Weak Feature Background Visual Risk Control Method and System Based on Image Depth Analysis

Abstract

This application relates to a weak feature background visual risk control method based on image depth analysis. The method includes: obtaining an image to be detected; inputting the image to be detected into a multi-scale convolutional neural network to extract multi-scale features of the background region from the image to be detected through the multi-scale convolutional neural network; calculating the feature change difference of the background region based on the multi-scale features through the multi-scale convolutional neural network; and identifying the weak feature background in the background region according to the feature change difference through the multi-scale convolutional neural network. The solution provided by this application can automatically and accurately identify the weak feature background, reduce manual intervention, improve the robustness and accuracy of similar background detection, be applicable to a variety of complex scenarios, and has strong generalization ability.

Description

Technical Field

[0001] This application relates to the field of remote risk control technology, and particularly to a weak feature background visual risk control method and system based on image depth analysis. Background Art

[0002] Similar background detection refers to finding and marking image regions in a set of images that are similar to the background but have specific targets. Through similar background detection, the risk of group cases can be effectively reduced.

[0003] In similar background detection, weak feature backgrounds (such as white walls, single-color backgrounds) often lack obvious texture or feature information and are easily misidentified as having a high similarity to the target background. In this case, the detection system may misclassify or over-aggregate weak feature backgrounds, resulting in deviations in the similar background detection results and even affecting the execution of subsequent tasks.

[0004] In related technologies, it is necessary to manually pre-define weak feature background types, which not only increases the operation complexity but also is difficult to adapt to various complex scenarios. In addition, related technologies rely on single feature (such as texture) matching, which is not sensitive enough to weak feature backgrounds such as white walls, is easily interfered by complex backgrounds, and is prone to over-aggregation of weak feature backgrounds, and is not sufficient to cover all weak feature background scenarios, resulting in low recognition accuracy. Summary of the Invention

[0005] To solve or partially solve the problems existing in related technologies, this application provides a weak feature background visual risk control method based on image depth analysis, which can automatically and accurately identify weak feature backgrounds, reduce manual intervention, improve the robustness and accuracy of similar background detection, be applicable to various complex scenarios, and have strong generalization ability.

[0006] In the first aspect of this application, a weak feature background visual risk control method based on image depth analysis is provided. The method includes:

[0007] Obtain the image to be detected;

[0008] Input the image to be detected into a multi-scale convolutional neural network to extract multi-scale features of the background region from the image to be detected through the multi-scale convolutional neural network;

[0009] Calculate the feature change difference of the background region based on the multi-scale features through the multi-scale convolutional neural network;

[0010] Identify the weak feature background in the background region according to the feature change difference through the multi-scale convolutional neural network.

[0011] In one embodiment, the method further includes:

[0012] Remove the weak feature background from the background region to obtain a non-weak feature region;

[0013] Perform similar background detection on the non-weak feature region.

[0014] In one embodiment, the multi-scale convolutional neural network includes first convolutional kernels and second convolutional kernels of different sizes, and the multi-scale features include large-scale features and small-scale features; the extracting of the multi-scale features of the background region from the image to be detected by the multi-scale convolutional neural network includes:

[0015] Input the image to be detected into the first convolutional kernel and the second convolutional kernel respectively through the multi-scale convolutional neural network;

[0016] Extract a plurality of the large-scale features of the background region from the image to be detected through the first convolutional kernel; and,

[0017] Extract a plurality of the small-scale features of the background region from the image to be detected through the second convolutional kernel.

[0018] In one embodiment, the extracting of a plurality of the large-scale features of the background region from the image to be detected through the first convolutional kernel includes:

[0019] Divide the background region in the image to be detected into a plurality of first blocks through the first convolutional kernel;

[0020] Extract the feature information of each of the first blocks through the first convolutional kernel to obtain each of the large-scale features of the background region;

[0021] The extracting of a plurality of the small-scale features of the background region from the image to be detected through the second convolutional kernel includes:

[0022] Divide the background region in the image to be detected into a plurality of second blocks through the second convolutional kernel;

[0023] Extract the feature information of each of the second blocks through the second convolutional kernel to obtain each of the small-scale features of the background region;

[0024] Wherein, the size of the first block is larger than that of the second block, and the feature information of the first block and the feature information of the second block both include texture, color and shape.

[0025] In one embodiment, the calculating of the feature change difference of the background region based on the multi-scale features by the multi-scale convolutional neural network includes:

[0026] Based on the feature information of two adjacent first blocks, calculate the feature change difference between the two adjacent first blocks through the first convolution kernel; and,

[0027] Based on the feature information of two adjacent second blocks, calculate the feature change difference between the two adjacent second blocks through the second convolution kernel.

[0028] In one embodiment, the step of identifying the weak feature background in the background region by the multi-scale convolutional neural network according to the feature change difference includes:

[0029] Compare the feature change difference between two adjacent first blocks with a first change threshold through the first convolution kernel; and,

[0030] Compare the feature change difference between two adjacent second blocks with a second change threshold through the second convolution kernel;

[0031] If the feature change difference between two adjacent first blocks is less than or equal to the first change threshold, obtain a first recognition result output by the first convolution kernel; the first recognition result is used to represent that the two adjacent first blocks form a first weak feature region; and / or,

[0032] If the feature change difference between two adjacent second blocks is less than or equal to the second change threshold, obtain a second recognition result output by the second convolution kernel; the second recognition result is used to represent that the two adjacent second blocks form a second weak feature region;

[0033] Identify the sub-blocks jointly included in the first weak feature region and the second weak feature region as the weak feature background in the background region.

[0034] In one embodiment, the step of obtaining the image to be detected includes:

[0035] During the remote interaction with the lender, collect the visual data of the scene where the lender is located; wherein, the visual data includes at least one of image data and video data;

[0036] Perform a preprocessing operation on the visual data to obtain the image to be detected; wherein, the preprocessing operation includes at least one of image denoising and image normalization.

[0037] A second aspect of the present application provides a weak feature background visual risk control system based on image depth parsing, and the system includes:

[0038] An image acquisition module, configured to acquire an image to be detected;

[0039] A feature extraction module, configured to input the image to be detected into a multi-scale convolutional neural network, so as to extract multi-scale features of the background region from the image to be detected through the multi-scale convolutional neural network;

[0040] A feature change difference calculation module, configured to calculate the feature change difference of the background region based on the multi-scale features through the multi-scale convolutional neural network;

[0041] A weak feature background recognition module, configured to recognize the weak feature background in the background region according to the feature change difference through the multi-scale convolutional neural network.

[0042] A third aspect of the present application provides an electronic device, including:

[0043] A processor; and

[0044] A memory, on which executable code is stored, and when the executable code is executed by the processor, the processor is caused to execute the method as described above.

[0045] A fourth aspect of the present application provides a computer-readable storage medium, on which executable code is stored, and when the executable code is executed by a processor of an electronic device, the processor is caused to execute the method as described above.

[0046] The technical solution provided by the present application may include the following beneficial effects:

[0047] The solution provided by the present application obtains an image to be detected; inputs the image to be detected into a multi-scale convolutional neural network, so as to extract multi-scale features of the background region from the image to be detected through the multi-scale convolutional neural network; calculates the feature change difference of the background region based on the multi-scale features through the multi-scale convolutional neural network; and recognizes the weak feature background in the background region according to the feature change difference through the multi-scale convolutional neural network. The present application uses a multi-scale convolutional neural network to automatically recognize weak feature backgrounds, thereby reducing manual intervention and improving recognition efficiency. Moreover, the multi-scale convolutional neural network can extract multi-scale features of the background region, and the multi-scale features can accurately reflect the feature change difference of the background region. Furthermore, based on the feature change difference, weak feature backgrounds can be effectively recognized, thereby improving the robustness and accuracy of similar background detection, and it is also applicable to various complex scenarios and has strong generalization ability.

[0048] It should be understood that the above general description and subsequent detailed description are only exemplary and explanatory, and cannot limit the present application. Description of the Drawings

[0049] The above and other objects, features, and advantages of the present application will become more apparent by describing the exemplary embodiments of the present application in more detail in conjunction with the accompanying drawings, wherein, in the exemplary embodiments of the present application, the same reference numerals generally represent the same components.

[0050] Figure 1 is a schematic flowchart of a weak feature background visual risk control method based on image depth parsing shown in an embodiment of the present application;

[0051] Figure 2 is another schematic flowchart of a weak feature background visual risk control method based on image depth parsing shown in an embodiment of the present application;

[0052] Figure 3 is a schematic structural diagram of a weak feature background visual risk control system based on image depth parsing shown in an embodiment of the present application;

[0053] Figure 4 is a schematic structural diagram of an electronic device shown in an embodiment of the present application. Detailed Embodiments

[0054] Embodiments of the present application will be described in more detail below with reference to the accompanying drawings. Although the embodiments of the present application are shown in the drawings, it should be understood that the present application can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided to make the present application more thorough and complete, and to fully convey the scope of the present application to those skilled in the art.

[0055] The terms used in the present application are for the purpose of describing specific embodiments only and are not intended to limit the present application. The singular forms "a", "the", and "said" used in the present application and the appended claims are also intended to include the plural forms unless the context clearly dictates otherwise. It should also be understood that the term "and / or" as used herein refers to and encompasses any and all possible combinations of one or more of the associated listed items.

[0056] It should be understood that although the terms "first", "second", "third", etc. may be used in the present application to describe various information, such information should not be limited to these terms. These terms are only used to distinguish the same type of information from each other. For example, without departing from the scope of the present application, the first information may also be referred to as the second information, and similarly, the second information may also be referred to as the first information. Thus, the features defined with "first" and "second" may explicitly or implicitly include one or more of such features. In the description of the present application, "a plurality" means two or more unless otherwise specifically defined.

[0057] Similar background detection can effectively reduce the risk of case clusters. Specifically, during the loan application process, the applicant needs to pass a face recognition authentication. The background area in this process can reflect the scene where the applicant is located. By matching the features (such as texture) of the entire background area with the features (such as texture) of the target background (pre-collected backgrounds with the risk of case clusters), if the two match, it is determined that the background area is similar to the target background, and then it is evaluated that the scene where the applicant is located has the risk of case clusters.

[0058] However, if there are weak-feature backgrounds in the background area (such as white walls, single-color backgrounds), due to the lack of obvious texture or feature information in weak-feature backgrounds, they are easily misjudged as having a high similarity to the target background. Exemplarily, assume that multiple applicants apply for loans in different places within a short period of time, but their backgrounds are all white walls. This can easily cause the detection system to determine that these applicants are in the same place, and then wrongly aggregate these unrelated applicants together and evaluate that the scenes where they are located have the risk of case clusters. It can be seen that weak-feature backgrounds easily cause deviations in the results of similar background detection and even affect the execution of subsequent tasks. Therefore, it is necessary to identify weak-feature backgrounds to reduce their interference with similar background detection.

[0059] In related technologies for identifying weak-feature backgrounds, it is necessary to manually pre-define the types of weak-feature backgrounds. However, this not only increases the operation complexity but also is difficult to adapt to various complex scenarios. In addition, related technologies rely on single-feature (such as texture) matching. However, this is not sensitive enough to weak-feature backgrounds such as white walls, is easily interfered by complex backgrounds, and is prone to over-aggregation of weak-feature backgrounds, and is not sufficient to cover all weak-feature background scenarios, resulting in low recognition accuracy.

[0060] To address the above problems, the embodiments of the present application provide a visual risk control method for weak-feature backgrounds based on image depth parsing, which uses a multi-scale convolutional neural network to automatically identify weak-feature backgrounds, thereby reducing manual intervention and improving the recognition efficiency. Moreover, the multi-scale convolutional neural network can extract multi-scale features of the background area, and the multi-scale features can accurately reflect the feature change differences of the background area. Then, based on the feature change differences, weak-feature backgrounds can be effectively identified, thereby improving the robustness and accuracy of similar background detection, and it is also applicable to various complex scenarios and has strong generalization ability.

[0061] The technical solutions of the embodiments of the present application are described in detail below with reference to the accompanying drawings.

[0062] Figure 1 It is a schematic flowchart of the visual risk control method for weak-feature backgrounds based on image depth parsing shown in the embodiments of the present application.

[0063] See Figure 1 , the visual risk control method for weak-feature backgrounds based on image depth parsing of the present application includes:

[0064] S110, Obtain the image to be detected.

[0065] In the embodiments of the present application, it can be applied to a weak feature background visual risk control system based on image depth parsing (hereinafter referred to as "visual risk control system" for short). When a user needs video face signing or loan application, the user serves as a lender. The visual risk control system can obtain visual data of the scene where the lender is located, and the visual data can include at least one of static image data and dynamic video data. Since the similar background detection essentially analyzes the background area except for people (such as the lender), the visual risk control system can perform portrait segmentation on the visual data to obtain the image to be detected, and the image to be detected can only contain the background area except for people (such as the lender).

[0066] S120, Input the image to be detected into a multi-scale convolutional neural network to extract multi-scale features of the background area from the image to be detected through the multi-scale convolutional neural network.

[0067] In the embodiments of the present application, a multi-scale convolutional neural network (MSCNN) can be pre-trained. The multi-scale convolutional neural network is a deep learning model for processing features of different scales. In the embodiments of the present application, the trained multi-scale convolutional neural network can be deployed into the visual risk control system for online use by the visual risk control system.

[0068] In practical applications, the visual risk control system can call the pre-trained multi-scale convolutional neural network, and then input the image to be detected into the multi-scale convolutional neural network, so as to extract multi-scale features of the background area from the image to be detected through the multi-scale convolutional neural network.

[0069] It should be noted that compared with the traditional convolutional neural network that only performs convolution and pooling operations at one scale, the embodiments of the present application utilize the multi-scale convolutional neural network (MSCNN) to be able to operate at different scales. Specifically, by constructing multiple parallel convolutional paths, each path uses a convolutional kernel of a different size to capture feature information of different scales. For example, in the multi-scale convolutional neural network, a smaller convolutional kernel can capture local microscopic features of the background area from the image to be detected, and a larger convolutional kernel can capture global macroscopic features of the background area from the image to be detected. Therefore, the multi-scale features extracted in the embodiments of the present application can cover the details and overall information of the background area.

[0070] S130, Calculate the feature change difference of the background area based on the multi-scale features through the multi-scale convolutional neural network.

[0071] After extracting the multi-scale features of the background region, the multi-scale convolutional neural network can calculate the difference in feature changes of the background region based on the multi-scale features, and the difference in feature changes is used to characterize the degree of difference in the feature information of two adjacent blocks in the same background region.

[0072] S140. Identify the weak-feature background in the background region through the multi-scale convolutional neural network according to the difference in feature changes.

[0073] In one example, if the difference in feature changes is small or even close to zero, it indicates that there is no obvious change or even no change in the feature information of two adjacent blocks in the background region. Then, the multi-scale convolutional neural network can identify these two adjacent blocks as the weak-feature background in the background region. The weak-feature background can include but is not limited to: white walls, single-color backgrounds, and backgrounds with common decoration styles.

[0074] In another example, if the difference in feature changes is large, it indicates that there are large changes in the feature information of two adjacent blocks in the background region. Then, the multi-scale convolutional neural network can identify these two adjacent blocks as non-weak-feature regions in the background region, so that the visual risk control system can focus on analyzing the non-weak-feature regions subsequently, thereby greatly improving the recognition ability in complex backgrounds.

[0075] It can be seen from this example that the solution provided in this application is to obtain the image to be detected; input the image to be detected into the multi-scale convolutional neural network to extract the multi-scale features of the background region from the image to be detected through the multi-scale convolutional neural network; calculate the difference in feature changes of the background region based on the multi-scale features through the multi-scale convolutional neural network; identify the weak-feature background in the background region through the multi-scale convolutional neural network according to the difference in feature changes. This application uses the multi-scale convolutional neural network to automatically identify the weak-feature background, thereby reducing manual intervention and improving the recognition efficiency. Moreover, the multi-scale convolutional neural network can extract the multi-scale features of the background region, and the multi-scale features can accurately reflect the difference in feature changes of the background region. Furthermore, based on the difference in feature changes, the weak-feature background can be effectively identified, thereby improving the robustness and accuracy of similar background detection and being applicable to a variety of complex scenarios, with strong generalization ability.

[0076] Figure 2 It is another schematic flowchart of the weak-feature background visual risk control method based on image depth parsing shown in this application.

[0077] See Figure 2 The weak-feature background visual risk control method based on image depth parsing in this application includes:

[0078] S210. Obtain the image to be detected.

[0079] This step can be referred to the description in S110 and will not be elaborated here.

[0080] In one embodiment, obtaining the image to be detected may include:

[0081] During the remote interaction with the lender, collect the visual data of the scene where the lender is located; wherein, the visual data includes at least one of image data and video data; perform preprocessing operations on the visual data to obtain the image to be detected; wherein, the preprocessing operations include at least one of image denoising and image normalization.

[0082] When the user needs video face-to-face signing or loan application, the user serves as the lender. The lender can initiate a remote interaction request to the visual risk control system through a terminal device (such as a mobile phone, computer, etc.). The visual risk control system responds to the remote interaction request and conducts a remote interaction with the lender. The visual risk control system can perform visual risk control on the scene where the lender is located in the current remote interaction scenario.

[0083] In one example, after receiving the remote interaction request, the visual risk control system can send a collection instruction to the terminal device currently used by the lender. The collection instruction is used to instruct the camera of the terminal device to collect the visual data of the scene where the lender is located in real time. Therefore, during the remote interaction with the lender, the visual risk control system can obtain the visual data of the scene where the lender is located in real time.

[0084] In another example, the remote interaction request may carry the visual data of the scene where the lender is located uploaded by the lender. Therefore, during the remote interaction with the lender, the visual risk control system can obtain the visual data of the scene where the lender is located from the remote interaction request.

[0085] Wherein, the visual data may include at least one of static image data and dynamic video data, and both the image data and the video data can be in RGB format.

[0086] In order to enhance the depth representation of weak features, the visual risk control system can perform preprocessing operations such as image denoising and image normalization on the visual data, thereby obtaining the image to be detected. Among them, image denoising can remove noise interference. For example, in a scene with large noise, the visual risk control system can use Gaussian filtering to remove noise interference. Image normalization is to convert the visual data into the corresponding unique standard form of the image to be detected through a series of transformations (that is, using the invariant moments of the visual data to find a set of parameters to eliminate the influence of other transformation functions on the transformation of the visual data). The image to be detected has invariant characteristics for affine transformations such as translation, rotation, and scaling.

[0087] In addition, since the similar background detection is essentially an analysis of the background area other than humans (such as borrowers), the visual risk control system can also perform human segmentation on the visual data to obtain the image to be detected, and the image to be detected may only include the background area other than humans (such as borrowers).

[0088] S220, input the image to be detected into a multi-scale convolutional neural network to extract multi-scale features of the background area from the image to be detected through the multi-scale convolutional neural network.

[0089] This step can refer to the description in S120 and will not be elaborated here.

[0090] In one embodiment, the multi-scale convolutional neural network includes first convolutional kernels and second convolutional kernels of different sizes, and the multi-scale features include large-scale features and small-scale features; extracting multi-scale features of the background area from the image to be detected through the multi-scale convolutional neural network may include:

[0091] Input the image to be detected into the first convolutional kernel and the second convolutional kernel respectively through the multi-scale convolutional neural network; extract several large-scale features of the background area from the image to be detected through the first convolutional kernel; and extract several small-scale features of the background area from the image to be detected through the second convolutional kernel.

[0092] In the embodiment of the present application, multiple parallel convolutional paths can be pre-constructed in the multi-scale convolutional neural network (MSCNN), and each path uses convolutional kernels of different sizes, so that the multi-scale convolutional neural network can include convolutional kernels of different sizes to extract multi-scale features of the background area by using convolutional kernels of different sizes. Specifically, the convolutional kernels of different sizes at least include a relatively large first convolutional kernel and a relatively small second convolutional kernel, the multi-scale features of the background area at least include large-scale features and small-scale features, and the multi-scale convolutional neural network inputs the image to be detected into the first convolutional kernel and the second convolutional kernel respectively through two parallel convolutional paths. Since the size of the first convolutional kernel is larger than that of the second convolutional kernel, the first convolutional kernel is responsible for extracting several large-scale features of the background area from the image to be detected, and the second convolutional kernel is responsible for extracting several small-scale features of the background area from the image to be detected.

[0093] Among them, large-scale features are used to characterize the global macroscopic features of the background region, and small-scale features are used to characterize the local microscopic features of the background region. As an example, assuming that the color of the background region changes slowly, if only small-scale features are extracted, the color basically does not change microscopically, but the slow change of the color can be found macroscopically. As another example, assuming that the color of a small block in the background region changes drastically, if only large-scale features are extracted, the detailed information of this small block is likely to be lost macroscopically, but the drastic change in the color of this small block can be found microscopically. Therefore, the embodiments of the present application can cover the details and overall information of the background region by combining large-scale features and small-scale features.

[0094] In one embodiment, extracting a plurality of large-scale features of the background region from the image to be detected through a first convolution kernel may include:

[0095] Dividing the background region in the image to be detected into a plurality of first blocks through the first convolution kernel; extracting the feature information of each first block through the first convolution kernel to obtain each large-scale feature of the background region; wherein, the feature information of the first block includes texture, color, and shape.

[0096] Since similar background detection is essentially an analysis of the background region except for people (such as borrowers), and the image to be detected only contains the background region except for people (such as borrowers), the first convolution kernel can divide the background region in the image to be detected into a plurality of first blocks, and then extract the feature information of each first block. The feature information of each first block can be used as the large-scale feature of the background region. Among them, the feature information of each first block can include texture, color, and shape.

[0097] In one embodiment, extracting a plurality of small-scale features of the background region from the image to be detected through a second convolution kernel may include:

[0098] Dividing the background region in the image to be detected into a plurality of second blocks through the second convolution kernel; extracting the feature information of each second block through the second convolution kernel to obtain each small-scale feature of the background region; wherein, the feature information of the second block includes texture, color, and shape.

[0099] Since similar background detection is essentially an analysis of the background region except for people (such as borrowers), and the image to be detected only contains the background region except for people (such as borrowers), the second convolution kernel can divide the background region in the image to be detected into a plurality of second blocks, and then extract the feature information of each second block. The feature information of each second block can be used as the small-scale feature of the background region. Among them, the feature information of each second block can include texture, color, and shape.

[0100] It should be noted that since the size of the first convolution kernel is larger than that of the second convolution kernel, the size of the first block is larger than that of the second block.

[0101] S230. Calculate the feature change difference of the background region based on multi-scale features through a multi-scale convolutional neural network.

[0102] This step can refer to the description in S130 and will not be elaborated here.

[0103] In one embodiment, calculating the feature change difference of the background region based on multi-scale features through a multi-scale convolutional neural network may include:

[0104] Calculate the feature change difference between two adjacent first blocks based on the feature information of the two adjacent first blocks through the first convolution kernel; and calculate the feature change difference between two adjacent second blocks based on the feature information of the two adjacent second blocks through the second convolution kernel.

[0105] The feature change difference can be represented by the change in feature mean or the change in feature variance.

[0106] In one example, after the background region is divided into multiple first blocks, the first convolution kernel can calculate the texture mean, color mean, and shape mean of all pixel points in the same first block, and then add the texture mean, color mean, and shape mean to obtain the feature mean of this first block. In this way, the feature mean of each first block is obtained, and then based on the feature means of two adjacent first blocks, the change in the feature means of the two adjacent first blocks is calculated, and this change in the feature means can be used as the feature change difference between the two adjacent first blocks.

[0107] In another example, after the background region is divided into multiple first blocks, the first convolution kernel can calculate the texture variance, color variance, and shape variance of all pixel points in the same first block, and then add the texture variance, color variance, and shape variance to obtain the feature variance of this first block. In this way, the feature variance of each first block is obtained, and then based on the feature variances of two adjacent first blocks, the change in the feature variances of the two adjacent first blocks is calculated, and this change in the feature variances can be used as the feature change difference between the two adjacent first blocks.

[0108] Similarly, after the background region is divided into multiple second blocks, the second convolution kernel can calculate the texture mean, color mean, and shape mean of all pixel points in the same second block, and then add the texture mean, color mean, and shape mean to obtain the feature mean of the second block. In this way, the feature mean of each second block is obtained, and then based on the feature means of two adjacent second blocks, the change in the feature means of the two adjacent second blocks is calculated, and this change in the feature means can be used as the feature change difference between the two adjacent second blocks.

[0109] Similarly, after the background region is divided into multiple second blocks, the second convolution kernel can calculate the texture variance, color variance, and shape variance of all pixel points in the same second block, and then add the texture variance, color variance, and shape variance to obtain the feature variance of the second block. In this way, the feature variance of each second block is obtained, and then based on the feature variances of two adjacent second blocks, the change in the feature variances of the two adjacent second blocks is calculated, and this change in the feature variances can be used as the feature change difference between the two adjacent second blocks.

[0110] It should be noted that the first convolution kernel and the second convolution kernel can be processed in parallel, thereby improving the calculation efficiency.

[0111] S240. Identify the weak feature background in the background region according to the feature change difference through a multi-scale convolutional neural network.

[0112] This step can refer to the description in S140 and will not be elaborated here.

[0113] In one embodiment, identifying the weak feature background in the background region according to the feature change difference through a multi-scale convolutional neural network may include:

[0114] Compare the feature change difference between two adjacent first blocks with a first change threshold through the first convolution kernel; and compare the feature change difference between two adjacent second blocks with a second change threshold through the second convolution kernel; if the feature change difference between two adjacent first blocks is less than or equal to the first change threshold, obtain a first recognition result output by the first convolution kernel; the first recognition result is used to represent that two adjacent first blocks form a first weak feature region; and / or, if the feature change difference between two adjacent second blocks is less than or equal to the second change threshold, obtain a second recognition result output by the second convolution kernel; the second recognition result is used to represent that two adjacent second blocks form a second weak feature region; identify the sub-blocks jointly included in the first weak feature region and the second weak feature region as the weak feature background in the background region.

[0115] The embodiments of the present application can identify potential weak feature backgrounds in the background region based on the change threshold.

[0116] In one example, the first convolutional kernel can compare the feature change difference between two adjacent first blocks with a preset first change threshold. If the feature change difference between two adjacent first blocks ≤ the first change threshold, the first convolutional kernel outputs a first recognition result for the two adjacent first blocks, and the first recognition result is used to represent that the two adjacent first blocks constitute a first weak feature region. Or, if the feature change difference between two adjacent first blocks > the first change threshold, the first convolutional kernel outputs a third recognition result for the two adjacent first blocks, and the third recognition result is used to represent that the two adjacent first blocks constitute a first non-weak feature region.

[0117] In another example, the second convolutional kernel can compare the feature change difference between two adjacent second blocks with a preset second change threshold. If the feature change difference between two adjacent second blocks ≤ the second change threshold, the second convolutional kernel outputs a second recognition result for the two adjacent second blocks, and the second recognition result is used to represent that the two adjacent second blocks constitute a second weak feature region. If the feature change difference between two adjacent second blocks > the second change threshold, the second convolutional kernel outputs a fourth recognition result for the two adjacent second blocks, and the fourth recognition result is used to represent that the two adjacent second blocks constitute a second non-weak feature region.

[0118] After that, the embodiments of the present application can identify the sub-blocks jointly included in the first weak feature region and the second weak feature region as the weak feature background in the background region.

[0119] In one example, if the first recognition result and the second recognition result are obtained, the visual risk control system can first determine the sub-blocks jointly included in the first weak feature region corresponding to the first recognition result and the second weak feature region corresponding to the second recognition result, so as to identify the sub-blocks as the weak feature background in the background region. Exemplarily, assume that the first block A and the second block B jointly include the sub-block C. If the feature change difference between the first block A and its adjacent first block A' ≤ the first change threshold, and the feature change difference between the second block B and its adjacent second block B' ≤ the second change threshold, then the sub-block C can be identified as the weak feature background in the background region.

[0120] In addition, the embodiments of the present application can identify the sub-blocks jointly included in the first non-weak feature region and the second non-weak feature region as the non-weak feature region in the background region; or, the embodiments of the present application can identify the sub-blocks jointly included in the first non-weak feature region and the second weak feature region as the non-weak feature region in the background region; or, the embodiments of the present application can identify the sub-blocks jointly included in the first weak feature region and the second non-weak feature region as the non-weak feature region in the background region.

[0121] It can be seen that, compared with the related art where only a single feature (such as texture) is used to identify a weak-feature background, the embodiments of the present application identify the weak-feature background by integrating all features (such as texture, color, and shape), which can improve the recognition accuracy.

[0122] It can be seen that, compared with the related art where the features of the entire background area are analyzed, the embodiments of the present application analyze the entire background area by dividing it into large-scale features and small-scale features, which can cover the details and overall information of the background area, thereby further improving the recognition accuracy.

[0123] S250. Eliminate the weak-feature background from the background area to obtain a non-weak-feature area.

[0124] The visual risk control system can mark the weak-feature background as low priority and then eliminate the weak-feature background from the background area. The background area after elimination can be used as a non-weak-feature area.

[0125] Compared with the related art where high-precision similar background detection usually relies on expensive hardware devices, the embodiments of the present application eliminate the weak-feature background, so that the weak-feature background does not need to participate in the subsequent similar background detection (i.e., step S260). This can reduce the participation of invalid areas (i.e., weak-feature backgrounds), significantly reduce the hardware requirements and computational burden of the visual risk control system, thereby reducing the computational cost, and can also exclude the interference of the weak-feature background on the similar background detection, thereby improving the detection accuracy.

[0126] In addition, the embodiments of the present application can adjust the degree of discarding the weak-feature background by adjusting the first change threshold and the second change threshold. As an example, by increasing the first change threshold and the second change threshold, backgrounds with not very strong feature changes can be discarded, which can reduce the hardware resources required for subsequent similar background detection and reduce the computational cost. As another example, by decreasing the first change threshold and the second change threshold, only backgrounds with very smooth feature changes can be discarded, which can increase the hardware resources required for subsequent similar background detection and optimize the system performance. By adjusting the parameters based on different hardware resources, the embodiments of the present application can achieve a dynamic balance between resources and performance.

[0127] In addition, in the related art, there is a lack of an analysis method that effectively combines depth information for identifying weak feature backgrounds, resulting in insufficient expressiveness of weak features in three-dimensional scenes. In response, embodiments of the present application can achieve efficient recognition and classification of weak features in complex backgrounds by combining the depth information of images on the basis of feature extraction technology. Specifically, before eliminating the weak feature background, the visual risk control system can first obtain the depth information of two adjacent first blocks and the depth information of two adjacent second blocks collected by an ambient light sensor or a laser ranging device, and then further optimize the elimination result based on the comparison of the depth information of the two adjacent first blocks and the comparison of the depth information of the two adjacent second blocks, thereby further improving the recognition accuracy and recognition efficiency of the weak feature background.

[0128] S260, perform similar background detection on non-weak feature regions.

[0129] The visual risk control system can perform similar background detection on non-weak feature regions. Specifically, embodiments of the present application pre-collect various target backgrounds with the risk of case clustering, and then store these target backgrounds in the bottom library of the visual risk control system, so that various target backgrounds can be obtained from the bottom library in this step, and then the non-weak feature regions are respectively matched with various target backgrounds. If a non-weak feature region matches one of the target backgrounds, the visual risk control system can determine that there is a risk of case clustering in the scene where the lender is located, so the loan application of the lender can be rejected; if the non-weak feature region does not match any of the target backgrounds, the visual risk control system can determine that there is no risk of case clustering in the scene where the lender is located, so the loan application of the lender can be approved.

[0130] As can be seen from this example, the solution provided by this application is as follows: obtain the image to be detected; input the image to be detected into a multi-scale convolutional neural network to extract multi-scale features of the background region from the image to be detected through the multi-scale convolutional neural network; calculate the difference in feature changes of the background region based on the multi-scale features through the multi-scale convolutional neural network; identify the weak-feature background in the background region according to the difference in feature changes through the multi-scale convolutional neural network; remove the weak-feature background from the background region to obtain a non-weak-feature region; perform similar background detection on the non-weak-feature region. This application uses a multi-scale convolutional neural network to automatically identify the weak-feature background, thereby reducing manual intervention and improving the recognition efficiency. Moreover, the multi-scale convolutional neural network can extract multi-scale features of the background region, and the multi-scale features accurately reflect the difference in feature changes of the background region. Furthermore, based on the difference in feature changes, the weak-feature background can be effectively identified, thereby improving the robustness and accuracy of similar background detection, and it is also applicable to various complex scenarios and has strong generalization ability. Further, by reducing the participation of invalid regions (i.e., weak-feature backgrounds), this application can significantly reduce the hardware requirements and computational burden. In this way, only the non-weak-feature regions need to be analyzed intensively, which can greatly improve the recognition ability in complex backgrounds.

[0131] Corresponding to the foregoing method embodiment for implementing application functions, this application also provides a weak-feature background visual risk control system, an electronic device, a computer-readable storage medium, and corresponding embodiments based on image depth parsing.

[0132] Figure 3 It is a schematic structural diagram of a weak-feature background visual risk control system based on image depth parsing shown in an embodiment of this application.

[0133] See Figure 3 , a weak-feature background visual risk control system based on image depth parsing provided by this application, the system may include:

[0134] An image acquisition module 310, configured to acquire an image to be detected;

[0135] A feature extraction module 320, configured to input the image to be detected into a multi-scale convolutional neural network to extract multi-scale features of the background region from the image to be detected through the multi-scale convolutional neural network;

[0136] A feature change difference calculation module 330, configured to calculate the difference in feature changes of the background region based on the multi-scale features through the multi-scale convolutional neural network;

[0137] A weak-feature background recognition module 340, configured to identify the weak-feature background in the background region according to the difference in feature changes through the multi-scale convolutional neural network.

[0138] In one embodiment, the system may further include:

[0139] An elimination module, configured to eliminate the weak feature background from the background region to obtain a non-weak feature region;

[0140] A similar background detection module, configured to perform similar background detection on the non-weak feature region.

[0141] In one embodiment, the multi-scale convolutional neural network includes first convolutional kernels and second convolutional kernels of different sizes, and the multi-scale features include large-scale features and small-scale features; the feature extraction module 320 may include:

[0142] An input sub-module, configured to input the image to be detected into the first convolutional kernel and the second convolutional kernel respectively through the multi-scale convolutional neural network;

[0143] A large-scale feature extraction sub-module, configured to extract a plurality of large-scale features of the background region from the image to be detected through the first convolutional kernel; and,

[0144] A small-scale feature extraction sub-module, configured to extract a plurality of small-scale features of the background region from the image to be detected through the second convolutional kernel.

[0145] In one embodiment, the large-scale feature extraction sub-module may include:

[0146] A first block division unit, configured to divide the background region in the image to be detected into a plurality of first blocks through the first convolutional kernel;

[0147] A feature information extraction unit of the first block, configured to extract the feature information of each first block through the first convolutional kernel to obtain each large-scale feature of the background region;

[0148] The small-scale feature extraction sub-module may include:

[0149] A second block division unit, configured to divide the background region in the image to be detected into a plurality of second blocks through the second convolutional kernel;

[0150] A feature information extraction unit of the second block, configured to extract the feature information of each second block through the second convolutional kernel to obtain each small-scale feature of the background region;

[0151] Wherein, the size of the first block is larger than that of the second block, and the feature information of the first block and the feature information of the second block both include texture, color and shape.

[0152] In one embodiment, the feature change difference calculation module 330 may include:

[0153] A large-scale feature change difference calculation sub-module for calculating the feature change difference between two adjacent first blocks based on the feature information of the two adjacent first blocks through a first convolution kernel; and,

[0154] A small-scale feature change difference calculation sub-module for calculating the feature change difference between two adjacent second blocks based on the feature information of the two adjacent second blocks through a second convolution kernel.

[0155] In one embodiment, the weak feature background recognition module 340 may include:

[0156] A large-scale feature change difference comparison sub-module for comparing the feature change difference between two adjacent first blocks with a first change threshold through a first convolution kernel, and,

[0157] A small-scale feature change difference comparison sub-module for comparing the feature change difference between two adjacent second blocks with a second change threshold through a second convolution kernel;

[0158] A first recognition result obtaining sub-module for obtaining a first recognition result output by the first convolution kernel if the feature change difference between two adjacent first blocks is less than or equal to the first change threshold; the first recognition result is used to represent that two adjacent first blocks form a first weak feature region; and,

[0159] A second recognition result obtaining sub-module for obtaining a second recognition result output by the second convolution kernel if the feature change difference between two adjacent second blocks is less than or equal to the second change threshold; the second recognition result is used to represent that two adjacent second blocks form a second weak feature region;

[0160] A weak feature background recognition sub-module for recognizing the sub-blocks jointly included in the first weak feature region and the second weak feature region as the weak feature background in the background region.

[0161] In one embodiment, the image acquisition module 310 may include:

[0162] A visual data acquisition sub-module for acquiring visual data of the scene where the lender is located during the remote interaction with the lender; wherein, the visual data includes at least one of image data and video data;

[0163] A preprocessing operation sub-module for performing preprocessing operations on the visual data to obtain a to-be-detected image; wherein, the preprocessing operations include at least one of image denoising and image normalization.

[0164] As can be seen from this example, the solution provided by this application is as follows: obtain the image to be detected; input the image to be detected into a multi-scale convolutional neural network to extract multi-scale features of the background region from the image to be detected through the multi-scale convolutional neural network; calculate the feature change difference of the background region based on the multi-scale features through the multi-scale convolutional neural network; identify the weak feature background in the background region according to the feature change difference through the multi-scale convolutional neural network. This application uses a multi-scale convolutional neural network to automatically identify the weak feature background, thereby reducing manual intervention and improving the recognition efficiency. Moreover, the multi-scale convolutional neural network can extract multi-scale features of the background region, and the multi-scale features can accurately reflect the feature change difference of the background region. Furthermore, based on the feature change difference, the weak feature background can be effectively identified, thereby improving the robustness and accuracy of similar background detection, and it is also applicable to various complex scenarios and has strong generalization ability.

[0165] Regarding the system in the above embodiments, the specific manners in which each module performs operations have been described in detail in the embodiments related to the method, and will not be elaborated here.

[0166] Figure 4 It is a schematic structural diagram of an electronic device shown in an embodiment of this application.

[0167] See Figure 4 , the electronic device 400 includes a memory 410 and a processor 420.

[0168] The processor 420 may be a central processing unit (CPU), or may also be other general-purpose processors, digital signal processors (DSPs), application specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor, or the processor may also be any conventional processor, etc.

[0169] The memory 410 may include various types of storage units, such as system memory, read-only memory (ROM), and permanent storage devices. Among them, the ROM may store static data or instructions required by the processor 420 or other modules of the computer. The permanent storage device may be a readable and writable storage device. The permanent storage device may be a non-volatile storage device that does not lose the stored instructions and data even when the computer is powered off. In some embodiments, the permanent storage device employs a mass storage device (such as a magnetic or optical disk, flash memory) as the permanent storage device. In some other embodiments, the permanent storage device may be a removable storage device (such as a floppy disk, optical drive). The system memory may be a readable and writable storage device or a volatile readable and writable storage device, such as dynamic random access memory. The system memory may store some or all of the instructions and data required by the processor during operation. In addition, the memory 410 may include any combination of computer-readable storage media, including various types of semiconductor storage chips (such as DRAM, SRAM, SDRAM, flash memory, programmable read-only memory), and magnetic disks and / or optical disks may also be used. In some embodiments, the memory 410 may include a removable storage device that is readable and / or writable, such as a compact disc (CD), read-only digital versatile disc (such as DVD-ROM, dual-layer DVD-ROM), read-only Blu-ray disc, super density disc, flash memory card (such as SD card, min SD card, Micro-SD card, etc.), magnetic floppy disk, etc. The computer-readable storage medium does not include carrier waves and instantaneous electronic signals transmitted wirelessly or by wire.

[0170] Executable code is stored on the memory 410, and when the executable code is processed by the processor 420, it may cause the processor 420 to execute some or all of the methods described above.

[0171] In addition, the method according to the present application may also be implemented as a computer program or a computer program product, which includes computer program code instructions for executing some or all of the steps in the above method of the present application.

[0172] Alternatively, the present application may also be implemented as a computer-readable storage medium (or non-transitory machine-readable storage medium or machine-readable storage medium), on which executable code (or computer program or computer instruction code) is stored. When the executable code (or computer program or computer instruction code) is executed by a processor of an electronic device (or a server, etc.), it causes the processor to execute some or all of the steps of the above method according to the present application.

[0173] The embodiments of the present application have been described above. The above description is exemplary and not exhaustive, and is also not limited to the disclosed embodiments. Many modifications and variations are obvious to those of ordinary skill in the art without departing from the scope and spirit of the described embodiments. The choice of terms used herein is intended to best explain the principles of the embodiments, practical applications, or improvements to technologies in the market, or to enable other ordinary skilled persons in the art to understand the embodiments disclosed herein.

Claims

1. A visual risk control method for weak feature background based on image depth analysis, characterized in that: The method comprises: Acquire the image to be detected; Inputting the image to be detected into a multi-scale convolutional neural network to extract multi-scale features of a background area from the image to be detected through the multi-scale convolutional neural network; Calculating the feature change difference of the background area based on the multi-scale features by the multi-scale convolutional neural network; Identifying weak feature background in the background area according to the feature change difference by the multi-scale convolutional neural network; The multi-scale convolutional neural network includes a first convolution kernel and a second convolution kernel of different sizes, the first convolution kernel is used to divide the background area in the image to be detected into a plurality of first blocks, and the second convolution kernel is used to divide the background area in the image to be detected into a plurality of second blocks; The identifying the weak feature background in the background area according to the feature change difference by the multi-scale convolutional neural network includes: Comparing the feature change difference between two adjacent first blocks with a first change threshold through the first convolution kernel; and Compare the feature change difference between two adjacent second blocks with a second change threshold using the second convolution kernel; If the feature change difference between two adjacent first blocks is less than or equal to the first change threshold, a first recognition result output by the first convolution kernel is obtained; the first recognition result is used to characterize that the two adjacent first blocks constitute a first weak feature area; and / or, If the feature change difference between two adjacent second blocks is less than or equal to the second change threshold, a second recognition result output by the second convolution kernel is obtained; the second recognition result is used to indicate that the two adjacent second blocks constitute a second weak feature area; A sub-block commonly included in the first weak feature region and the second weak feature region is identified as a weak feature background in the background region.

2. The method according to claim 1, characterized in that The method further comprises: Eliminating the weak feature background from the background area to obtain a non-weak feature area; Perform similar background detection on the non-weak feature area.

3. The method according to claim 1, characterized in that The multi-scale features include large-scale features and small-scale features; The extracting multi-scale features of the background area from the image to be detected by the multi-scale convolutional neural network includes: Inputting the image to be detected into the first convolution kernel and the second convolution kernel respectively through the multi-scale convolution neural network; Extracting a plurality of the large-scale features of the background area from the image to be detected by using the first convolution kernel; and A plurality of the small-scale features of the background area are extracted from the image to be detected by using the second convolution kernel.

4. The method according to claim 3, characterized in that: The extracting a plurality of large-scale features of the background area from the image to be detected by using the first convolution kernel includes: Extracting feature information of each of the first blocks by using the first convolution kernel to obtain each of the large-scale features of the background area; The extracting a plurality of the small-scale features of the background area from the image to be detected by using the second convolution kernel includes: Extracting feature information of each of the second blocks by using the second convolution kernel to obtain each of the small-scale features of the background area; The size of the first block is larger than that of the second block, and the feature information of the first block and the feature information of the second block both include texture, color and shape.

5. The method according to claim 4, characterized in that The calculating the feature change difference of the background area based on the multi-scale features by the multi-scale convolutional neural network includes: Calculating the feature change difference between two adjacent first blocks based on feature information of two adjacent first blocks by using the first convolution kernel; and The feature change difference between two adjacent second blocks is calculated based on the feature information of two adjacent second blocks by using the second convolution kernel.

6. The method according to claim 1, characterized in that The step of acquiring the image to be detected comprises: During the remote interaction with the lender, collecting visual data of the scene in which the lender is located; wherein the visual data includes at least one of image data and video data; The visual data is preprocessed to obtain an image to be detected; wherein the preprocessing operation includes at least one of image denoising and image normalization.

7. A weak feature background visual wind control system based on image depth analysis, characterized in that: The system comprises: An image acquisition module, used for acquiring an image to be detected; A feature extraction module, used for inputting the image to be detected into a multi-scale convolutional neural network, so as to extract multi-scale features of a background area from the image to be detected through the multi-scale convolutional neural network; A feature change difference calculation module, used to calculate the feature change difference of the background area based on the multi-scale feature through the multi-scale convolutional neural network; A weak feature background recognition module, used to recognize the weak feature background in the background area according to the feature change difference through the multi-scale convolutional neural network; The multi-scale convolutional neural network includes a first convolution kernel and a second convolution kernel of different sizes, the first convolution kernel is used to divide the background area in the image to be detected into a plurality of first blocks, and the second convolution kernel is used to divide the background area in the image to be detected into a plurality of second blocks; The weak feature background recognition module comprises: A large-scale feature change difference comparison submodule is used to compare the feature change difference between two adjacent first blocks with a first change threshold through a first convolution kernel, and, A small-scale feature change difference comparison submodule, used for comparing the feature change difference between two adjacent second blocks with a second change threshold through a second convolution kernel; A first recognition result obtaining submodule is used to obtain a first recognition result output by a first convolution kernel if the feature change difference between two adjacent first blocks is less than or equal to a first change threshold; the first recognition result is used to characterize that the two adjacent first blocks constitute a first weak feature area; and / or, A second recognition result obtaining submodule is used to obtain a second recognition result output by a second convolution kernel if the feature change difference between two adjacent second blocks is less than or equal to a second change threshold; the second recognition result is used to characterize that the two adjacent second blocks constitute a second weak feature area; The weak feature background recognition submodule is used to recognize the sub-block contained in the first weak feature area and the second weak feature area as the weak feature background in the background area.

8. An electronic device, characterized in that: include: processor; as well as A memory having executable codes stored thereon, which, when executed by the processor, causes the processor to execute the method according to any one of claims 1 to 6.

9. A computer-readable storage medium having executable codes stored thereon, which, when executed by a processor of an electronic device, causes the processor to execute the method according to any one of claims 1 to 6.

Images