium and computer program product for identifying abnormal loan behavior of an enterprise, which can quickly identify whether the loan behavior of an enterprise is abnormal, and has high efficiency and accuracy in identification.

[0005] In a first aspect, the present application provides a method for identifying abnormal loan behavior of an enterprise, the method comprising:

[0006] In response to a loan application from a pre-loan enterprise, performing a loan requirement check on an initial scene image of the pre-loan enterprise to obtain a check result;

[0007] If the inspection result shows that the initial scene image meets the loan requirements, multiple key information of enterprise loans are extracted from the initial scene image through the key information recognition model; the key information of enterprise loans is used to characterize the characteristics that distinguish the pre-loan enterprise from other pre-loan enterprises;

[0008] Enterprise information matching is performed on key loan information of multiple enterprises. If the enterprise information matching is successful, it is determined that the loan behavior of the pre-loan enterprise is normal.

[0009] In a second aspect, the present application also provides a device for identifying abnormal loan behavior of an enterprise, the device comprising:

[0010] A verification module, for responding to a loan application from a pre-loan enterprise, performing a loan requirement verification on an initial scene image of the pre-loan enterprise, and obtaining a verification result;

[0011] The extraction module is used to extract multiple key information of enterprise loans from the initial scene image through the key information recognition model if the inspection result shows that the initial scene image meets the loan requirements; the key information of enterprise loans is used to characterize the characteristics that distinguish the pre-loan enterprise from other pre-loan enterprises;

[0012] The matching determination module is used to match the enterprise information of multiple enterprise loan key information. If the enterprise information is successfully matched, it is determined that the loan behavior of the pre-loan enterprise is normal.

[0013] In a third aspect, the present application further provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of any of the above methods when executing the computer program.

[0014] In a fourth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, which implements the steps of any of the above methods when the computer program is executed by a processor.

[0015] In a fifth aspect, the present application also provides a computer program product, including a computer program, which implements the steps of any of the above methods when the computer program is executed by a processor.

[0016] The present application provides a method, device, computer equipment, computer-readable storage medium and computer program product for identifying abnormal loan behavior of an enterprise. The method includes: a credit institution server responds to a loan application of a pre-loan enterprise, performs a loan requirement check on the initial scene image of the pre-loan enterprise, and extracts multiple key enterprise loan information from the initial scene image through a key information recognition model when the initial scene image meets the loan requirements; and performs enterprise information matching on multiple key enterprise loan information, and determines that the loan behavior of the pre-loan enterprise is normal when the enterprise information matching is successful. The present application reviews the scene image used by the pre-loan enterprise for loan review by the server, replacing manual review, and can achieve the technical effect of improving review efficiency; at the same time, after obtaining the scene image of the pre-loan enterprise, the server will perform a loan requirement check on the scene image, and realize the preliminary screening of the loan information of the pre-loan enterprise. Further, when the server determines that the scene image of the pre-loan enterprise meets the loan requirements, it extracts the key enterprise loan information from the scene image through the key information recognition model, and matches the extracted multiple key enterprise loan information for enterprise information, so as to realize further review of the loan behavior of the pre-loan enterprise. Through the above-mentioned preliminary screening and precise matching, the accuracy of judging the loan behavior of the pre-loan enterprise can be improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 This is an application environment diagram of a method for identifying abnormal loan behavior of an enterprise in one embodiment;

[0018] Figure 2 A flowchart of a method for identifying abnormal loan behavior of an enterprise in one embodiment;

[0019] Figure 3 A flowchart of a method for identifying abnormal loan behavior of an enterprise in another embodiment;

[0020] Figure 4 A flowchart of a method for identifying abnormal loan behavior of an enterprise in another embodiment;

[0021] Figure 5 A flowchart of a method for identifying abnormal loan behavior of an enterprise in another embodiment;

[0022] Figure 6 A flowchart of a method for identifying abnormal loan behavior of an enterprise in another embodiment;

[0023] Figure 7 A flowchart of a method for identifying abnormal loan behavior of an enterprise in another embodiment;

[0024] Figure 8 It is a structural block diagram of a device for identifying abnormal loan behavior of an enterprise in one embodiment;

[0025] Fig. 9 FIG. 4 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0026] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0027] The method for identifying abnormal loan behavior of an enterprise provided in the embodiment of the present application can be applied to Figure 1In the application environment shown. Among them, the terminal 102 communicates with the server 104 through the network. The data storage system can store the data that the server 104 needs to process. The data storage system can be integrated on the server 104, or it can be placed on the cloud or other network servers. The terminal 102 sends the initial scene image to the server 104 through the network, and the server 104 performs a loan requirement inspection on the initial scene image, and then further extracts multiple enterprise loan key information from the initial scene image through the key information recognition model when the initial scene image meets the loan requirements, and performs enterprise information matching on the multiple enterprise loan key information; when the enterprise information is successfully matched, it is determined that the loan behavior of the pre-loan enterprise is normal. The intelligent recognition process replaces manual recognition to improve the efficiency and accuracy of recognition. Among them, the terminal 102 can be, but is not limited to, various personal computers, laptops, smart phones, tablet computers, Internet of Things devices and portable wearable devices, and the Internet of Things devices can be smart speakers, smart TVs, smart air conditioners, smart car-mounted devices, etc. Portable wearable devices can be smart watches, smart bracelets, head-mounted devices, etc. The server 104 can be implemented with an independent server or a server cluster consisting of multiple servers.

[0028] In one embodiment, Figure 2 As shown, a method for identifying abnormal loan behavior of an enterprise is provided, and the method is applied to Figure 1 The server in the example is used to illustrate the following steps:

[0029] Step S202, in response to the loan application of the pre-loan enterprise, a loan requirement check is performed on the initial scene image of the pre-loan enterprise.

[0030] Among them, the initial scene image is used for the integration of all information provided by the pre-loan enterprise to the lending institution through the terminal for loan review, and may include image information such as enterprise office scene images, enterprise asset images, enterprise culture images, enterprise logo images, and enterprise material images. The image information may be presented in the form of pictures or in the form of videos, and there is no limitation on this. In order to facilitate the credit institution to conduct a unified review of various scene images of different pre-loan enterprises, the credit institution may set corresponding image requirements for different scene images, such as size, clarity, etc., so the credit institution needs to conduct loan requirement inspection on the initial scene images provided by all pre-loan enterprises to avoid the pre-loan enterprise providing images that do not meet the loan requirements, so that the subsequent images cannot be recognized, resulting in low recognition efficiency and low accuracy. The loan requirement inspection may be to inspect the initial scene image through various algorithms, or to inspect the initial scene image through different models, and there is no limitation on this. The inspection may be to inspect the clarity of the image, the size of the image, the authenticity of the image, the similarity of the image, etc.

[0031] Exemplarily, a pre-loan enterprise uploads various pictures, such as panoramic pictures of office space, pictures of office space segmentation, pictures of external office buildings of the enterprise, pictures of signs and markers of the surrounding environment of the enterprise's geographical location, pictures of the property and logistics work of the enterprise, pictures of the fixed assets of the enterprise, pictures of products produced and sold by the enterprise, product stock images, pictures of seals, pictures of raw materials for product production, pictures of main business, pictures of water meters and electricity meters, pictures of corporate culture promotion, pictures of corporate honorary certificates, pictures of corporate certificates, pictures of corporate logos, pictures of corporate organizational structures, pictures of corporate ownership certificates, pictures of corporate cash flow sheets, pictures of detailed inventory bills, pictures of corporate employee salary lists, pictures of corporate financing forms, pictures of corporate project schedules, etc., through computer equipment provided by a credit institution. All pictures can be input into a picture verification model, and the output results of the model can be used to determine whether the pictures meet the loan requirements.

[0032] Step S204: if the initial scene image meets the loan requirements, multiple key information of enterprise loans are extracted from the initial scene image through the key information recognition model; the key information of enterprise loans is used to characterize the characteristics that distinguish the pre-loan enterprise from other pre-loan enterprises.

[0033] Among them, according to the above steps, when the server determines that the initial scene image provided by the pre-loan enterprise meets the loan requirements, it can continue to extract key information from the image, and extract multiple key information of the enterprise loan. The key information of the enterprise loan is used to characterize some characteristics of other pre-loan enterprises, such as enterprise name, enterprise logo, enterprise geographic location information, enterprise account information, etc. The key information recognition model is used to extract the key information of the enterprise loan from the initial scene image provided by the pre-loan enterprise. The key information recognition model may include an enterprise name recognition model, an enterprise logo recognition model, an enterprise account information recognition model, etc. Different enterprise loan key information is recognized by different models, and the recognition accuracy is higher, and the key information of the enterprise obtained is more accurate. The key information recognition model can be obtained by training a neural network model based on different pictures, such as RNN, CNN, Faster RCNN and other neural networks, which are not limited in this application. The training samples can be obtained from the historical database of the server. It should be noted that after the server performs loan requirement verification on the initial scene image, it can classify the initial scene image to obtain images of different categories, or it can mark information on each image, such as office panorama, financial report image, corporate logo image, corporate seal image, corporate product image, etc., to facilitate the subsequent targeted input of the target image into different key information recognition models to obtain multiple different corporate loan key information.

[0034] For example, a picture containing a company name may be input into a company name recognition model, and the company name recognition model may recognize that the company name is Xi'an Runzejing Consulting Service Cable Company; an image containing a company logo may be input into a company logo recognition model to obtain the company logo; an image of a company's financial statement may be input into a company account information recognition model to obtain the company's account data, etc. Finally, multiple company loan information may be obtained.

[0035] Step S206, performing enterprise information matching on multiple key enterprise loan information, if the enterprise information matching is successful, it is determined that the loan behavior of the pre-loan enterprise is normal.

[0036] Among them, based on the above recognition of different images through different key information recognition models, multiple key information of corporate loans are obtained, and further corporate information matching can be performed on multiple key information of corporate loans to improve the accuracy of identifying corporate loan behavior. The so-called corporate information matching refers to verifying the authenticity and validity of multiple key information of corporate loans. Since a formal enterprise will be registered in the early stage of its establishment, the registration information of the enterprise will be queried on the official website of the enterprise, other external websites, enterprise query software and other websites, APPs, and some key information of the enterprise can be queried. Some key information of the enterprise can also be obtained from relevant departments such as the audit department, financial supervision department, and tax department. Therefore, the server can obtain some key information about the enterprise through the above channels, and then match the obtained key information of the enterprise with the multiple key information of corporate loans extracted by the above key information recognition model to realize the verification of the key information of corporate loans. If the multiple key information of enterprise loans extracted by the key information identification model successfully matches the key information of enterprises obtained by the server from other channels, it means that there is no problem with the loan review information provided by the enterprise, which further indicates that the loan behavior of the pre-loan enterprise is normal, that is, the possibility of loan fraud by the pre-loan enterprise is small, and the subsequent evaluation of the loan amount and rate of the enterprise can be continued to complete the loan application of the pre-loan enterprise. On the contrary, if the multiple key information of enterprise loans extracted by the key information identification model fails to match the key information of enterprises obtained by the server from other channels, it means that there is a problem with the loan review information provided by the enterprise, which further indicates that the loan behavior of the pre-loan enterprise is abnormal, that is, the possibility of loan fraud by the pre-loan enterprise is large, and the credit institution may reject the loan application of the pre-loan enterprise.

[0037] The present application provides a method for identifying abnormal loan behavior of an enterprise. The method includes: a credit institution server responds to a loan application of a pre-loan enterprise, performs a loan requirement check on the initial scene image of the pre-loan enterprise, and extracts multiple key enterprise loan information from the initial scene image through a key information recognition model when the initial scene image meets the loan requirements; and performs enterprise information matching on multiple key enterprise loan information, and when the enterprise information matching is successful, it is determined that the loan behavior of the pre-loan enterprise is normal. The present application reviews the scene image used by the pre-loan enterprise for loan review by the server, replacing manual review, and can achieve the technical effect of improving review efficiency; at the same time, after obtaining the scene image of the pre-loan enterprise, the server will perform a loan requirement check on the scene image, and realize the preliminary screening of the loan information of the pre-loan enterprise. Further, when the server determines that the scene image of the pre-loan enterprise meets the loan requirements, it extracts the key enterprise loan information from the scene image through a key information recognition model, and matches the extracted multiple key enterprise loan information for enterprise information, so as to realize further review of the loan behavior of the pre-loan enterprise. Through the above-mentioned preliminary screening and precise matching, the accuracy of judging the loan behavior of the pre-loan enterprise can be improved.

[0038] In one embodiment, Figure 3 As shown, an optional method embodiment of performing loan requirement inspection on an initial scene image provided by an embodiment of the present application is provided, and the method embodiment comprises the following steps:

[0039] Step S302, inspecting the image clarity, image authenticity and image repeatability of the initial scene image of the pre-loan enterprise;

[0040] Step S304: If the image clarity, image authenticity and image repeatability are all checked and passed, the test result is that the initial scene image meets the loan requirements.

[0041] Among them, the clarity of the image represents whether the appearance of the initial scene image provided by the pre-loan enterprise meets the requirements, because low image clarity will affect the subsequent recognition of key information of the image, and the accuracy of the key information of the enterprise loan obtained cannot be guaranteed, further affecting the judgment of the loan behavior of the pre-loan enterprise. The clarity of the image can be determined by various clarity algorithms, such as various gradient functions, variance functions, energy functions and other algorithms, or it can be determined according to a model, for example, all initial scene images can be input into a pre-trained clarity trial model, and the clarity of the image is determined according to the output of the model. This application does not limit this.

[0042] The authenticity of the image indicates whether the scene image provided by the pre-loan enterprise is authentic and whether it can truly reflect the real information of the pre-loan enterprise. The authenticity of the image can be determined by whether the image has been tampered with. The method for determining the authenticity of the image can also be determined by extracting the overall pixels, eigenvalues, noise, etc. of the image, or it can be determined by a model. For example, all initial scene images can be input into an image eigenvalue recognition model, and the output of the model is used to determine whether the image is a tampered image. This application does not limit this.

[0043] The repeatability of the image indicates whether there are repeated images among the multiple different initial scene images provided by the pre-loan enterprise, because uploading repeated images can easily lead to the lack of some necessary initial scene images in the initial scene images provided by the pre-loan enterprise, affecting the subsequent judgment of the loan behavior of the pre-loan enterprise. The method for determining whether an image is uploaded repeatedly can be determined by comparing the similarity between two images, or it can be determined by a model. For example, all initial scene images can be input into a similarity comparison model, and the output of the model can be used to determine whether there are similar images. After determining the similar images, the server can send a prompt message to the pre-loan enterprise, prompting the pre-loan enterprise to re-upload the initial scene image of the corresponding type, or it can delete the repeated initial scene images and other operations.

[0044] After the server verifies the image clarity, image authenticity and image repeatability of the initial scene image through the above steps, and after the initial scene image passes the inspection from three dimensions, it is determined that all the initial scene images provided by the pre-loan enterprise meet the loan requirements of the credit institution.

[0045] The present application provides a method for identifying abnormal loan behavior of an enterprise. After obtaining the initial scene image provided by the pre-loan enterprise, the server in the present application can inspect the initial scene image from the aspects of image clarity, image authenticity and image repeatability respectively. Only after all three aspects are inspected, can subsequent identification operations be performed, thereby achieving an initial screening of the loan behavior of the pre-loan enterprise, providing reliable data support for subsequent identification, and being able to help quickly identify the loan behavior of the pre-loan enterprise in the subsequent period, and also improving the accuracy of identifying the loan behavior of the pre-loan enterprise.

[0046] In one embodiment, Figure 4 As shown, an optional method embodiment of performing image clarity inspection on an initial scene image provided by an embodiment of the present application, the method embodiment comprises the following steps:

[0047] Step S402, obtaining the clarity parameters of the initial scene image of the pre-loan enterprise in the spatial domain and the frequency domain respectively.

[0048] Step S404: If the clarity parameter is within the preset clarity threshold range, it is determined that the image clarity test of the initial scene image of the pre-loan enterprise has passed.

[0049] Among them, the clarity of the initial scene image provided by the pre-loan enterprise can be determined by obtaining clarity parameters of the initial scene image in the spatial domain and the frequency domain. The method of obtaining the clarity parameters can be obtained through a variety of methods such as a variety of gradient functions, variance functions, energy functions, SSIM structural similarity models, SPSMD algorithms, image FFT transforms, power spectrum algorithms, etc. The clarity parameters can be image brightness, image contrast, image structure, etc., that is, the brightness of the initial scene image can be calculated by an algorithm for calculating image brightness to obtain the brightness of the initial scene image, and then determine whether the image meets a specified brightness threshold based on a set image brightness threshold, and then further determine whether the image meets a specified brightness threshold based on a set image brightness threshold. According to the brightness judgment result, it is determined whether the image clarity of the initial scene image has passed the inspection; it can also be calculated by an algorithm for calculating the image contrast, and then the image contrast is determined to meet the specified contrast threshold according to the set contrast threshold, and then further determined whether the image clarity of the initial scene image has passed the inspection according to the contrast judgment result; it can also be obtained according to the brightness judgment result, the contrast judgment result, the structure judgment result, etc. A clarity value is obtained, and then the image clarity of the initial scene image of the pre-loan enterprise is determined to pass the inspection according to whether the clarity value is within the preset clarity threshold range. This application does not limit this. Further, the server can determine that the image clarity of the initial scene image of the pre-loan enterprise has passed the inspection when the clarity parameter is within the preset clarity threshold range; conversely, the server can determine that the image clarity of the initial scene image of the pre-loan enterprise has failed the inspection when the clarity parameter exceeds the preset clarity threshold range.

[0050] The present application provides a method for identifying abnormal loan behavior of an enterprise. The server can obtain the clarity parameters of the initial scene image in the spatial domain and the frequency domain respectively through different algorithms, and determine whether the image clarity of the initial scene image of the pre-loan enterprise has passed the inspection based on whether the clarity parameters are within a preset clarity threshold range. Since it is relatively easy to obtain the clarity parameters of the initial scene image in the spatial domain and the frequency domain, and does not require a complex calculation process, by inspecting the clarity of the initial scene image in this way, the inspection result of the clarity of the initial scene image provided by the pre-loan enterprise can be quickly obtained, so as to further improve the efficiency of identifying the loan behavior of the pre-loan enterprise.

[0051] Optionally, the initial scene images of the pre-loan enterprise can be input into the clarity determination model one by one, and then the image clarity of the initial scene image can be determined based on the output result of the clarity model to determine whether it has passed the inspection. The clarity determination model can be obtained by the server through neural network model training based on a large number of cleaned images in the memory, and this application does not limit this.

[0052] In one embodiment, Figure 5 As shown, an optional method embodiment of performing image authenticity verification on an initial scene image provided by an embodiment of the present application, the method embodiment comprises the following steps:

[0053] Step S502, extracting overall image features and image noise features from the initial scene image of the pre-loan enterprise through a preset feature extraction network model;

[0054] Step S504: If the uniformity of the overall image features and the image noise features are both greater than the preset uniformity threshold, it is determined that the image authenticity check of the initial scene image of the pre-loan enterprise has passed.

[0055] The feature extraction network model may include a network model for extracting RGB feature values ​​of an image and a network model for extracting image noise. The network model for extracting RGB feature values ​​of an image may be obtained through deep learning using a neural network model such as ResNet, VGGNet, BERT, or a HyperNet model. If the initial scene image has not been tampered with, the overall features of the image should be evenly distributed and there will not be large differences. For example, the present application may use the HyperNet model to extract the RGB feature values ​​(i.e., overall features) of the initial scene image from the shallow, medium, and deep levels. Specifically, the corresponding feature information may be extracted from the first, third, and fifth convolution groups, respectively, and the feature information of the shallow and medium levels may be deconvolved so that the resolutions of both are the same as the resolution extracted from the third level. The features extracted from the three levels are then smoothed and fused to obtain the overall features of the initial scene image. The overall features extracted are then judged to determine whether the uniformity of the overall features is greater than a preset uniformity threshold, and if the uniformity of the overall features of the image is greater than the preset uniformity threshold, the image authenticity test of the initial scene image of the pre-loan enterprise is determined to have passed. Otherwise, it has failed. By extracting the overall features of the initial scene image through the HyperNet model, the extracted feature details are richer, and the contrast difference between the tampered area of ​​the image and the real area of ​​the image can be captured more accurately, thereby more accurately judging the authenticity of the current initial scene image.

[0056] At the same time, the server can also extract the noise features of the initial scene image based on the network model for image noise extraction. This is based on the principle that if the initial scene image has not been tampered with, the noise features of each area of ​​the image should be relatively evenly distributed and there will not be large differences. The noise features of each area of ​​the initial scene image can be extracted by using the network model for image noise extraction to compare the noise features of each area and determine whether the uniformity of the noise features of each area is greater than a preset uniformity threshold. If the uniformity of the noise features is less than the preset uniformity threshold, it is determined that the initial scene image is a tampered image, and the image authenticity test result of the initial image is failed, otherwise, it is failed.

[0057] The server may be a server that extracts the overall features of all the initial scene images through a network model that extracts RGB feature values ​​of the image to obtain an image authenticity verification result; and may also be a server that extracts the noise features of all the initial scene images through a network model that extracts image noise to obtain another image authenticity verification result, and then determines whether the image authenticity verification of the initial scene images provided by the pre-loan enterprise has passed based on the two verification results. Furthermore, if both verification results are passed, it may be determined that the image authenticity verification of the initial scene images of the pre-loan enterprise has passed; otherwise, if one of the two verification results is not passed, it may be determined that the image authenticity verification of the initial scene images of the pre-loan enterprise has failed.

[0058] The present application provides a method for identifying abnormal loan behavior of an enterprise. The server can obtain the overall characteristics and image noise characteristics of the initial scene image through a network model that extracts RGB feature values ​​of the image and a network model that extracts image noise, respectively, to determine the image authenticity verification result of the initial image according to the uniformity of the overall image characteristics and the image noise characteristics, and determine the authenticity verification result of the image from different dimensions, so that the authenticity verification result of the initial scene image is more accurate, so as to further ensure the accuracy of the subsequent identification of the loan behavior of the pre-loan enterprise.

[0059] Optionally, if the authenticity check of the initial scene image fails, the server may send a message to the pre-loan enterprise rejecting the loan application; the server may also send a re-upload prompt message to the pre-loan enterprise to instruct the pre-loan enterprise to re-upload the initial scene image if the authenticity check of the initial scene image fails. This application is not limited to this.

[0060] Optionally, the server may also verify the image authenticity of the initial scene image of the pre-loan enterprise through a preset authenticity trial network model. If the authenticity trial network model outputs a result of passing, it is determined that the image authenticity test of the initial scene image of the pre-loan enterprise has passed.

[0061] The authenticity trial network model can be obtained by the server through training of a neural network model based on a large number of real images stored in the memory. When the server needs to perform an image authenticity check on the initial scene images of the pre-loan enterprise, the initial scene images can be directly input into the pre-stored authenticity trial network model one by one, so as to determine whether there are tampered images in the initial scene images based on the results output by the authenticity trial network model, so as to realize the image authenticity check of the initial scene images. The authenticity trial network model can quickly determine the image authenticity check results of the initial scene images, thereby improving the efficiency of identifying the loan behavior of the pre-loan enterprise.

[0062] In one embodiment, Figure 6 As shown, an optional method embodiment of performing image repeatability inspection on an initial scene image provided by an embodiment of the present application is provided, and the method embodiment comprises the following steps:

[0063] Step S602, obtaining the similarity between any two images in the initial scene image of the pre-loan enterprise through a similarity comparison algorithm;

[0064] Step S604, if the number of normal images in the initial scene image of the pre-loan enterprise is greater than a preset number, it is determined that the image repeatability test of the initial scene image of the pre-loan enterprise has passed; wherein the normal image refers to an image whose similarity is less than a preset similarity threshold.

[0065] Among them, the similarity comparison algorithm is used to calculate the similarity of the two images to obtain the similarity between the two images. The similarity comparison algorithm can be a histogram method, a gray value method, a PSNR peak signal-to-noise ratio algorithm, etc., which is not limited in this application. The initial scene image is compared for similarity because the initial scene image includes a variety of different types of scene images. The credit institution needs to determine whether the loan behavior of the pre-loan enterprise is normal from multiple different dimensions. Therefore, if the pre-loan enterprise only repeatedly uploads one type of image, it will affect the credit institution's subsequent judgment on the loan behavior of the pre-loan enterprise. At the same time, if the pre-loan enterprise uploads a large number of similar or identical images for loan review under the clear regulations of the credit institution, the server can directly determine that the loan behavior of the pre-loan enterprise is abnormal, and no subsequent identification operation is required. The loan application of the pre-loan enterprise can be rejected to avoid wasting resources. Therefore, the server needs to calculate the similarity of the initial scene image to determine the proportion of similar images in the initial scene image, that is, the proportion of normal images, so as to further determine the image repeatability test result of the initial scene image of the pre-loan enterprise based on whether the number of normal images is greater than the preset number. It should be noted that the normal image refers to an image whose similarity with other images in the initial image is less than a preset similarity threshold. Then, further, if the number of long images is greater than the preset threshold, it is determined that the image repeatability test of the initial scene image of the pre-loan enterprise has passed, otherwise, it means that it has failed.

[0066] The present application provides a method for identifying abnormal loan behavior of an enterprise. The server in the method can verify the image repeatability of the initial scene image by calculating the similarity between any two images in the initial scene image, so as to avoid the pre-loan enterprise providing too many repetitive scene images, which is not conducive to improving the accuracy of identifying the loan behavior of the pre-loan enterprise.

[0067] Optionally, the server can also obtain the similarity of the initial scene image through a similarity comparison model. The similarity comparison model can be obtained by the server through historical similar images through neural network model training, which is not limited in this application. The similarity between the initial scene images can be quickly obtained through the similarity comparison model to improve the subsequent recognition efficiency of the loan behavior of the pre-loan enterprise.

[0068] In one embodiment, the server may be able to send a prompt message to the pre-loan enterprise, instructing the pre-loan enterprise to re-upload clear images, real images and missing images, if there is a situation that the image clarity check, image authenticity check and image repeatability check performed on the initial scene image through the above steps do not pass, for example, the image clarity check fails, the image authenticity check fails or the image repeatability check fails, so as to avoid subsequent recognition operations when the initial scene image does not meet the loan requirements, resulting in a waste of resources.

[0069] In one embodiment, this embodiment is an embodiment in which the loan behavior of the pre-loan enterprise can be determined to be abnormal when the server determines that the inspection result of either the image authenticity or the image repeatability of the initial scene image meets the abnormal loan conditions.

[0070] Among them, the strength of the failure of the inspection in this embodiment is greater than the strength of the failure of the above-mentioned inspection. For example, if the pre-loan enterprise provides 58 scene images, if the authenticity inspection results of less than 10 scene images fail, a prompt message can be sent to the pre-loan enterprise to instruct the pre-loan enterprise to re-upload; if the authenticity inspection results of more than 10 scene images fail, then the credit institution can directly determine that the loan behavior of the pre-loan enterprise is abnormal, that is, there is a greater risk of loan fraud, and directly reject the loan application of the pre-loan enterprise. Similarly, if the number of normal images is less than 30, a prompt message can be sent to the pre-loan enterprise to instruct the pre-loan enterprise to re-upload the missing scene images. If the number of normal images is greater than 30, then the credit institution can directly determine that the loan behavior of the pre-loan enterprise is abnormal, that is, there is a greater risk of loan fraud, and directly reject the loan application of the pre-loan enterprise. In this way, the preliminary screening of the loan behavior of the pre-loan enterprise is achieved to avoid waste of resources and improve the efficiency of identifying the loan behavior of the pre-loan enterprise.

[0071] In one embodiment, Figure 7 As shown, when multiple key enterprise loan information includes enterprise name information, enterprise logo information and enterprise account information, an optional method embodiment for matching multiple key enterprise loan information is provided in this embodiment. The implementation of the method includes the following steps:

[0072] Step S702, matching the enterprise name information, enterprise logo information, enterprise account information with the pre-stored standard information of the pre-loan enterprise; the standard information is obtained through official information and / or third-party public information;

[0073] Step S704: If the enterprise name information, enterprise logo information, and enterprise account information are all matched, it is determined that the enterprise information is matched successfully.

[0074] Among them, this application obtains multiple key information of enterprise loans through different key information recognition models, wherein multiple key information of enterprise loans includes enterprise name information, enterprise logo information and enterprise account information. After obtaining these key information of enterprise loans, the server can match the pre-stored standard information of the pre-loan enterprise with the above-mentioned enterprise loan application information to further determine the authenticity of the content of the initial scene information provided by the pre-loan enterprise, and realize the identification of the loan behavior of the pre-loan enterprise. It should be noted that the standard information of the pre-loan enterprise can be obtained by the server from the official website of the pre-loan enterprise, such as logging in to the official website matching the enterprise name; it can be obtained from a public website, such as Baidu Knows; it can also be obtained from the enterprise query software, such as from the enterprise check APP; it can also be obtained from other functional departments, such as from the audit department, the financial department, etc.; it can also be obtained after comprehensive results of multiple queries, and this application does not limit this. After obtaining the standard information of the pre-loan enterprise, the server stores the standard information in the corresponding memory address, and when a comparison is required, it extracts the standard information from the memory and matches it with the extracted key information of the enterprise loan, so as to determine whether the loan behavior of the pre-loan enterprise is normal according to the matching result. When the enterprise name information, enterprise logo information and enterprise account information are matched, it can be determined that the enterprise information is matched successfully, and further determined that the loan behavior of the pre-loan enterprise is normal. Otherwise, it is determined that the loan behavior of the pre-loan enterprise is abnormal. The server can also send a prompt message to the pre-loan enterprise to reject the loan application after determining that the loan behavior of the pre-loan enterprise is abnormal.

[0075] The present application provides a method for identifying abnormal loan behavior of an enterprise. The method matches standard information obtained from official information and / or third-party public information with key loan information of the enterprise, such as the enterprise name information, enterprise logo information and enterprise account information of the pre-loan enterprise. The authenticity of the content in the initial image provided by the enterprise can be further verified to achieve the identification of the loan behavior of the pre-loan enterprise. The matching process is simple and does not require complex calculation processes. The matching results can be obtained quickly, further improving the efficiency of identifying the loan behavior of the pre-loan enterprise. At the same time, the loan information of the pre-loan enterprise is further verified based on standard information obtained from other channels, and the identification of the loan behavior of the pre-loan enterprise is more accurate.

[0076] It should be understood that, although the various steps in the flowcharts involved in the above-mentioned embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps does not have a strict order restriction, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above-mentioned embodiments can include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.

[0077] Based on the same inventive concept, the embodiment of the present application also provides an enterprise abnormal loan behavior identification device for implementing the above-mentioned enterprise abnormal loan behavior identification method. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above-mentioned method, so the specific limitations of one or more embodiments of the enterprise abnormal loan behavior identification device provided below can refer to the above-mentioned limitations on the audit method, and will not be repeated here.

[0078] In one embodiment, Figure 8 As shown, a device for identifying abnormal loan behavior of an enterprise is provided, comprising: a verification module 802, an extraction module 804 and a matching determination module 806, wherein:

[0079] The verification module 802 is used to respond to the loan application of the pre-loan enterprise and perform a loan requirement verification on the initial scene image of the pre-loan enterprise to obtain a verification result;

[0080] The extraction module 804 is used to extract multiple key information of enterprise loans from the initial scene image through the key information recognition model if the inspection result shows that the initial scene image meets the loan requirements; the key information of enterprise loans is used to characterize the characteristics that distinguish the pre-loan enterprise from other pre-loan enterprises;

[0081] The matching determination module 806 is used to match the enterprise information of multiple enterprise loan key information. If the enterprise information is successfully matched, it is determined that the loan behavior of the pre-loan enterprise is normal.

[0082] In one embodiment, the inspection module 802 is specifically used to inspect the image clarity, image authenticity and image repeatability of the initial scene image of the pre-loan enterprise; when the image clarity, image authenticity and image repeatability are all inspected and passed, the inspection result is that the initial scene image meets the loan requirements.

[0083] In one embodiment, the inspection module 802 includes a first inspection unit,

[0084] The first inspection unit is used to obtain the clarity parameters of the initial scene image of the pre-loan enterprise in the spatial domain and the frequency domain respectively; when the clarity parameters are within the preset clarity threshold range, it is determined that the image clarity inspection of the initial scene image of the pre-loan enterprise has passed.

[0085] In one embodiment, the inspection module 802 includes a second inspection unit,

[0086] The second inspection unit is used to extract the overall image features and image noise features from the initial scene image of the pre-loan enterprise through a preset feature extraction network model; when the uniformity of the overall image features and the image noise features are both greater than a preset uniformity threshold, it is determined that the image authenticity inspection of the initial scene image of the pre-loan enterprise has passed.

[0087] In one embodiment, the above-mentioned second verification unit is also used to verify the image authenticity of the initial scene image of the pre-loan enterprise through a preset authenticity trial network model. If the output result of the authenticity trial network model is passed, it is determined that the image authenticity test of the initial scene image of the pre-loan enterprise has passed.

[0088] In one embodiment, the inspection module 802 includes a third inspection unit,

[0089] The third inspection unit is used to obtain the similarity between any two images in the initial scene image of the pre-loan enterprise through a similarity comparison algorithm; when the number of normal images in the initial scene image of the pre-loan enterprise is greater than a preset number, it is determined that the image repeatability inspection of the initial scene image of the pre-loan enterprise has passed; wherein the normal image represents a similarity less than a preset similarity.

[0090] In one embodiment, the above device further includes a sending module.

[0091] The sending module is used to send a prompt message to the pre-loan enterprise when any of the inspections of image clarity, image authenticity and image repeatability fails, and the prompt message is used to instruct the pre-loan enterprise to re-upload the initial scene image that meets the loan requirements.

[0092] In one embodiment, the above device further includes a determination module.

[0093] The determination module is used to determine that the loan behavior of the pre-loan enterprise is abnormal when the inspection result of either the image authenticity or the image repeatability meets the abnormal loan conditions.

[0094] In one embodiment, multiple key information of enterprise loans include enterprise name information, enterprise logo information and enterprise account information; the above-mentioned matching determination module 806 is specifically used to match the enterprise name information, enterprise logo information, enterprise account information with the pre-stored standard information of the pre-loan enterprise; the standard information is obtained through official information and / or third-party public information; when the enterprise name information, enterprise logo information and enterprise account information are all matched, it is determined that the enterprise information is matched successfully.

[0095] Each module in the above-mentioned audit device can be implemented in whole or in part by software, hardware or a combination thereof. Each module can be embedded in or independent of a processor in a computer device in the form of hardware, or can be stored in a memory in a computer device in the form of software, so that the processor can call and execute the operations corresponding to each module above.

[0096] In one embodiment, a computer device is provided. The computer device may be a server, and its internal structure diagram may be as follows: Fig. 9 As shown. The computer device includes a processor, a memory and a network interface connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used for scene image data. The network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a method for identifying abnormal loan behavior of an enterprise is implemented.

[0097] Those skilled in the art will understand that Fig. 9 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0098] In one embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and when the processor executes the computer program, the following steps are implemented:

[0099] In response to a loan application from a pre-loan enterprise, performing a loan requirement check on an initial scene image of the pre-loan enterprise to obtain a check result;

[0100] If the inspection result shows that the initial scene image meets the loan requirements, multiple key information of enterprise loans are extracted from the initial scene image through the key information recognition model; the key information of enterprise loans is used to characterize the characteristics that distinguish the pre-loan enterprise from other pre-loan enterprises;

[0101] Enterprise information matching is performed on key loan information of multiple enterprises. If the enterprise information matching is successful, it is determined that the loan behavior of the pre-loan enterprise is normal.

[0102] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0103] The image clarity, image authenticity and image repeatability of the initial scene image of the pre-loan enterprise are inspected; when the image clarity, image authenticity and image repeatability are all tested and passed, the test result is that the initial scene image meets the loan requirements.

[0104] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0105] Obtain clarity parameters of the initial scene image of the pre-loan enterprise in the spatial domain and the frequency domain respectively; when the clarity parameters are within a preset clarity threshold range, determine that the image clarity test of the initial scene image of the pre-loan enterprise has passed.

[0106] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0107] The overall image features and image noise features are extracted from the initial scene image of the pre-loan enterprise through a preset feature extraction network model; when the uniformity of the overall image features and the image noise features are both greater than the preset uniformity threshold, it is determined that the image authenticity test of the initial scene image of the pre-loan enterprise has passed.

[0108] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0109] By presetting the authenticity trial network model, the image authenticity of the initial scene image of the pre-loan enterprise is tested. If the output result of the authenticity trial network model is passed, it is determined that the image authenticity test of the initial scene image of the pre-loan enterprise has passed.

[0110] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0111] The similarity between any two images in the initial scene image of the pre-loan enterprise is obtained through a similarity comparison algorithm; when the number of normal images in the initial scene image of the pre-loan enterprise is greater than a preset number, it is determined that the image repeatability test of the initial scene image of the pre-loan enterprise has passed; wherein, the normal image represents a similarity less than a preset similarity.

[0112] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0113] In the event that any one of the inspections of image clarity, image authenticity and image repeatability fails, a prompt message is sent to the pre-loan enterprise, and the prompt message is used to instruct the pre-loan enterprise to re-upload the initial scene image that meets the loan requirements.

[0114] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0115] When the inspection result of either the authenticity of the image or the repeatability of the image meets the abnormal loan conditions, the loan behavior of the pre-loan enterprise is determined to be abnormal.

[0116] In one embodiment, the plurality of key enterprise loan information includes enterprise name information, enterprise logo information, and enterprise account information; when the processor executes the computer program, the following steps are also implemented:

[0117] The enterprise name information, enterprise logo information, enterprise account information and the pre-stored standard information of the pre-loan enterprise are matched; the standard information is obtained through official information and / or third-party public information; when the enterprise name information, enterprise logo information and enterprise account information are matched, the enterprise information is determined to be matched successfully.

[0118] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored, and when the computer program is executed by a processor, the following steps are implemented:

[0119] In response to a loan application from a pre-loan enterprise, performing a loan requirement check on an initial scene image of the pre-loan enterprise to obtain a check result;

[0120] If the inspection result shows that the initial scene image meets the loan requirements, multiple key information of enterprise loans are extracted from the initial scene image through the key information recognition model; the key information of enterprise loans is used to characterize the characteristics that distinguish the pre-loan enterprise from other pre-loan enterprises;

[0121] Enterprise information matching is performed on key loan information of multiple enterprises. If the enterprise information matching is successful, it is determined that the loan behavior of the pre-loan enterprise is normal.

[0122] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0123] The image clarity, image authenticity and image repeatability of the initial scene image of the pre-loan enterprise are inspected; when the image clarity, image authenticity and image repeatability are all tested and passed, the test result is that the initial scene image meets the loan requirements.

[0124] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0125] Obtain clarity parameters of the initial scene image of the pre-loan enterprise in the spatial domain and the frequency domain respectively; when the clarity parameters are within a preset clarity threshold range, determine that the image clarity test of the initial scene image of the pre-loan enterprise has passed.

[0126] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0127] The overall image features and image noise features are extracted from the initial scene image of the pre-loan enterprise through a preset feature extraction network model; when the uniformity of the overall image features and the image noise features are both greater than the preset uniformity threshold, it is determined that the image authenticity test of the initial scene image of the pre-loan enterprise has passed.

[0128] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0129] By presetting the authenticity trial network model, the image authenticity of the initial scene image of the pre-loan enterprise is tested. If the output result of the authenticity trial network model is passed, it is determined that the image authenticity test of the initial scene image of the pre-loan enterprise has passed.

[0130] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0131] The similarity between any two images in the initial scene image of the pre-loan enterprise is obtained through a similarity comparison algorithm; when the number of normal images in the initial scene image of the pre-loan enterprise is greater than a preset number, it is determined that the image repeatability test of the initial scene image of the pre-loan enterprise has passed; wherein, the normal image represents a similarity less than a preset similarity.

[0132] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0133] In the event that any one of the inspections of image clarity, image authenticity and image repeatability fails, a prompt message is sent to the pre-loan enterprise, and the prompt message is used to instruct the pre-loan enterprise to re-upload the initial scene image that meets the loan requirements.

[0134] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0135] When the inspection result of either the authenticity of the image or the repeatability of the image meets the abnormal loan conditions, the loan behavior of the pre-loan enterprise is determined to be abnormal.

[0136] In one embodiment, the plurality of key enterprise loan information includes enterprise name information, enterprise logo information, and enterprise account information; when the computer program is executed by the processor, the following steps are also implemented:

[0137] The enterprise name information, enterprise logo information, enterprise account information and the pre-stored standard information of the pre-loan enterprise are matched; the standard information is obtained through official information and / or third-party public information; when the enterprise name information, enterprise logo information and enterprise account information are matched, the enterprise information is determined to be matched successfully.

[0138] In one embodiment, a computer program product is provided, comprising a computer program, which, when executed by a processor, implements the following steps:

[0139] In response to a loan application from a pre-loan enterprise, performing a loan requirement check on an initial scene image of the pre-loan enterprise to obtain a check result;

[0140] If the inspection result shows that the initial scene image meets the loan requirements, multiple key information of enterprise loans are extracted from the initial scene image through the key information recognition model; the key information of enterprise loans is used to characterize the characteristics that distinguish the pre-loan enterprise from other pre-loan enterprises;

[0141] Enterprise information matching is performed on key loan information of multiple enterprises. If the enterprise information matching is successful, it is determined that the loan behavior of the pre-loan enterprise is normal.

[0142] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0143] The image clarity, image authenticity and image repeatability of the initial scene image of the pre-loan enterprise are inspected; when the image clarity, image authenticity and image repeatability are all tested and passed, the test result is that the initial scene image meets the loan requirements.

[0144] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0145] Obtain clarity parameters of the initial scene image of the pre-loan enterprise in the spatial domain and the frequency domain respectively; when the clarity parameters are within a preset clarity threshold range, determine that the image clarity test of the initial scene image of the pre-loan enterprise has passed.

[0146] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0147] The overall image features and image noise features are extracted from the initial scene image of the pre-loan enterprise through a preset feature extraction network model; when the uniformity of the overall image features and the image noise features are both greater than the preset uniformity threshold, it is determined that the image authenticity test of the initial scene image of the pre-loan enterprise has passed.

[0148] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0149] By presetting the authenticity trial network model, the image authenticity of the initial scene image of the pre-loan enterprise is tested. If the output result of the authenticity trial network model is passed, it is determined that the image authenticity test of the initial scene image of the pre-loan enterprise has passed.

[0150] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0151] The similarity between any two images in the initial scene image of the pre-loan enterprise is obtained through a similarity comparison algorithm; when the number of normal images in the initial scene image of the pre-loan enterprise is greater than a preset number, it is determined that the image repeatability test of the initial scene image of the pre-loan enterprise has passed; wherein, the normal image represents a similarity less than a preset similarity.

[0152] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0153] In the event that any one of the inspections of image clarity, image authenticity and image repeatability fails, a prompt message is sent to the pre-loan enterprise, and the prompt message is used to instruct the pre-loan enterprise to re-upload the initial scene image that meets the loan requirements.

[0154] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0155] When the inspection result of either the authenticity of the image or the repeatability of the image meets the abnormal loan conditions, the loan behavior of the pre-loan enterprise is determined to be abnormal.

[0156] In one embodiment, the plurality of key enterprise loan information includes enterprise name information, enterprise logo information, and enterprise account information; when the computer program is executed by the processor, the following steps are also implemented:

[0157] The enterprise name information, enterprise logo information, enterprise account information and the pre-stored standard information of the pre-loan enterprise are matched; the standard information is obtained through official information and / or third-party public information; when the enterprise name information, enterprise logo information and enterprise account information are matched, the enterprise information is determined to be matched successfully.

[0158] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but are not limited to this.

[0159] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0160] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A method for identifying abnormal loan behavior of enterprises. It is characterized in that The method comprises: In response to a loan application from a pre-loan enterprise, the image clarity, image authenticity and image repeatability of the initial scene image of the pre-loan enterprise are inspected, wherein the image clarity indicates whether the appearance presentation of the initial scene image provided by the pre-loan enterprise meets the requirements, the image authenticity indicates whether the scene image provided by the pre-loan enterprise is authentic, and the image repeatability indicates whether there are repeated images among a plurality of different initial scene images provided by the pre-loan enterprise; If the image clarity, image authenticity and image repeatability are all tested and passed, the test result is that the initial scene image meets the loan requirements; If the inspection result is that the initial scene image meets the loan requirements, a plurality of key enterprise loan information is extracted from the initial scene image through a key information recognition model; the key enterprise loan information is used to characterize the characteristics that distinguish the pre-loan enterprise from other pre-loan enterprises, and the plurality of key enterprise loan information includes enterprise name information, enterprise logo information, and enterprise account information; Performing enterprise information matching on the plurality of enterprise loan key information, wherein the enterprise information matching refers to at least verifying the authenticity and validity of the plurality of loan key information; If the enterprise information is matched successfully, it is determined that the loan behavior of the pre-loan enterprise is normal.

2. The method according to claim 1, It is characterized in that The image clarity of the initial scene image of the pre-loan enterprise is tested, including: Obtaining clarity parameters of the initial scene image of the pre-loan enterprise in the spatial domain and the frequency domain respectively; If the clarity parameter is within a preset clarity threshold range, it is determined that the image clarity test of the initial scene image of the pre-loan enterprise has passed.

3. The method according to claim 1, It is characterized in that The authenticity of the initial scene image of the pre-loan enterprise is checked, including: Extracting overall image features and image noise features from the initial scene image of the pre-loan enterprise through a preset feature extraction network model; If the uniformity of the overall image features and the image noise features are both greater than a preset uniformity threshold, it is determined that the image authenticity test of the initial scene image of the pre-loan enterprise has passed.

4. The method according to claim 1, It is characterized in that The authenticity of the initial scene image of the pre-loan enterprise is checked, including: By presetting the authenticity trial network model, the image authenticity of the initial scene image of the pre-loan enterprise is checked. If the output result of the authenticity trial network model is passed, it is determined that the image authenticity check of the initial scene image of the pre-loan enterprise has passed.

5. The method according to claim 1, It is characterized in that The image repeatability of the initial scene image of the pre-loan enterprise is tested, including: Obtaining the similarity between any two images in the initial scene image of the pre-loan enterprise through a similarity comparison algorithm; If the number of normal images in the initial scene image of the pre-loan enterprise is greater than a preset number, it is determined that the image repeatability test of the initial scene image of the pre-loan enterprise has passed; wherein the normal image represents a similarity less than a preset similarity.

6. The method according to any one of claims 2 to 5, It is characterized in that The method further comprises: If any one of the image clarity, image authenticity and image repeatability tests fails, a prompt message is sent to the pre-loan enterprise, and the prompt message is used to instruct the pre-loan enterprise to re-upload the initial scene image that meets the loan requirements.

7. The method according to any one of claims 2 to 5, It is characterized in that The method further comprises: If the inspection result of either the authenticity of the image or the repeatability of the image meets the abnormal loan conditions, it is determined that the loan behavior of the pre-loan enterprise is abnormal.

8. The method according to any one of claims 1 to 5, It is characterized in that The enterprise information matching of the plurality of key enterprise loan information includes: Matching the enterprise name information, the enterprise logo information, and the enterprise account information with the pre-stored standard information of the pre-loan enterprise; the standard information is obtained through official information and / or third-party public information; If the enterprise name information, the enterprise logo information, and the enterprise account information are all matched, it is determined that the enterprise information is matched successfully.

9. A device for identifying abnormal loan behavior of an enterprise, It is characterized in that The device comprises: A verification module is used to verify the image clarity, image authenticity and image repeatability of the initial scene image of the pre-loan enterprise in response to the loan application of the pre-loan enterprise, wherein the image clarity indicates whether the appearance presentation of the initial scene image provided by the pre-loan enterprise meets the requirements, the image authenticity indicates whether the scene image provided by the pre-loan enterprise is authentic, and the image repeatability indicates whether there are repeated images among a plurality of different initial scene images provided by the pre-loan enterprise; The inspection module is used to verify that the initial scene image meets the loan requirements when the image clarity, image authenticity and image repeatability are all verified; An extraction module is used to extract multiple key information of enterprise loans from the initial scene image through a key information recognition model if the inspection result shows that the initial scene image meets the loan requirements; the key information of enterprise loans is used to characterize the characteristics that distinguish the pre-loan enterprise from other pre-loan enterprises, and the multiple key information of enterprise loans includes enterprise name information, enterprise logo information and enterprise account information; A matching determination module, used for performing enterprise information matching on the plurality of enterprise loan key information, wherein the enterprise information matching refers to at least verifying the authenticity and validity of the plurality of loan key information; The matching determination module is used to determine that the loan behavior of the pre-loan enterprise is normal if the enterprise information is successfully matched.

10. A computer device comprising a memory and a processor, wherein the memory stores a computer program. It is characterized in that When the processor executes the computer program, the steps of the method according to any one of claims 1 to 8 are implemented.

Images