Collateral risk detection method and device, storage medium and electronic equipment

Abstract

The invention discloses a collateral risk detection method and device, a storage medium and electronic equipment. The method relates to the technical field of artificial intelligence, and comprises the steps: obtaining an initial terahertz image of a to-be-detected target collateral; performing feature enhancement processing on the initial terahertz image through a target network model to obtain a target terahertz image; performing classification processing on the target terahertz image through an image classification model to obtain target category information of the target collateral; and performing risk detection according to the target category information and the target attribute information corresponding to the target collateral to obtain a detection result. Through the risk detection method and device, the problem that the risk detection accuracy is low due to the fact that risk detection is carried out on the collateral in a manual mode in the related technology is solved.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and in particular, to a method and device for risk detection of collateral, a storage medium, and an electronic device. Background Art

[0002] In the mortgage loan business of financial institutions, the risk detection of collateral is a key link to ensure asset security. Traditionally, financial institutions mainly conduct open-box inspections of collateral manually, but this method has certain problems. Manual inspection requires a large amount of manpower and time, and the accuracy of manual inspection is limited by the professional level and experience of the evaluators. Therefore, there is a problem that the accuracy of risk detection is relatively low.

[0003] In view of the problem that the accuracy of risk detection is relatively low due to the manual method of risk detection of collateral in the related art, no effective solution has been proposed yet. Summary of the Invention

[0004] The main purpose of the present application is to provide a method and device for risk detection of collateral, a storage medium, and an electronic device, so as to solve the problem that the accuracy of risk detection is relatively low due to the manual method of risk detection of collateral in the related art.

[0005] To achieve the above object, according to one aspect of the present application, a method for risk detection of collateral is provided. The method includes: obtaining an initial terahertz image of a target collateral to be detected; performing feature enhancement processing on the initial terahertz image through a target network model to obtain a target terahertz image; performing classification processing on the target terahertz image through an image classification model to obtain target category information of the target collateral; and performing risk detection based on the target category information and target attribute information corresponding to the target collateral to obtain a detection result.

[0006] Further, performing feature enhancement processing on the initial terahertz image through a target network model to obtain a target terahertz image includes: processing the initial terahertz image through a first feature processing module in the target network model to obtain a first target feature vector; performing feature fusion processing on the first target feature vector through a second feature processing module in the target network model to obtain a second target feature vector; and processing the second target feature vector through a convolution module in the target network model to obtain the target terahertz image.

[0007] Further, processing the initial terahertz image through the first feature processing module in the target network model to obtain a first target feature vector includes: processing the initial terahertz image through the first convolutional layer and the first non-linear activation layer in the first feature processing module to obtain a first feature vector; extracting features from the first feature vector through the first attention layer in the first feature processing module to obtain a second feature vector; processing the second feature vector through the second convolutional layer in the first feature processing module to obtain a third feature vector; performing a splicing process on the initial terahertz image and the third feature vector through the splicing layer in the first feature processing module to obtain the first target feature vector.

[0008] Further, extracting features from the first feature vector through the first attention layer in the first feature processing module to obtain a second feature vector includes: performing a first dimension size transformation on the first feature vector to obtain a first transformed feature vector; performing a second dimension size transformation on the first feature vector to obtain a second transformed feature vector; changing the dimension sorting of the first feature vector to obtain a third transformed feature vector; performing a normalization process on the second transformed feature vector and the third transformed feature vector to obtain a first processed feature vector; performing a dimension size transformation on the first processed feature vector and the first transformed feature vector to obtain a second processed feature vector; performing a splicing process on the second processed feature vector and the first feature vector to obtain the second feature vector.

[0009] Further, performing feature fusion processing on the first target feature vector through the second feature processing module in the target network model to obtain a second target feature vector includes: processing the first target feature vector through the third convolutional layer and the second non-linear activation layer in the second feature processing module to obtain a fourth feature vector; extracting features from the fourth feature vector through the feature pyramid attention layer in the second feature processing module to obtain a fifth feature vector; processing the fifth feature vector through the fourth convolutional layer and the third non-linear activation layer in the second feature processing module to obtain a sixth feature vector; obtaining the second target feature vector based on the fourth feature vector and the sixth feature vector.

[0010] Further, the feature pyramid attention layer includes a plurality of feature extraction groups, each feature extraction group consists of a sampling residual block and an attention layer, and the fifth feature vector is obtained by performing feature extraction on the fourth feature vector through the feature pyramid attention layer in the second feature processing module, which includes: for each feature extraction group, performing feature extraction on the fourth feature vector through the sampling residual block and the attention layer in the feature extraction group to obtain a feature vector corresponding to each feature extraction group; and splicing the feature vectors corresponding to each feature extraction group to obtain the fifth feature vector.

[0011] Further, classifying the target terahertz image through an image classification model to obtain the target category information of the target collateral includes: performing grayscale processing on the target terahertz image to obtain a processed terahertz image; extracting shape features from the processed terahertz image to obtain a shape feature vector; and processing the shape feature vector to obtain the target category information of the target collateral.

[0012] To achieve the above object, according to another aspect of the present application, there is provided a risk detection device for collateral. The device includes: an acquisition unit for acquiring an initial terahertz image of a target collateral to be detected; a first processing unit for performing feature enhancement processing on the initial terahertz image through a target network model to obtain a target terahertz image; a second processing unit for classifying the target terahertz image through an image classification model to obtain the target category information of the target collateral; and a detection unit for performing risk detection based on the target category information and the target attribute information corresponding to the target collateral to obtain a detection result.

[0013] Further, the first processing unit includes: a first processing subunit for processing the initial terahertz image through the first feature processing module in the target network model to obtain a first target feature vector; a second processing subunit for performing feature fusion processing on the first target feature vector through the second feature processing module in the target network model to obtain a second target feature vector; and a third processing subunit for processing the second target feature vector through the convolution module in the target network model to obtain the target terahertz image.

[0014] Further, the first processing subunit includes: a first processing module, configured to process the initial terahertz image through a first convolutional layer and a first non-linear activation layer in the first feature processing module to obtain a first feature vector; a first extraction module, configured to perform feature extraction on the first feature vector through a first attention layer in the first feature processing module to obtain a second feature vector; a second processing module, configured to process the second feature vector through a second convolutional layer in the first feature processing module to obtain a third feature vector; a splicing module, configured to perform splicing processing on the initial terahertz image and the third feature vector through a splicing layer in the first feature processing module to obtain the first target feature vector.

[0015] Further, the extraction module includes: a first transformation sub-module, configured to perform a first dimension size transformation on the first feature vector to obtain a first transformed feature vector; a second transformation sub-module, configured to perform a second dimension size transformation on the first feature vector to obtain a second transformed feature vector; a first variation sub-module, configured to vary the dimension sorting of the first feature vector to obtain a third transformed feature vector; a processing sub-module, configured to perform normalization processing on the second transformed feature vector and the third transformed feature vector to obtain a first processed feature vector; a second variation sub-module, configured to perform a dimension size transformation on the first processed feature vector and the first transformed feature vector to obtain a second processed feature vector; a first splicing sub-module, configured to perform splicing processing on the second processed feature vector and the first feature vector to obtain the second feature vector.

[0016] Further, the second processing subunit includes: a third processing module, configured to process the first target feature vector through a third convolutional layer and a second non-linear activation layer in the second feature processing module to obtain a fourth feature vector; a second extraction module, configured to perform feature extraction on the fourth feature vector through a feature pyramid attention layer in the second feature processing module to obtain a fifth feature vector; a fourth processing module, configured to process the fifth feature vector through a fourth convolutional layer and a third non-linear activation layer in the second feature processing module to obtain a sixth feature vector; a determination module, configured to obtain the second target feature vector based on the fourth feature vector and the sixth feature vector.

[0017] Further, the feature pyramid attention layer includes a plurality of feature extraction groups, each feature extraction group consists of a sampling residual block and an attention layer, and the second extraction module includes: an extraction sub-module, for each feature extraction group, performing feature extraction on the fourth feature vector through the sampling residual block and the attention layer in the feature extraction group to obtain a feature vector corresponding to each feature extraction group; a second splicing sub-module, for splicing the feature vectors corresponding to each feature extraction group to obtain the fifth feature vector.

[0018] Further, the second processing unit includes: a fourth processing sub-unit, for grayscale processing the target terahertz image to obtain a processed terahertz image; an extraction sub-unit, for extracting shape features from the processed terahertz image to obtain a shape feature vector; a fifth processing sub-unit, for processing the shape feature vector to obtain the target category information of the target collateral.

[0019] To achieve the above object, according to one aspect of the present application, there is provided a computer-readable storage medium storing a program, wherein when the program runs, it controls the device where the storage medium is located to execute the risk detection method of the collateral described in any one of the above.

[0020] To achieve the above object, according to another aspect of the present application, there is further provided an electronic device, which includes one or more processors and a memory, and the memory is used to store the risk detection method of the collateral described in any one of the above implemented by the one or more processors.

[0021] In the embodiments of the present application, the following method is adopted: obtaining an initial terahertz image of a target collateral to be detected; performing feature enhancement processing on the initial terahertz image through a target network model to obtain a target terahertz image; performing classification processing on the target terahertz image through an image classification model to obtain target category information of the target collateral; and performing risk detection based on the target category information and target attribute information corresponding to the target collateral to obtain a detection result, which solves the technical problem in the related art that the accuracy of risk detection is relatively low when manually detecting the risk of collateral. In this solution, the initial terahertz image of the target collateral to be detected is adopted, and the initial terahertz image is subjected to feature enhancement through the target network model, and then the target terahertz image is subjected to classification processing through the image classification model to obtain the target category information of the target collateral. Finally, risk detection is performed according to the target category information and target attribute information corresponding to the target collateral to obtain a detection result. The internal structure and features of an object can be more accurately identified through the terahertz image, and the feature enhancement of the initial terahertz image through the target network model can make the image boundary information clearer, improve the accuracy of subsequent classification and recognition, and avoid unpacking and detecting the collateral manually, thereby achieving the technical effect of improving the risk detection efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] The drawings constituting a part of this application are used to provide a further understanding of this application. The schematic embodiments of this application and their descriptions are used to explain this application and do not constitute an improper limitation to this application. In the drawings:

[0023] Figure 1 shows a hardware structure block diagram of a computer terminal for implementing a method for risk detection of collateral;

[0024] Figure 2 is a flowchart of a method for risk detection of collateral provided according to an embodiment of this application;

[0025] Figure 3 is a schematic diagram of a target network model provided according to an embodiment of this application;

[0026] Figure 4 is a schematic diagram of a first feature processing module provided according to an embodiment of this application;

[0027] Figure 5 is a schematic diagram of a second feature processing module provided according to an embodiment of this application;

[0028] Figure 6 is a schematic diagram of a risk detection device for collateral provided according to an embodiment of this application;

[0029] Figure 7It is a structural block diagram of an electronic device according to an embodiment of the present application. Detailed implementation manners

[0030] In order to enable those skilled in the art to better understand the solution of the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present application.

[0031] It should be noted that the terms "first", "second", etc. in the description and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and do not necessarily need to be used to describe a specific order or sequence. It should be understood that such data used can be interchanged under appropriate circumstances so that the embodiments of the present application described here can be implemented in an order other than those illustrated or described here. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device including a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0032] First, some nouns or terms that appear in the process of describing the embodiments of the present application are applicable to the following explanations:

[0033] Terahertz image: Terahertz waves generally refer to electromagnetic radiation with a frequency range between microwaves and infrared rays, and its frequency range is about 0.1 to 10 THz. A terahertz image refers to an image generated using terahertz technology.

[0034] It should be noted that the information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data for display, data for analysis, etc.) collected in the present application are information and data that have been authorized by the user or fully authorized by all parties, and the collection, storage, use, processing, transmission, provision, disclosure, and application of the relevant data and other processing comply with relevant laws, regulations and standards, take necessary confidentiality measures, do not violate public order and good customs, and provide corresponding operation entrances for users to choose to authorize or refuse. For example, there is an interface between the present system and relevant users or institutions to provide corresponding operation entrances for users to choose to agree or refuse the automated decision-making results; if the user chooses to refuse, the expert decision-making process will be entered.

[0035] Embodiment 1

[0036] According to an embodiment of the present application, an embodiment of a method for risk detection of collateral is further provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. And although the logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in a different order than here.

[0037] The method embodiment provided by the first embodiment of the present application can be executed in a mobile terminal, a computer terminal or a similar computing device. Figure 1 The hardware structure block diagram of a computer terminal (or mobile device) for implementing the method for risk detection of collateral is shown. As Figure 1 shown, the computer terminal 10 (or mobile device) may include one or more (shown as 102a, 102b,..., 102n in the figure) processors 102 (the processor 102 may include, but is not limited to, a processing device such as a microprocessor MCU or a programmable logic device FPGA), a memory 104 for storing data, and a transmission device 106 for communication functions. In addition, it may further include: a display, an input / output interface (I / O interface), a universal serial bus (USB) port (which can be included as one of the ports of the BUS bus), a network interface, a power supply and / or a camera. Those of ordinary skill in the art can understand that Figure 1 the structure shown is only schematic and does not limit the structure of the above-mentioned electronic device. For example, the computer terminal 10 may further include more or fewer components than Figure 1 shown, or have a different configuration from Figure 1 shown.

[0038] It should be noted that the above one or more processors 102 and / or other data processing circuits are generally referred to as "data processing circuits" in this article. The data processing circuit can be embodied as software, hardware, firmware or any combination thereof in whole or in part. In addition, the data processing circuit can be a single independent processing module, or be incorporated in whole or in part into any one of the other elements in the computer terminal 10 (or mobile device). As involved in the embodiment of the present application, the data processing circuit is used for processor control (such as the selection of a variable resistor terminal path connected to an interface).

[0039] The memory 104 can be used to store software programs and modules of application software, such as the program instructions / data storage device corresponding to the risk detection method of the collateral in the embodiments of the present application. The processor 102 executes various functional applications and data processing by running the software programs and modules stored in the memory 104, that is, implements the above-mentioned risk detection method of the collateral. The memory 104 may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memories, or other non-volatile solid-state memories. In some instances, the memory 104 may further include a memory remotely disposed relative to the processor 102, and these remote memories can be connected to the computer terminal 10 through a network. Examples of the above network include but are not limited to the Internet, enterprise intranets, local area networks, mobile communication networks, and combinations thereof.

[0040] The transmission device 106 is used to receive or send data via a network. Specific examples of the above network may include a wireless network provided by a communication provider of the computer terminal 10. In one instance, the transmission device 106 includes a network adapter (Network Interface Controller, NIC), which can be connected to other network devices through a base station and thus can communicate with the Internet. In one instance, the transmission device 106 can be a radio frequency (RF) module, which is used to communicate with the Internet wirelessly.

[0041] The display can be a touch-screen liquid crystal display (LCD), and the liquid crystal display enables a user to interact with the user interface of the computer terminal 10 (or mobile device).

[0042] Under the above operating environment, the present application provides a risk detection method of the collateral as Figure 2 shown. Figure 2 is a flowchart of the risk detection method of the collateral according to Embodiment 1 of the present application.

[0043] Step S201, obtain an initial terahertz image of a target collateral to be detected.

[0044] Optionally, to determine the target collateral to be detected, it should be noted that the target collateral can be semi-finished products or finished products produced by a loan user (for example, a related enterprise) and have packaged collaterals. After determining the target collateral that needs to be risk-detected, obtain the initial terahertz image of the target collateral.

[0045] Step S202, perform feature enhancement processing on the initial terahertz image through a target network model to obtain a target terahertz image.

[0046] Optionally, after obtaining the initial terahertz image of the target collateral, in order to improve the resolution of the initial terahertz image, the initial terahertz image is input into the target network model, and the initial terahertz image is subjected to feature enhancement processing through the target network model to obtain the target terahertz image. For example, a low-level feature map is extracted from the initial terahertz image through the target network model. Then, this feature map is upsampled and passed as input to the next feature layer. This process is repeated until a high-resolution feature map is obtained, which contains rich image detail information. Finally, all the feature maps are aggregated to obtain the above-mentioned target terahertz image.

[0047] Step S203: Classify the target terahertz image through an image classification model to obtain the target category information of the target collateral.

[0048] Optionally, after obtaining the target terahertz image, the target terahertz image is classified through an image classification model. For example, features are extracted from the target terahertz image, and the target category information of the target collateral is determined based on the extracted features.

[0049] Step S204: Perform risk detection based on the target category information and the target attribute information corresponding to the target collateral to obtain a detection result.

[0050] Optionally, risk detection is performed based on the target category information and the target attribute information corresponding to the target collateral. It should be noted that the target attribute information may include, but is not limited to, weight, shape, quantity. The weight can be obtained through a corresponding weighing device, and the shape and quantity can be obtained by identifying the target terahertz image through a relevant deep learning model. For example, obtain the category information and relevant attribute information (such as weight, shape, quantity, etc.) provided by the loan user, and then compare the information provided by the loan user with the target category information and the target attribute information corresponding to the target collateral to obtain the detection result corresponding to the target collateral. For example, if the information is consistent, there is no risk. If the information is inconsistent, there is a risk.

[0051] In summary, the initial terahertz image of the target collateral to be detected is adopted, and the initial terahertz image is subjected to feature enhancement through the target network model. Then, the target terahertz image is classified through the image classification model to obtain the target category information of the target collateral. Finally, risk detection is performed according to the target category information and the target attribute information corresponding to the target collateral to obtain the detection result. Through the terahertz image, the internal structure and features of the item can be identified more accurately. Moreover, the feature enhancement of the initial terahertz image through the target network model can make the image boundary information clearer, improve the accuracy of subsequent classification and recognition, and avoid unpacking and detecting the collateral manually, thereby achieving the technical effect of improving the risk detection efficiency.

[0052] Optionally, in the collateral risk detection method provided in the embodiment of the present application, the process of obtaining the target terahertz image by performing feature enhancement processing on the initial terahertz image through the target network model includes: processing the initial terahertz image through the first feature processing module in the target network model to obtain the first target feature vector; performing feature fusion processing on the first target feature vector through the second feature processing module in the target network model to obtain the second target feature vector; and processing the second target feature vector through the convolution module in the target network model to obtain the target terahertz image.

[0053] In an optional embodiment, the process of obtaining the target terahertz image by performing feature enhancement processing on the initial terahertz image through the target network model includes the following steps: As Figure 3 shown, the target network model includes a first feature processing module, a second feature processing module, and a convolution module.

[0054] The first target feature vector is obtained by performing feature extraction processing on the initial terahertz image through the first feature processing module. It should be noted that the first target feature vector is a low-level feature map. Then, the second target feature vector is obtained by performing feature fusion processing on the first target feature vector through the second feature processing module. It should be noted that the second target feature vector is a high-resolution feature map. Finally, the target terahertz image is obtained by processing the second target feature vector through the convolution module.

[0055] Through the first feature processing module and the second feature processing module, the detailed information of the reconstructed terahertz image is completely retained, the image quality of the target terahertz image is improved, and the accuracy of subsequent classification and recognition is thereby improved.

[0056] Optionally, in the method for detecting the risk of collateral provided in the embodiments of the present application, processing the initial terahertz image through the first feature processing module in the target network model to obtain the first target feature vector includes: processing the initial terahertz image through the first convolutional layer and the first non-linear activation layer in the first feature processing module to obtain the first feature vector; extracting features from the first feature vector through the first attention layer in the first feature processing module to obtain the second feature vector; processing the second feature vector through the second convolutional layer in the first feature processing module to obtain the third feature vector; and splicing the initial terahertz image and the third feature vector through the splicing layer in the first feature processing module to obtain the first target feature vector.

[0057] In an optional embodiment, processing the initial terahertz image through the first feature processing module in the target network model to obtain the first target feature vector includes the following steps: As Figure 4 shown, the first feature processing module includes a first convolutional layer, a first non-linear activation layer, a first attention layer, a second convolutional layer, and a splicing layer.

[0058] Performing initial feature extraction on the initial terahertz image through the first convolutional layer and the first non-linear activation layer to obtain the above-mentioned first feature vector, then extracting features from the first feature vector through the first attention layer to obtain the second feature vector. The first attention layer can improve the feature representation of specific semantic information to explicitly model the interdependence between channels. To better fuse features, the second feature vector is processed through the second convolutional layer to obtain the third feature vector. Finally, a splicing operation is performed on the input feature and the output feature at the dimension level to achieve the fusion of low-level features and high-level features, that is, the initial terahertz image and the third feature vector are spliced through the splicing layer to obtain the first target feature vector.

[0059] The convolutional layer and the attention layer can more comprehensively extract the feature information of the initial terahertz image. Splicing the initial terahertz image and the third feature vector to achieve the fusion of low-level features and high-level features can effectively avoid losing the detailed features of the image.

[0060] Optionally, in the risk detection method of the collateral provided in the embodiments of the present application, the second feature vector is obtained by extracting features from the first feature vector through the first attention layer in the first feature processing module, including: performing a first dimensionality size transformation on the first feature vector to obtain a first transformed feature vector; performing a second dimensionality size transformation on the first feature vector to obtain a second transformed feature vector; changing the dimensionality order of the first feature vector to obtain a third transformed feature vector; performing a normalization process on the second transformed feature vector and the third transformed feature vector to obtain a first processed feature vector; performing a dimensionality size transformation on the first processed feature vector and the first transformed feature vector to obtain a second processed feature vector; and performing a concatenation process on the second processed feature vector and the first feature vector to obtain a second feature vector.

[0061] In an alternative embodiment, the second feature vector is obtained by extracting features from the first feature vector through the first attention layer in the first feature processing module, including the following steps: The first attention layer can be composed of multiple convolutional layers, and the convolutional layers are used to perform transformation operations on the dimensionality size and dimensionality order of the feature vector, so as to obtain the second feature vector. For example, performing a first dimensionality size transformation on the first feature vector to obtain a first transformed feature vector, performing a second dimensionality size transformation on the first feature vector to obtain a second transformed feature vector, and changing the dimensionality order of the first feature vector to obtain a third transformed feature vector.

[0062] After obtaining the above first transformed feature vector, second transformed feature vector, and third transformed feature vector, first multiply the second transformed feature vector and the third transformed feature vector, and then perform a normalization process on the multiplied feature vector to obtain a first processed feature vector. Then multiply the first processed feature vector and the first transformed feature vector, and then perform a dimensionality size transformation on the multiplied feature vector. It should be noted that the dimensionality size can be the same as the first dimensionality size, or the second dimensionality size, or the same as both. Finally, perform a concatenation process on the second processed feature vector and the first feature vector to obtain a second feature vector.

[0063] Through the above transformation operations of the dimensionality size and dimensionality order, the feature representation of specific semantic information can be improved to explicitly model the interdependence between channels and improve the accuracy of the second feature vector.

[0064] Optionally, in the method for detecting the risk of the collateral provided in the embodiments of the present application, the second target feature vector is obtained by performing feature fusion processing on the first target feature vector through the second feature processing module in the target network model, including: the first target feature vector is processed by the third convolutional layer and the second non-linear activation layer in the second feature processing module to obtain a fourth feature vector; the fourth feature vector is subjected to feature extraction by the feature pyramid attention layer in the second feature processing module to obtain a fifth feature vector; the fifth feature vector is processed by the fourth convolutional layer and the third non-linear activation layer in the second feature processing module to obtain a sixth feature vector; and the second target feature vector is obtained based on the fourth feature vector and the sixth feature vector.

[0065] In an alternative embodiment, the steps of obtaining the second target feature vector by performing feature fusion processing on the first target feature vector through the second feature processing module in the target network model include: as Figure 5 shown, the second feature processing module includes a third convolutional layer, a second non-linear activation layer, a feature pyramid attention layer, and a fourth convolutional layer and a third non-linear activation layer.

[0066] The first target feature vector is subjected to convolutional calculation and gradient processing by the third convolutional layer and the second non-linear activation layer to obtain a fourth feature vector, and then the fourth feature vector is subjected to multi-scale feature extraction by the feature pyramid attention layer to obtain a fifth feature vector. After obtaining the fifth feature vector, the fifth feature vector is again subjected to convolutional calculation and gradient processing by the fourth convolutional layer and the third non-linear activation layer to obtain a sixth feature vector. Finally, the fourth feature vector and the sixth feature vector are added together to obtain the final second target feature vector.

[0067] The feature pyramid attention layer can obtain feature vectors at different scales, improving the accuracy of subsequent determination of the second target feature vector.

[0068] Optionally, in the method for detecting the risk of the collateral provided in the embodiments of the present application, the feature pyramid attention layer includes a plurality of feature extraction groups, and each feature extraction group is composed of a sampling residual block and an attention layer. The steps of obtaining the fifth feature vector by performing feature extraction on the fourth feature vector through the feature pyramid attention layer in the second feature processing module include: for each feature extraction group, the fourth feature vector is subjected to feature extraction by the sampling residual block and the attention layer in the feature extraction group to obtain the feature vector corresponding to each feature extraction group; and the feature vectors corresponding to each feature extraction group are subjected to splicing processing to obtain the fifth feature vector.

[0069] In an alternative embodiment, the feature pyramid attention layer includes a plurality of feature extraction groups, and each feature extraction group consists of a sampling residual block and an attention layer. When performing feature extraction on the fourth feature vector through the feature pyramid attention layer in the second feature processing module, the following steps are included: For each feature extraction group, perform feature extraction on the fourth feature vector through the sampling residual block and the attention layer in the feature extraction group to obtain the feature vector corresponding to each feature extraction group. It should be noted that the scale features of the feature vectors corresponding to each feature extraction group are different. After obtaining the feature vectors corresponding to each of the above-mentioned feature extraction groups, perform feature fusion on the feature vectors corresponding to each feature extraction group to obtain the final fifth feature vector.

[0070] Modeling the correlation of different scale features through the sampling residual block and the attention layer improves the accuracy of determining the fifth feature vector.

[0071] Optionally, in the method for detecting the risk of the collateral provided in the embodiment of the present application, classifying the target terahertz image through an image classification model to obtain the target category information of the target collateral includes: performing grayscale processing on the target terahertz image to obtain the processed terahertz image; extracting shape feature vectors from the processed terahertz image; and processing the shape feature vectors to obtain the target category information of the target collateral.

[0072] In an alternative embodiment, classifying the target terahertz image through an image classification model includes the following steps: performing grayscale processing on the target terahertz image to obtain the processed terahertz image, then extracting shape feature vectors from the processed terahertz image, and finally obtaining the target category information of the target collateral according to the shape feature vectors.

[0073] The accuracy of classification is improved through the target terahertz image with enhanced features, thereby achieving the effect of improving the accuracy of risk detection.

[0074] In an alternative embodiment, in order to improve the reconstruction quality of the terahertz image, the above-mentioned target network model can be obtained by using the training method of a generative adversarial network.

[0075] The risk detection method for mortgaged property provided by the embodiments of the present application includes: obtaining an initial terahertz image of a target mortgaged property to be detected; performing feature enhancement processing on the initial terahertz image through a target network model to obtain a target terahertz image; performing classification processing on the target terahertz image through an image classification model to obtain target category information of the target mortgaged property; and performing risk detection based on the target category information and the target attribute information corresponding to the target mortgaged property to obtain a detection result, which solves the technical problem in the related art that the accuracy of risk detection is relatively low when manually detecting the risk of mortgaged property. In this solution, an initial terahertz image of the target mortgaged property to be detected is adopted, and the initial terahertz image is subjected to feature enhancement through a target network model, and then the target terahertz image is subjected to classification processing through an image classification model to obtain the target category information of the target mortgaged property. Finally, risk detection is performed according to the target category information and the target attribute information corresponding to the target mortgaged property to obtain a detection result. The internal structure and features of an object can be more accurately identified through a terahertz image, and the feature enhancement of the initial terahertz image through the target network model can make the image boundary information clearer, improve the accuracy of subsequent classification and recognition, and avoid unpacking and detecting the mortgaged property manually, thereby achieving the technical effect of improving the risk detection efficiency.

[0076] It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although the logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in a different order than here.

[0077] Embodiment 2

[0078] The embodiments of the present application also provide a risk detection device for mortgaged property. It should be noted that the risk detection device for mortgaged property in the embodiments of the present application can be used to execute the risk detection method for mortgaged property provided by the embodiments of the present application. The following introduces the risk detection device for mortgaged property provided by the embodiments of the present application.

[0079] According to the embodiments of the present application, there is also provided a device for implementing the above-mentioned risk detection method for mortgaged property, as Figure 6 shown, the device includes: an acquisition unit 601, a first processing unit 602, a second processing unit 603, and a detection unit 604.

[0080] The acquisition unit 601 is configured to acquire an initial terahertz image of a target mortgaged property to be detected;

[0081] The first processing unit 602 is configured to perform feature enhancement processing on the initial terahertz image through a target network model to obtain a target terahertz image;

[0082] A second processing unit 603, configured to classify the target terahertz image through an image classification model to obtain target category information of the target collateral;

[0083] A detection unit 604, configured to perform risk detection based on the target category information and the target attribute information corresponding to the target collateral to obtain a detection result.

[0084] The risk detection device for collateral provided by the embodiment of the present application obtains an initial terahertz image of the target collateral to be detected through an acquisition unit 601; a first processing unit 602 performs feature enhancement processing on the initial terahertz image through a target network model to obtain a target terahertz image; a second processing unit 603 classifies the target terahertz image through an image classification model to obtain target category information of the target collateral; a detection unit 604 is configured to perform risk detection based on the target category information and the target attribute information corresponding to the target collateral to obtain a detection result, solving the technical problem in the related art that risk detection of collateral is performed manually, resulting in relatively low accuracy of risk detection. In this solution, an initial terahertz image of the target collateral to be detected is used, and the initial terahertz image is subjected to feature enhancement through a target network model, and then the target terahertz image is classified through an image classification model to obtain target category information of the target collateral. Finally, risk detection is performed based on the target category information and the target attribute information corresponding to the target collateral to obtain a detection result. The internal structure and features of an item can be more accurately identified through a terahertz image, and feature enhancement of the initial terahertz image through a target network model can make the image boundary information clearer, improving the accuracy of subsequent classification and recognition, and avoiding unpacking and detecting the collateral manually, thereby achieving the technical effect of improving the risk detection efficiency.

[0085] Optionally, in the risk detection device for collateral provided by the embodiment of the present application, the first processing unit includes: a first processing subunit, configured to process the initial terahertz image through a first feature processing module in the target network model to obtain a first target feature vector; a second processing subunit, configured to perform feature fusion processing on the first target feature vector through a second feature processing module in the target network model to obtain a second target feature vector; a third processing subunit, configured to process the second target feature vector through a convolution module in the target network model to obtain a target terahertz image.

[0086] Optionally, in the risk detection device for collateral provided in the embodiments of the present application, the first processing subunit includes: a first processing module, configured to process the initial terahertz image through the first convolutional layer and the first non-linear activation layer in the first feature processing module to obtain a first feature vector; a first extraction module, configured to perform feature extraction on the first feature vector through the first attention layer in the first feature processing module to obtain a second feature vector; a second processing module, configured to process the second feature vector through the second convolutional layer in the first feature processing module to obtain a third feature vector; a splicing module, configured to perform splicing processing on the initial terahertz image and the third feature vector through the splicing layer in the first feature processing module to obtain a first target feature vector.

[0087] Optionally, in the risk detection device for collateral provided in the embodiments of the present application, the extraction module includes: a first transformation sub-module, configured to perform a first dimension size transformation on the first feature vector to obtain a first transformed feature vector; a second transformation sub-module, configured to perform a second dimension size transformation on the first feature vector to obtain a second transformed feature vector; a first variation sub-module, configured to change the dimension sorting of the first feature vector to obtain a third transformed feature vector; a processing sub-module, configured to perform normalization processing on the second transformed feature vector and the third transformed feature vector to obtain a first processed feature vector; a second variation sub-module, configured to perform dimension size transformation on the first processed feature vector and the first transformed feature vector to obtain a second processed feature vector; a first splicing sub-module, configured to perform splicing processing on the second processed feature vector and the first feature vector to obtain a second feature vector.

[0088] Optionally, in the risk detection device for collateral provided in the embodiments of the present application, the second processing subunit includes: a third processing module, configured to process the first target feature vector through the third convolutional layer and the second non-linear activation layer in the second feature processing module to obtain a fourth feature vector; a second extraction module, configured to perform feature extraction on the fourth feature vector through the feature pyramid attention layer in the second feature processing module to obtain a fifth feature vector; a fourth processing module, configured to process the fifth feature vector through the fourth convolutional layer and the third non-linear activation layer in the second feature processing module to obtain a sixth feature vector; a determination module, configured to obtain a second target feature vector based on the fourth feature vector and the sixth feature vector.

[0089] Optionally, in the risk detection device for collateral provided in the embodiments of the present application, the feature pyramid attention layer includes a plurality of feature extraction groups, and each feature extraction group is composed of a sampling residual block and an attention layer. The second extraction module includes: an extraction sub-module, configured to, for each feature extraction group, perform feature extraction on the fourth feature vector through the sampling residual block and the attention layer in the feature extraction group to obtain a feature vector corresponding to each feature extraction group; and a second splicing sub-module, configured to splice the feature vectors corresponding to each feature extraction group to obtain a fifth feature vector.

[0090] Optionally, in the risk detection device for collateral provided in the embodiments of the present application, the second processing unit includes: a fourth processing sub-unit, configured to perform grayscale processing on the target terahertz image to obtain a processed terahertz image; an extraction sub-unit, configured to extract shape features from the processed terahertz image to obtain a shape feature vector; and a fifth processing sub-unit, configured to process the shape feature vector to obtain target category information of the target collateral.

[0091] It should be noted here that the above-mentioned obtaining unit 601, first processing unit 602, second processing unit 603, and detection unit 604 correspond to steps S201 to S204 in Embodiment 1. The functions of the four units are the same as those of the corresponding steps in terms of implementation examples and application scenarios, but are not limited to the content disclosed in Embodiment 1 above. It should be noted that the above-mentioned modules or units may be hardware components or software components stored in a memory (for example, memory 104) and processed by one or more processors (for example, processors 102a, 102b,..., 102n). The above-mentioned modules may also be part of a device and can run in the computer terminal 10 provided in Embodiment 1.

[0092] Embodiment 3

[0093] Embodiments of the present application may provide an electronic device. Figure 7 It is a structural block diagram of an electronic device according to an embodiment of the present application. As Figure 7 shown, the electronic device may include: one or more ( Figure 7 only one is shown in the figure) processors 702, a memory 704, a storage controller, and a peripheral interface, where the peripheral interface is connected to a radio frequency module, an audio module, and a display.

[0094] Among them, the memory can be used to store software programs and modules, such as the program instructions / modules corresponding to the methods and devices in the embodiments of the present application. The processor executes various functional applications and data processing by running the software programs and modules stored in the memory, that is, the above-mentioned method is implemented. The memory may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memories. In some instances, the memory may further include a memory remotely provided with respect to the processor, and these remote memories may be connected to the terminal through a network. Examples of the above-mentioned network include but are not limited to the Internet, enterprise intranet, local area network, mobile communication network, and combinations thereof.

[0095] The processor can call the information and application programs stored in the memory through the transmission device to perform the following steps: obtaining an initial terahertz image of the target collateral to be detected; performing feature enhancement processing on the initial terahertz image through the target network model to obtain a target terahertz image; performing classification processing on the target terahertz image through the image classification model to obtain target category information of the target collateral; and performing risk detection based on the target category information and the target attribute information corresponding to the target collateral to obtain a detection result.

[0096] The processor can call the information and application programs stored in the memory through the transmission device to perform the following steps: performing feature enhancement processing on the initial terahertz image through the target network model to obtain a target terahertz image, including: processing the initial terahertz image through the first feature processing module in the target network model to obtain a first target feature vector; performing feature fusion processing on the first target feature vector through the second feature processing module in the target network model to obtain a second target feature vector; and processing the second target feature vector through the convolution module in the target network model to obtain a target terahertz image.

[0097] The processor can call the information and application programs stored in the memory through the transmission device to perform the following steps: processing the initial terahertz image through the first feature processing module in the target network model to obtain a first target feature vector, including: processing the initial terahertz image through the first convolutional layer and the first non-linear activation layer in the first feature processing module to obtain a first feature vector; performing feature extraction on the first feature vector through the first attention layer in the first feature processing module to obtain a second feature vector; processing the second feature vector through the second convolutional layer in the first feature processing module to obtain a third feature vector; and performing splicing processing on the initial terahertz image and the third feature vector through the splicing layer in the first feature processing module to obtain a first target feature vector.

[0098] The processor can call the information and application programs stored in the memory through the transmission device to execute the following steps: extracting features from the first feature vector through the first attention layer in the first feature processing module to obtain a second feature vector, including: performing a first-dimensionality size transformation on the first feature vector to obtain a first transformed feature vector; performing a second-dimensionality size transformation on the first feature vector to obtain a second transformed feature vector; changing the dimension sorting of the first feature vector to obtain a third transformed feature vector; performing a normalization process on the second transformed feature vector and the third transformed feature vector to obtain a first processed feature vector; performing a dimensionality size transformation on the first processed feature vector and the first transformed feature vector to obtain a second processed feature vector; performing a concatenation process on the second processed feature vector and the first feature vector to obtain a second feature vector.

[0099] The processor can call the information and application programs stored in the memory through the transmission device to execute the following steps: performing feature fusion processing on the first target feature vector through the second feature processing module in the target network model to obtain a second target feature vector, including: processing the first target feature vector through the third convolutional layer and the second non-linear activation layer in the second feature processing module to obtain a fourth feature vector; extracting features from the fourth feature vector through the feature pyramid attention layer in the second feature processing module to obtain a fifth feature vector; processing the fifth feature vector through the fourth convolutional layer and the third non-linear activation layer in the second feature processing module to obtain a sixth feature vector; obtaining the second target feature vector based on the fourth feature vector and the sixth feature vector.

[0100] The processor can call the information and application programs stored in the memory through the transmission device to execute the following steps: The feature pyramid attention layer includes multiple feature extraction groups, and each feature extraction group consists of a sampling residual block and an attention layer. Extracting features from the fourth feature vector through the feature pyramid attention layer in the second feature processing module to obtain a fifth feature vector, including: for each feature extraction group, extracting features from the fourth feature vector through the sampling residual block and the attention layer in the feature extraction group to obtain a feature vector corresponding to each feature extraction group; performing a concatenation process on the feature vectors corresponding to each feature extraction group to obtain a fifth feature vector.

[0101] The processor can call the information and application programs stored in the memory through the transmission device to execute the following steps: classifying the target terahertz image through an image classification model to obtain the target category information of the target collateral, including: performing grayscale processing on the target terahertz image to obtain a processed terahertz image; extracting shape features from the processed terahertz image to obtain a shape feature vector; processing the shape feature vector to obtain the target category information of the target collateral.

[0102] Those of ordinary skill in the art can understand that Figure 7 the structure shown is only illustrative, and the electronic device can also be a smart phone (such as an Android phone, an iOS phone, etc.), a tablet computer, a handheld computer, and terminal devices such as Mobile Internet Devices (MID), PAD, etc. Figure 7 It does not limit the structure of the above-mentioned electronic device. For example, the electronic device may further include more or fewer components (such as a network interface, a display device, etc.) than those shown in Figure 7 , or have a different configuration from that shown in Figure 7 .

[0103] Those of ordinary skill in the art can understand that all or part of the steps in the various methods of the above embodiments can be completed by instructing the hardware related to the terminal device through a program, and the program can be stored in a computer-readable storage medium. The storage medium may include: a flash drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disc, etc.

[0104] Embodiment 4

[0105] The embodiment of the present application also provides a storage medium. Optionally, in this embodiment, the above storage medium can be used to store the program code executed by the risk detection method of the collateral provided in the first embodiment above.

[0106] Optionally, in this embodiment, the above storage medium can be located in any one of the computer terminals in the computer terminal group in the computer network, or in any one of the mobile terminals in the mobile terminal group.

[0107] The present application also provides a computer program product, which is suitable for executing a program for the steps of the risk detection method of the collateral when executed on a data processing device.

[0108] The serial numbers of the above embodiments of the present application are only for description and do not represent the advantages and disadvantages of the embodiments.

[0109] In the above embodiments of the present application, the descriptions of the various embodiments have their own emphases. For the parts not detailed in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0110] In several embodiments provided by this application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are merely illustrative. For example, the division of the units is only a logical function division. In actual implementation, there may be other division methods. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed coupling or direct coupling or communication connection between each other can be through some interfaces. The indirect coupling or communication connection of units or modules can be in electrical or other forms.

[0111] The units described as separate components may or may not be physically separated. The components displayed as units may or may not be physical units, that is, they can be located in one place, or they can be distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0112] In addition, in each embodiment of this application, each functional unit can be integrated in a processing unit, or each unit can exist physically alone, or two or more units can be integrated in one unit. The above-mentioned integrated units can be implemented in the form of hardware or in the form of software functional units.

[0113] If the above-mentioned integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in each embodiment of this application. The foregoing storage medium includes: various media such as USB flash drives, read-only memories (ROMs), random access memories (RAMs), mobile hard disks, magnetic disks, or optical discs that can store program codes.

[0114] The above is only the preferred embodiment of this application. It should be noted that for those of ordinary skill in the art, without departing from the principle of this application, several improvements and refinements can still be made, and these improvements and refinements should also be regarded as the protection scope of this application.

Claims

1. A risk detection method for collateral, characterized in that: include: Acquire an initial terahertz image of the target collateral to be detected; Performing feature enhancement processing on the initial terahertz image through a target network model to obtain a target terahertz image; Classify the target terahertz image through an image classification model to obtain target category information of the target collateral; A risk detection is performed based on the target category information and the target attribute information corresponding to the target collateral to obtain a detection result.

2. The method according to claim 1, characterized in that Performing feature enhancement processing on the initial terahertz image through the target network model to obtain a target terahertz image includes: Processing the initial terahertz image by a first feature processing module in the target network model to obtain a first target feature vector; Performing feature fusion processing on the first target feature vector by a second feature processing module in the target network model to obtain a second target feature vector; The second target feature vector is processed by a convolution module in the target network model to obtain the target terahertz image.

3. The method according to claim 2, characterized in that Processing the initial terahertz image by a first feature processing module in the target network model to obtain a first target feature vector includes: Processing the initial terahertz image through a first convolutional layer and a first nonlinear activation layer in the first feature processing module to obtain a first feature vector; Performing feature extraction on the first feature vector through the first attention layer in the first feature processing module to obtain a second feature vector; Processing the second feature vector through the second convolutional layer in the first feature processing module to obtain a third feature vector; The initial terahertz image and the third feature vector are spliced ​​together through the splicing layer in the first feature processing module to obtain the first target feature vector.

4. The method according to claim 3, characterized in that Performing feature extraction on the first feature vector by the first attention layer in the first feature processing module to obtain a second feature vector includes: Performing a first dimension size transformation on the first feature vector to obtain a first transformed feature vector; Performing a second dimension size transformation on the first feature vector to obtain a second transformed feature vector; Changing the dimensional order of the first feature vector to obtain a third transformed feature vector; Normalizing the second transformed feature vector and the third transformed feature vector to obtain the first processed feature vector; Performing dimension transformation on the first processed feature vector and the first transformed feature vector to obtain a second processed feature vector; The second processed feature vector and the first feature vector are concatenated to obtain the second feature vector.

5. The method according to claim 2, characterized in that: Performing feature fusion processing on the first target feature vector by a second feature processing module in the target network model to obtain a second target feature vector includes: Processing the first target feature vector through the third convolution layer and the second nonlinear activation layer in the second feature processing module to obtain a fourth feature vector; Performing feature extraction on the fourth feature vector through a feature pyramid attention layer in the second feature processing module to obtain a fifth feature vector; Processing the fifth eigenvector through the fourth convolutional layer and the third nonlinear activation layer in the second feature processing module to obtain a sixth eigenvector; The second target feature vector is obtained according to the fourth feature vector and the sixth feature vector.

6. The method according to claim 5, characterized in that The feature pyramid attention layer includes a plurality of feature extraction groups, each of which is composed of a sampling residual block and an attention layer. The feature extraction of the fourth feature vector is performed through the feature pyramid attention layer in the second feature processing module to obtain a fifth feature vector including: For each feature extraction group, extract features of the fourth feature vector through a sampling residual block and an attention layer in the feature extraction group to obtain a feature vector corresponding to each feature extraction group; The feature vectors corresponding to each feature extraction group are concatenated to obtain the fifth feature vector.

7. The method according to claim 1, characterized in that The target terahertz image is classified by an image classification model to obtain target category information of the target collateral, including: gray-scale the target terahertz image to obtain a processed terahertz image; Extracting shape features from the processed terahertz image to obtain a shape feature vector; The shape feature vector is processed to obtain target category information of the target collateral.

8. A risk detection device for collateral, characterized in that: include: An acquisition unit, used for acquiring an initial terahertz image of a target collateral to be detected; A first processing unit, configured to perform feature enhancement processing on the initial terahertz image through a target network model to obtain a target terahertz image; A second processing unit is used to classify the target terahertz image through an image classification model to obtain target category information of the target collateral; The detection unit is used to perform risk detection based on the target category information and the target attribute information corresponding to the target collateral to obtain a detection result.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a stored executable program, wherein when the executable program is executed, the device where the computer-readable storage medium is located is controlled to execute the risk detection method for collateral described in any one of claims 1 to 7.

10. An electronic device, characterized in that: include: A memory storing an executable program; A processor is used to run the program, wherein the risk detection method for collateral described in any one of claims 1 to 7 is executed when the program is run.

Images