Fraud risk identification method and device, computer equipment and storage medium

Abstract

The invention belongs to the field of artificial intelligence technology and financial science and technology, and relates to a fraud risk identification method and device, computer equipment and a storage medium, and the method comprises the steps: obtaining a to-be-detected image; performing target segmentation on the to-be-detected image to obtain a background region, a target region of the subject matter and a trunk region; extracting target global features according to the target region, extracting trunk local features according to the trunk region, and extracting background global features according to the background region; according to the target global feature, the trunk local feature and the background global feature, performing retrieval in a historical database to obtain a target retrieval result and a background retrieval result; determining whether the subject matter is a repeated claim settlement object according to the target retrieval result; and if yes, determining a fraud risk type according to a background retrieval result. In addition, the invention also relates to a block chain technology, and data such as a to-be-detected image can be stored in a block chain. According to the invention, digital processing can be carried out on the claim settlement auditing process, and the efficiency and accuracy of fraud risk identification are improved.

Description

Technical Field

[0001] The present application relates to the fields of artificial intelligence technology and financial technology, and in particular to a fraud risk identification method, apparatus, computer equipment and storage medium. Background Art

[0002] In the field of animal and plant breeding insurance, as farmers' insurance awareness increases, it has become a common practice to insure animals and plants to resist the impact of property losses caused by force majeure. However, there is a common fraud risk in the claims process, that is, repeated claims for the subject matter. This fraud usually involves tampering with part of the content of the same livestock (such as cattle or pigs) and repeating the claim fraud in different cases, causing economic losses to the insurance company.

[0003] In order to reduce the losses caused by fraudulent claims, the current practice is usually to send staff to the site for verification, take photos and upload them to the system for further verification by a large number of sales staff in the massive data. However, this traditional manual verification method is not only labor-intensive and inefficient, but also susceptible to fraudulent behavior such as on-site verification personnel and user fraud, thus failing to effectively prevent property losses.

[0004] Therefore, there is an urgent need to digitize the existing claims review process to improve the efficiency and accuracy of fraud risk identification. Summary of the invention

[0005] The purpose of the embodiments of the present application is to propose a fraud risk identification method, device, computer equipment and storage medium, which can digitize the claims review process and improve the efficiency and accuracy of fraud risk identification.

[0006] In order to solve the above technical problems, the present application embodiment provides a fraud risk identification method, which adopts the following technical solution:

[0007] Acquire an image to be detected, where the image to be detected includes at least one target object;

[0008] Performing target segmentation on the image to be detected to obtain a background area, a target area of ​​the target object, and a torso area;

[0009] Extracting the global features of the target from the target area, extracting the local features of the trunk from the trunk area, and extracting the global features of the background from the background area;

[0010] According to the target global feature, the torso local feature and the background global feature, a search is performed in a pre-constructed historical database to obtain a target search result and a background search result;

[0011] Determining whether the subject matter is a duplicate claim object according to the target search result;

[0012] If the subject matter is a duplicate claim object, the fraud risk type is determined based on the background search results.

[0013] In order to solve the above technical problems, the embodiment of the present application also provides a fraud risk identification device, which adopts the following technical solution:

[0014] An image acquisition module, used to acquire an image to be detected, wherein the image to be detected includes at least one target object;

[0015] A region segmentation module is used to perform target segmentation on the image to be detected to obtain a background region, a target region of the target object, and a trunk region;

[0016] A feature extraction module, configured to extract a target global feature based on the target area, extract a torso local feature based on the torso area, and extract a background global feature based on the background area;

[0017] A feature retrieval module, used to search in a pre-built historical database according to the target global feature, the torso local feature and the background global feature, to obtain a target retrieval result and a background retrieval result;

[0018] An object determination module, used to determine whether the subject matter is a duplicate claim object according to the target search result;

[0019] The type determination module is used to determine the fraud risk type based on the background search result if the subject matter is a repeated claim object.

[0020] In order to solve the above technical problems, an embodiment of the present application also provides a computer device, which adopts the technical solution described below: the computer device includes a memory and a processor, the memory stores computer-readable instructions, and the processor implements the steps of the fraud risk identification method as described in any one of the above items when executing the computer-readable instructions.

[0021] In order to solve the above technical problems, an embodiment of the present application also provides a computer-readable storage medium, which adopts the following technical solution: the computer-readable storage medium stores computer-readable instructions, and when the computer-readable instructions are executed by the processor, the steps of the fraud risk identification method as described in any one of the above items are implemented.

[0022] Compared with the prior art, the embodiments of the present application have the following beneficial effects:

[0023] The embodiment of the present application performs target segmentation on the image to be detected, accurately distinguishes the background area, the target area of ​​the target object and the trunk area, and ensures the accuracy and pertinence of subsequent feature extraction; by extracting the global features of the target and the global features of the background, it is helpful to quickly identify the overall attributes of the target object and the image background, and by extracting the local features of the trunk, it can more carefully reflect the detail differences of the target object, and combine the global and local feature extraction methods to more comprehensively capture the key information in the image; by searching in a pre-built historical database according to the global features of the target, the local features of the trunk and the global features of the background, it can compare the image features of the current image with those in the historical records, and quickly identify potential duplicate claims objects, which greatly shortens the time of fraud detection and improves work efficiency; by determining whether the target object is a duplicate claim object according to the target retrieval results; further determining the fraud risk type according to the background retrieval results, and using a hierarchical decision strategy to make fraud detection more accurate, it can more accurately identify different types of fraud risks. Based on the present application scheme, through refined segmentation and feature extraction, efficient retrieval and hierarchical decision strategies, the claims review process is digitized, and the efficiency and accuracy of fraud risk identification are improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] In order to more clearly illustrate the scheme in the present application, a brief introduction is given below to the drawings required for use in the description of the embodiments of the present application. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0025] Figure 1 is an exemplary system architecture diagram to which the present application may be applied;

[0026] Figure 2 It is a flowchart of an embodiment of the fraud risk identification method of the present application;

[0027] Figure 3 It is a structural schematic diagram of an embodiment of a fraud risk identification device of the present application;

[0028] Figure 4 It is a structural diagram of an embodiment of a computer device of the present application. DETAILED DESCRIPTION

[0029] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by technicians in the technical field of the present application; the terms used in the specification of the application herein are only for the purpose of describing specific embodiments and are not intended to limit the present application; the terms "including" and "having" and any variations thereof in the specification and claims of the present application and the above-mentioned drawings are intended to cover non-exclusive inclusions. The terms "first", "second", etc. in the specification and claims of the present application or the above-mentioned drawings are used to distinguish different objects, not to describe a specific order.

[0030] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0031] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the accompanying drawings.

[0032] like Figure 1 As shown, the system architecture 100 may include a terminal device 101, a network 102 and a server 103. The terminal device 101 may be a laptop 1011, a tablet computer 1012 or a mobile phone 1013. The network 102 is used to provide a medium for a communication link between the terminal device 101 and the server 103. The network 102 may include various connection types, such as wired, wireless communication links or optical fiber cables.

[0033] The user can use the terminal device 101 to interact with the server 103 through the network 102 to receive or send messages, etc. Various communication client applications can be installed on the terminal device 101, such as web browser applications, shopping applications, search applications, instant messaging tools, email clients, social platform software, etc.

[0034] The terminal device 101 can be any electronic device with a display screen and supporting web browsing. In addition to a laptop computer 1011, a tablet computer 1012 or a mobile phone 1013, the terminal device 101 can also be an e-book reader, an MP3 player (Moving Picture Experts Group Audio Layer III), an MP4 (Moving Picture Experts Group Audio Layer IV), a laptop computer, a desktop computer, etc.

[0035] The server 103 may be a server that provides various services, such as a background server that provides support for a web page displayed on the terminal device 101 .

[0036] It should be noted that the fraud risk identification method provided in the embodiment of the present application is generally executed by a server / terminal device, and accordingly, the fraud risk identification device is generally set in the server / terminal device.

[0037] It should be understood that Figure 1 The number of terminal devices, networks and servers in the embodiment is only for illustration. Any number of terminal devices, networks and servers may be provided according to the implementation requirements.

[0038] Continue to refer Figure 2 , showing a flow chart of an embodiment of the question-answering method according to the present application. The fraud risk identification method comprises the following steps:

[0039] Step S201, obtaining an image to be detected, wherein the image to be detected includes at least one target object;

[0040] In this embodiment, the fraud risk identification method is executed on the electronic device (eg Figure 1 The server / terminal device shown in the figure can obtain the image to be detected through a wired connection or a wireless connection. It should be noted that the above wireless connection method may include but is not limited to 2G / 3G / 4G / 5G connection, WiFi connection, Bluetooth connection, WiMAX connection, Zigbee connection, UWB (ultra wideband) connection, and other wireless connection methods currently known or to be developed in the future.

[0041] Specifically, an image to be detected provided by a claim settlement user is obtained, where the image to be detected refers to an image for which fraud risk identification is required, and the image to be detected includes at least one subject matter. The subject matter refers to an object used for claim settlement or risk review, and may include but is not limited to farmed animals (such as cattle, pigs, etc.) and precious plants.

[0042] Step S202, performing target segmentation on the image to be detected to obtain a background area, a target area of ​​the target object, and a body area;

[0043] In this embodiment, image processing technology (such as a deep learning model) is used to perform image detection on the acquired image to be detected, and the background part and the target object part (i.e., the target part) in the image to be detected are identified; image segmentation is performed on the identified background part and target object part to obtain the background area and the target area of ​​the target object; the trunk of the target object in the target area is identified, for example, if the target object is a farmed animal, the trunk part of the target object may include the head, body, forelimbs and hind limbs, etc.; if the target object is a plant, the trunk part of the target object may include roots, stems, leaves, flowers and fruits, etc.; the target area is image segmented according to the identified trunk part to obtain a plurality of trunk areas.

[0044] In some optional implementations, the above-mentioned object segmentation of the image to be detected to obtain the background area, the target area and the torso area of ​​the target object may include the following steps:

[0045] Performing target detection on the image to be detected to obtain a background area, a target area of ​​the target object, and a type of the target object;

[0046] Specifically, the trained target detection model is used to perform target detection on the image to be detected, and the background part and the target part (i.e., the target part) in the image to be detected are identified by the model; the identified background part and target part are segmented to obtain the background area and the target area of ​​the target; the target in the target area is classified and identified to obtain the type of the target. The target type may include pigs, cows, sheep, plants, etc.

[0047] Performing key point detection on the target area according to the type of the target object to obtain key point information of the target object;

[0048] Specifically, the corresponding key point detection rule is obtained according to the type of the target object, and the key point detection rule is used to perform key point detection on the segmented target area to obtain the key point information of the target object. For example, if the target object type is a pig, the key point detection rule for the pig is obtained. The key point detection rule sets 28 key points for the pig and a key point detection algorithm for the pig. The key point detection rule is used to perform key point detection on the target area where the pig is located to obtain the key point information of the target object (i.e., the pig). Among them, the key points reflect the important torso parts in the body structure of the target object, such as limbs, joints, head, tail, etc. The key point information may include information such as the outline of the target object, the outline of the torso part, and the position of the torso part.

[0049] The target area is segmented according to the key point information to obtain a torso area of ​​the target object.

[0050] Specifically, the target area is segmented according to the key point information of the target object. For example, the torso part is located according to the torso part position in the key point information, the torso part range is determined based on the torso part contour in the key point information, and the image is subtracted for the torso part range to obtain the torso area of ​​the target object.

[0051] The embodiments of the present application use a target detection model to perform target detection on the image to be detected, and can accurately identify the background area, target area (i.e., the target object) and target object type in the image; by determining the corresponding key point detection rules according to the target object type, and performing key point detection on the target area based on the key point detection rules, it is possible to quickly identify the key point information on the target object and accurately identify the position and contour of each body part of the target object; using the key point information to segment the target area can obtain a more refined segmentation result, which helps to improve the accuracy of fraud risk identification.

[0052] Step S203, extracting a target global feature based on the target area, extracting a torso local feature based on the torso area, and extracting a background global feature based on the background area;

[0053] In this embodiment, feature extraction is performed based on the segmented target area to obtain target global features, which are used to describe the overall attributes of the target object, such as overall shape, color, texture, etc. Feature extraction is performed based on the segmented torso area to obtain torso local features, which are used to describe the detailed features contained in the torso part of the target object, such as local texture, color distribution, spots, edges, etc. Feature extraction is performed based on the segmented background area to obtain background global features, which are used to describe the overall attributes of the image background, such as environmental information, shooting angle, light brightness, etc.

[0054] Step S204, searching in a pre-built historical database according to the target global feature, the torso local feature and the background global feature, to obtain a target search result and a background search result;

[0055] In this embodiment, a pre-constructed historical database is obtained, which contains a large number of historical claims images and their corresponding feature data. According to the extracted target global features, torso local features and background global features, searches are performed in the pre-constructed historical database respectively. For example, the extracted target global features are compared one by one with the global features of the claims object of each historical claims image in the historical database to obtain several global search results with high similarity; the extracted torso local features are compared one by one with the local features of the claims object of each historical claims image in the historical database to obtain several local search results with high similarity; the global search results and the historical claims images corresponding to the local search results are integrated to obtain the target search results. The background global features of the target search result are obtained from the historical database, and the extracted background global features of the image to be detected are compared one by one with the background global features of the target search result to obtain a similarity list between the background global features of the image to be detected and the background global features of the target search result as the background search result.

[0056] In some optional implementations, the above-mentioned searching in a pre-built historical database according to the target global feature, the torso local feature and the background global feature to obtain the target search result and the background search result may include the following steps:

[0057] Perform a global search in a historical database according to the target global feature to obtain a global search result;

[0058] Specifically, the global features of the claim objects contained in each historical claim image are obtained from the pre-built historical database, and the target global features of the image to be detected are compared with the global features of the claim objects contained in each historical claim image one by one to obtain a number of global retrieval results with high similarity. The global retrieval results are composed of historical claim images with high similarity between the global features of the claim objects and the target global features.

[0059] Acquire the local features of the claim object contained in the global search result from the historical database, search the local features of the claim object contained in the global search result according to the local features of the torso, and obtain the local search result as the target search result;

[0060] Specifically, the local features of the claim object contained in each global search result are obtained from a pre-constructed historical database, that is, the historical claim image corresponding to the global search result is determined, and the local features of the claim object contained in the historical claim image are obtained from the pre-constructed historical database; the local features of the torso of the image to be detected are compared with the local features of the claim object contained in the historical claim image, one by one, to obtain several local search results with high similarity. The local search result consists of historical claim images with high similarity between the local features of the claim object and the local features of the torso. The local search result is determined as the target search result.

[0061] The background global features corresponding to the target search result are obtained from the historical database, and the background global features corresponding to the target search result are searched according to the background global features to obtain the background search result.

[0062] Specifically, the background global features corresponding to each target retrieval result are obtained from a pre-constructed historical database, that is, the historical claims image corresponding to the target retrieval result (i.e., the local retrieval result) is determined, and the background global features contained in the historical claims image are obtained from the pre-constructed historical database; the background global features of the image to be detected are compared with the background global features contained in the historical claims image one by one to obtain the similarity value between the background global features of the image to be detected and the background global features of the target retrieval result, and each target retrieval result is sorted in ascending order according to the similarity value to obtain a similarity list as the background retrieval result.

[0063] The embodiment of the present application can quickly narrow the search scope by first using the global features of the target for global retrieval. The global features are used as the basis for preliminary screening, and can quickly exclude claims that are significantly different from the global features of the target. On the basis of global retrieval, the local features of the torso are further used to perform local retrieval in the global retrieval results, which can not only reduce unnecessary calculations and improve retrieval efficiency, but also significantly improve retrieval accuracy. Since the local features of the torso are more discriminative than the global features of the target, they can more accurately match the target in the image to be detected, thereby obtaining more accurate target retrieval results. Through the combination of global features and local features, the complementarity of coarse matching and fine matching is achieved, and the overall information of the image content is compatible with the matching of the local block details, avoiding the problem of insufficient discriminability caused by matching only the global features of similar targets. Furthermore, on the basis of local retrieval, the background global features of the image to be detected are used to perform background retrieval in the target retrieval results, providing a basis for subsequent judgment of what type of fraud risk the target belongs to.

[0064] In some optional implementations, the step of obtaining the local features of the claim object included in the global search result from the historical database may include the following steps:

[0065] Calculating the similarity value between the target global feature and the global search result;

[0066] Sort the global search results from high to low according to the similarity values, and select a preset number of target global search results that are ranked top;

[0067] The local features of the claim object included in the target global search result are obtained from the historical database.

[0068] Specifically, for the retrieved global search results, the global features of the claim objects of the historical claim images corresponding to the global search results are obtained, and the similarity value between the target global features and the global features of the claim objects is calculated. The global search results are sorted from high to low according to the similarity data, and a preset number (N) of global search results with the highest ranking are selected as the target global search results. The historical claim images corresponding to the target global search results are determined, and the local features of the claim objects contained in the historical claim images are obtained from the pre-built historical database.

[0069] Step S205, determining whether the subject matter is a duplicate claim object according to the target search result;

[0070] In this embodiment, based on the retrieved target search results, it is determined whether the subject in the object to be detected is a duplicate claim object. For example, the similarity of the target search results and the subject in the image to be detected is calculated in turn. If the subject is highly similar to the claim object in one or some target search results (that is, the target global features and the local features of the torso are matched), the subject is determined to be a duplicate claim object. If the similarity between the subject and the claim objects in all target search results is lower than a preset threshold, it is determined that the subject is not a duplicate claim object.

[0071] In some optional implementations, the above-mentioned determining whether the subject matter is a duplicate claim object according to the target search result may include the following steps:

[0072] Calculating a first matching value between the target search result and each torso region of the target object;

[0073] Comparing the first matching value with a preset first matching threshold;

[0074] If the first matching value exceeds the first matching threshold, the subject matter is determined as a duplicate claim object.

[0075] Specifically, determine the historical claim image corresponding to the target retrieval result, obtain the local features of the claim object contained in the historical claim image from a pre-constructed historical database; obtain the local features of the torso corresponding to each torso area of ​​the subject matter; and calculate the first matching value between the local features of the claim object and the local features of the torso. Obtain a pre-set first matching threshold, which is used to measure whether the torso area of ​​the subject matter is repeated with the torso area of ​​the claim object in the historical claim image. Compare the calculated first matching value with the pre-set first matching threshold to obtain a comparison result. When the comparison result is that the first matching value exceeds the first matching threshold, it indicates that there is a high degree of repeatability between the torso area of ​​the subject matter and the torso area of ​​the claim object in the historical claim image. At this time, the subject matter is determined to be a duplicate claim object.

[0076] Exemplarily, assuming that the trunk regions of the subject matter are the head, body, forelimbs and hind limbs, etc., respectively, the first matching values ​​between the head, body, forelimbs and hind limbs, etc., of the subject matter and the head, body, forelimbs and hind limbs, etc., of each claim object in the target search results are calculated, and the first matching threshold set for each trunk region is obtained. The first matching value of each trunk region is compared with the corresponding first matching threshold. When the comparison result shows that the first matching value exceeds the first matching threshold, one point is recorded. The total scores of each claim object after comparison are summarized. If there is a claim object with a total score greater than a preset threshold (for example, greater than 3), it means that the subject matter has a high degree of duplication with the claim object. At this time, the subject matter is determined to be a duplicate claim object.

[0077] Optionally, if the subject matter in the image to be detected is not a duplicate claim object, the subject matter in the image to be detected is determined to have no fraud risk.

[0078] Step S206: If the subject matter is a repeated claim object, determine the fraud risk type based on the background search result.

[0079] In this embodiment, if the subject matter in the image to be detected is a repeated claim object, the fraud risk type of the subject matter in the image to be detected is further determined based on the background retrieval results (such as the matching degree of the background global features, the shooting time, the location and other information). For example, if the background global features of one or some historical claims images exist in the background retrieval results, which are exactly the same or very similar to the background global features of the image to be detected, and the shooting time is similar, then the detection image is paired with the historical claims image and marked as similar background fraud of the same subject matter. If the background global features of one or some historical claims images exist in the background retrieval results, which are less similar or completely dissimilar to the background global features of the image to be detected, then the detection image is paired with the historical claims image and marked as background replacement fraud of the same subject matter.

[0080] In some optional implementations, the above-mentioned determining the fraud risk type according to the background search result may include the following steps:

[0081] If the background search result is not a null value, calculating a second matching value between the background search result and the background area of ​​the image to be detected;

[0082] Specifically, if the background retrieval result is not a null value, that is, there are several historical claims images in the historical database whose background areas are similar to the background area of ​​the image to be detected, then determine the historical claims image corresponding to the background retrieval result, and obtain the background global features contained in the historical claims image; obtain the background global features corresponding to the background area of ​​the image to be detected; calculate the second matching value between the background global features contained in the historical claims image and the background global features corresponding to the image to be detected.

[0083] Comparing the second matching value with a preset second matching threshold;

[0084] Specifically, a preset second matching threshold is obtained, and the second matching threshold is used to measure the similarity between the background area of ​​the image to be detected and the background area of ​​the historical claims image. The calculated second matching value is compared with the preset second matching threshold to obtain a comparison result.

[0085] If the second matching value exceeds the second matching threshold, the fraud risk type of the repeated claim object is determined as similar background fraud of the same subject matter;

[0086] Specifically, when the comparison result is that the second matching value exceeds the second matching threshold, it indicates that there is a high similarity between the background area of ​​the image to be detected and the background area of ​​the historical claim image. At this time, the fraud risk type of the repeated claim object (ie, the subject matter) is determined as similar background fraud of the same subject matter.

[0087] If the second matching value does not exceed the second matching threshold, the fraud risk type of the duplicate claim object is determined as background replacement fraud of the same subject matter.

[0088] Specifically, when the comparison result is that the second matching value does not exceed the second matching threshold, it indicates that there is a large difference between the background area of ​​the image to be detected and the background area of ​​the historical claim image. At this time, the fraud risk type of the repeated claim object (ie, the subject matter) is determined as background replacement fraud of the same subject matter.

[0089] In some optional implementations, the above-mentioned determining the fraud risk type according to the background search result may include the following steps:

[0090] If the background search result is a null value, the fraud risk type of the duplicate claim object is determined to be background replacement fraud of the same subject matter.

[0091] Specifically, if the background search result is a null value, that is, there is no historical claim image in the historical database whose background area is similar to the background area of ​​the image to be detected, then the fraud risk type of the duplicate claim object (i.e., the subject matter) is determined to be background replacement fraud of the same subject matter.

[0092] The embodiment of the present application achieves intelligent risk judgment by calculating the second matching value between the background search result and the background area of ​​the image to be detected, and compares it with the preset second matching threshold, thereby reducing manual intervention and improving detection efficiency and accuracy. The embodiment of the present application covers a variety of fraud risk types, such as background replacement fraud of the same subject matter and similar background fraud of the same subject matter. This subdivision makes fraud detection more comprehensive and can identify more types of fraud risks.

[0093] The embodiment of the present application, through the above scheme, specifically performs target segmentation on the image to be detected, accurately distinguishes the background area, the target area of ​​the target object and the trunk area, and ensures the accuracy and pertinence of subsequent feature extraction; by extracting the global features of the target and the global features of the background, it is helpful to quickly identify the overall attributes of the target object and the image background, and by extracting the local features of the trunk, it can more carefully reflect the detail differences of the target object, and combine the global and local feature extraction methods to more comprehensively capture the key information in the image; by searching in a pre-built historical database according to the global features of the target, the local features of the trunk and the global features of the background, it can compare the image features of the current image with those in the historical records, and quickly identify potential duplicate claims objects, greatly shortening the time of fraud detection and improving work efficiency; by determining whether the target object is a duplicate claim object according to the target retrieval results; further determining the fraud risk type according to the background retrieval results, and using the hierarchical decision strategy to make fraud detection more accurate, it can more accurately identify different types of fraud risks. Based on the scheme of the present application, through refined segmentation and feature extraction, efficient retrieval and hierarchical decision strategies, the claims review process is digitized, and the efficiency and accuracy of fraud risk identification are improved.

[0094] It should be emphasized that in order to further ensure the privacy and security of the above-mentioned images to be detected, background areas, target areas and torso areas of the subject matter and other relevant data, the above-mentioned images to be detected, background areas, target areas and torso areas of the subject matter and other relevant data can also be stored in a blockchain node.

[0095] The blockchain referred to in this application is a new application model of computer technologies such as distributed data storage, peer-to-peer transmission, consensus mechanism, encryption algorithm, etc. Blockchain is essentially a decentralized database, a string of data blocks generated by cryptographic methods. Each data block contains a batch of network transaction information, which is used to verify the validity of its information (anti-counterfeiting) and generate the next block. Blockchain can include the underlying blockchain platform, platform product service layer, and application service layer.

[0096] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through computer-readable instructions, and the computer-readable instructions can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, the aforementioned storage medium can be a non-volatile storage medium such as a disk, an optical disk, a read-only memory (ROM), or a random access memory (RAM).

[0097] It should be understood that, although the steps in the flowchart of the accompanying drawings are displayed in sequence as indicated by the arrows, these steps are not necessarily executed in sequence in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least a part of the steps in the flowchart of the accompanying drawings may include multiple sub-steps or multiple stages, and these sub-steps or stages are not necessarily executed at the same time, but can be executed at different times, and their execution order is not necessarily sequential, but can be executed in turn or alternately with other steps or at least a part of the sub-steps or stages of other steps.

[0098] Further references Figure 3 , as a response to the above Figure 2 In order to realize the method shown in the figure, the present application provides an embodiment of a fraud risk identification device. Figure 2 Corresponding to the method embodiment shown, the device can be specifically applied to various electronic devices.

[0099] like Figure 3 As shown, the fraud risk identification device 400 described in this embodiment includes: an image acquisition module 401, a region segmentation module 402, a feature extraction module 403, a feature retrieval module 404, an object determination module 405 and a type determination module 406. Among them:

[0100] An image acquisition module 401 is used to acquire an image to be detected, wherein the image to be detected includes at least one target object;

[0101] A region segmentation module 402 is used to perform target segmentation on the image to be detected to obtain a background region, a target region of the target object, and a torso region;

[0102] A feature extraction module 403 is used to extract a target global feature according to the target area, extract a torso local feature according to the torso area, and extract a background global feature according to the background area;

[0103] A feature retrieval module 404 is used to search in a pre-built historical database according to the target global feature, the torso local feature and the background global feature to obtain a target retrieval result and a background retrieval result;

[0104] The object determination module 405 is used to determine whether the subject matter is a duplicate claim object according to the target search result;

[0105] The type determination module 406 is used to determine the fraud risk type according to the background search result if the subject matter is a repeated claim object.

[0106] In this embodiment, the image acquisition module 401 is used to acquire the image to be detected provided by the claim user, the image to be detected refers to the image that needs to be identified for fraud risk, and the image to be detected includes at least one subject. The subject refers to the object used to apply for claims or conduct risk review, which may include but is not limited to farmed animals (such as cattle, pigs, etc.) and precious plants.

[0107] Specifically, the region segmentation module 402 is used to perform image detection on the acquired image to be detected by using image processing technology (such as a deep learning model), and identify the background part and the target object part (i.e., the target part) in the image to be detected; perform image segmentation on the identified background part and target object part to obtain the background area and the target area of ​​the target object; perform trunk recognition on the target object in the target area, for example, if the target object is a farmed animal, the trunk part of the target object may include the head, body, forelimbs and hind limbs, etc.; if the target object is a plant, the trunk part of the target object may include roots, stems, leaves, flowers and fruits, etc.; perform image segmentation on the target area according to the identified trunk part to obtain a plurality of trunk areas.

[0108] Specifically, the feature extraction module 403 is used to extract features from the segmented target area to obtain target global features, which are used to describe the overall attributes of the target, such as overall shape, color, texture, etc. Feature extraction is performed from the segmented torso area to obtain torso local features, which are used to describe the detailed features contained in the torso part of the target, such as local texture, color distribution, spots, edges, etc. Feature extraction is performed from the segmented background area to obtain background global features, which are used to describe the overall attributes of the image background, such as environmental information, shooting angle, light brightness, etc.

[0109] Specifically, the feature retrieval module 404 is used to obtain a pre-constructed historical database, which contains a large number of historical claims images and their corresponding feature data. According to the extracted target global features, torso local features and background global features, retrieval is performed in the pre-constructed historical database respectively. For example, the extracted target global features are compared one by one with the global features of the claims object of each historical claims image in the historical database to obtain several global retrieval results with high similarity; the extracted torso local features are compared one by one with the local features of the claims object of each historical claims image in the historical database to obtain several local retrieval results with high similarity; the global retrieval results and the historical claims images corresponding to the local retrieval results are integrated to obtain the target retrieval results. The background global features of the target retrieval result are obtained from the historical database, and the background global features of the extracted image to be detected are compared one by one with the background global features of the target retrieval result to obtain a similarity list between the background global features of the image to be detected and the background global features of the target retrieval result as the background retrieval result.

[0110] Specifically, the object determination module 405 is used to determine whether the subject matter in the object to be detected is a duplicate claim object based on the retrieved target retrieval results. For example, the similarity between the target retrieval results and the subject matter of the image to be detected is calculated in turn. If the subject matter is highly similar to the claim object in one or some target retrieval results (that is, the target global features and the local features of the torso are matched), the subject matter is determined to be a duplicate claim object. If the similarity between the subject matter and the claim objects in all target retrieval results is lower than a preset threshold, the subject matter is determined not to be a duplicate claim object.

[0111] Optionally, if the subject matter in the image to be detected is not a duplicate claim object, the subject matter in the image to be detected is determined to have no fraud risk.

[0112] Specifically, the type determination module 406 is used to further determine the fraud risk type of the subject matter in the image to be detected based on the background retrieval results (such as the matching degree of the background global features, the shooting time, the location and other information) if the subject matter in the image to be detected is a repeated claim object. For example, if the background global features of one or some historical claims images exist in the background retrieval results, which are exactly the same or very similar to the background global features of the image to be detected, and the shooting time is similar, then the detection image is paired with the historical claims image and marked as similar background fraud of the same subject matter. If the background global features of one or some historical claims images exist in the background retrieval results, which are less similar or completely dissimilar to the background global features of the image to be detected, then the detection image is paired with the historical claims image and marked as background replacement fraud of the same subject matter.

[0113] The fraud risk identification device 400 of the embodiment of the present application, by performing target segmentation on the image to be detected, accurately distinguishes the background area, the target area and the trunk area of ​​the target object, and ensures the accuracy and pertinence of subsequent feature extraction; by extracting the global features of the target and the global features of the background, it is helpful to quickly identify the overall attributes of the target object and the image background, and by extracting the local features of the trunk, it can more carefully reflect the detail differences of the target object, and combine the global and local feature extraction methods to more comprehensively capture the key information in the image; by searching in a pre-built historical database according to the global features of the target, the local features of the trunk and the global features of the background, it can compare the image features of the current image with those in the historical records, and quickly identify potential duplicate claim objects, which greatly shortens the time of fraud detection and improves work efficiency; by determining whether the target object is a duplicate claim object according to the target retrieval results; further determining the fraud risk type according to the background retrieval results, using the hierarchical decision strategy to make fraud detection more accurate, and can more accurately identify different types of fraud risks. Based on the present application scheme, through refined segmentation and feature extraction, efficient retrieval and hierarchical decision strategies, the claims review process is digitized, and the efficiency and accuracy of fraud risk identification are improved.

[0114] In some optional implementations of this embodiment, the region segmentation module 402 may include a target detection submodule, a key detection submodule and a target segmentation submodule.

[0115] The target detection submodule is used to perform target detection on the image to be detected to obtain the background area, the target area of ​​the target object and the type of the target object;

[0116] A key detection submodule, used to perform key point detection on the target area according to the target object type to obtain key point information of the target object;

[0117] The target segmentation submodule is used to perform target segmentation on the target area according to the key point information to obtain the torso area of ​​the target object.

[0118] Specifically, the target detection submodule is used to use the trained target detection model to perform target detection on the image to be detected, identify the background part and the target part (i.e., the target part) in the image to be detected through the model; perform image segmentation on the identified background part and target part to obtain the background area and the target area of ​​the target; classify and identify the target in the target area to obtain the type of the target. The target type may include pigs, cows, sheep, plants, etc.

[0119] Specifically, the key detection submodule is used to obtain the corresponding key point detection rules according to the target type, and use the key point detection rules to perform key point detection on the segmented target area to obtain the key point information of the target. For example, if the target type is a pig, the key point detection rules for pigs are obtained. The key point detection rules set 28 key points for pigs and a key point detection algorithm for pigs. The key point detection rules are used to perform key point detection on the target area where the pig is located to obtain the key point information of the target (i.e., the pig). Among them, the key points reflect the important torso parts in the body structure of the target, such as limbs, joints, head, tail, etc. The key point information may include information such as the outline of the target, the outline of the torso parts, and the position of the torso parts.

[0120] Specifically, the target segmentation submodule is used to perform target segmentation on the target area according to the key point information of the target object. For example, the torso part is located according to the torso part position in the key point information, the torso part range is determined based on the torso part contour in the key point information, and the image is subtracted for the torso part range to obtain the torso area of ​​the target object.

[0121] The region segmentation module 402 of the embodiment of the present application can accurately identify the background area, target area (i.e., target object) and target object type in the image by adopting a target detection model to perform target detection on the image to be detected; by determining the corresponding key point detection rule according to the target object type and performing key point detection on the target area based on the key point detection rule, it can quickly identify the key point information on the target object and accurately identify the position and contour of each body part of the target object; using the key point information to perform target segmentation on the target area can obtain a more refined segmentation result, which helps to improve the accuracy of fraud risk identification.

[0122] In some optional implementations of this embodiment, the feature retrieval module 404 may include a global retrieval submodule, a local retrieval submodule, and a background retrieval submodule.

[0123] A global search submodule, used to perform a global search in a historical database according to the target global feature to obtain a global search result;

[0124] A local search submodule, for obtaining the local features of the claim object contained in the global search result from the historical database, searching the local features of the claim object contained in the global search result according to the local features of the torso, and obtaining the local search result as the target search result;

[0125] The background retrieval submodule is used to obtain the background global features corresponding to the target retrieval results from the historical database, and to search the background global features corresponding to the target retrieval results according to the background global features to obtain background retrieval results.

[0126] Specifically, the global retrieval submodule is used to obtain the global features of the claim objects contained in each historical claim image from the pre-built historical database, and compare the target global features of the image to be detected with the global features of the claim objects contained in each historical claim image one by one to obtain a number of global retrieval results with high similarity. The global retrieval results are composed of historical claim images with high similarity between the global features of the claim objects and the target global features.

[0127] Specifically, the local retrieval submodule is used to obtain the local features of the claim object contained in each global retrieval result from a pre-constructed historical database, that is, to determine the historical claim image corresponding to the global retrieval result, and obtain the local features of the claim object contained in the historical claim image from the pre-constructed historical database; compare the local features of the torso of the image to be detected with the local features of the claim object contained in the historical claim image, one by one, to obtain several local retrieval results with high similarity. The local retrieval result consists of historical claim images with high similarity between the local features of the claim object and the local features of the torso. The local retrieval result is determined as the target retrieval result.

[0128] Specifically, the background retrieval submodule is used to obtain the background global features corresponding to each target retrieval result from a pre-constructed historical database, that is, to determine the historical claims image corresponding to the target retrieval result (i.e., the local retrieval result), and obtain the background global features contained in the historical claims image from the pre-constructed historical database; the background global features of the image to be detected are compared with the background global features contained in the historical claims image one by one to obtain the similarity value between the background global features of the image to be detected and the background global features of the target retrieval result, and each target retrieval result is sorted in ascending order according to the similarity value to obtain a similarity list as the background retrieval result.

[0129] The feature retrieval module 404 of the embodiment of the present application can quickly narrow the search scope by first performing a global search using the target global features. The global features are used as the basis for preliminary screening, and can quickly exclude claims that are significantly different from the target global features. On the basis of global search, the local features of the torso are further used to perform local search in the global search results, which can not only reduce unnecessary calculations and improve search efficiency, but also significantly improve the accuracy of the search. Since the local features of the torso are more discriminative than the global features of the target, they can more accurately match the target in the image to be detected, thereby obtaining more accurate target search results. Through the combination of global features and local features, the complementarity of coarse matching and fine matching is achieved, and the overall information of the image content is compatible with the matching of the local block details, avoiding the problem of insufficient discriminability caused by matching only the global features of similar targets. Furthermore, on the basis of local search, the background global features of the image to be detected are used to perform background search in the target search results, providing a basis for subsequent judgment of what type of fraud risk the target belongs to.

[0130] In some optional implementations of this embodiment, the local search submodule may include a similarity calculation unit, a result ranking unit, and a feature acquisition unit.

[0131] A similarity calculation unit, used to calculate the similarity value between the target global feature and the global search result;

[0132] A result sorting unit, used to sort the global search results from high to low according to the similarity values, and select a preset number of target global search results with top rankings;

[0133] The feature acquisition unit is used to acquire the local features of the claim object contained in the target global search result from the historical database.

[0134] Specifically, the similarity calculation unit is used to obtain the global features of the claim object of the historical claim image corresponding to each global search result for the retrieved global search results, and calculate the similarity value between the target global feature and the global feature of the claim object. The result sorting unit is used to sort the global search results in descending order according to the similarity data, and select a preset number (N) of global search results with the highest ranking as the target global search results. The feature acquisition unit is used to determine the historical claim image corresponding to the target global search result, and obtain the local features of the claim object contained in the historical claim image from a pre-constructed historical database.

[0135] In some optional implementations of this embodiment, the object determination module 405 may include a first calculation submodule, a first comparison submodule and a first determination submodule.

[0136] A first calculation submodule, used for calculating a first matching value between the target search result and each torso region of the target object;

[0137] A first comparison submodule, used for comparing the first matching value with a preset first matching threshold;

[0138] The first determination submodule is used to determine the subject matter as a duplicate claim object if the first matching value exceeds the first matching threshold.

[0139] Specifically, the first calculation submodule is used to determine the historical claim image corresponding to the target retrieval result, obtain the local features of the claim object contained in the historical claim image from a pre-constructed historical database; obtain the local features of the torso corresponding to each torso area of ​​the subject matter; and calculate the first matching value between the local features of the claim object and the local features of the torso. The first comparison submodule is used to obtain a pre-set first matching threshold, which is used to measure whether the torso area of ​​the subject matter is repeated with the torso area of ​​the claim object in the historical claim image; compare the calculated first matching value with the pre-set first matching threshold to obtain a comparison result. The first determination submodule is used to determine the subject matter as a repeated claim object when the comparison result is that the first matching value exceeds the first matching threshold, indicating that there is a high degree of repeatability between the torso area of ​​the subject matter and the torso area of ​​the claim object in the historical claim image.

[0140] Exemplarily, assuming that the trunk regions of the subject matter are the head, body, forelimbs and hind limbs, etc., respectively, the first matching values ​​between the head, body, forelimbs and hind limbs, etc., of the subject matter and the head, body, forelimbs and hind limbs, etc., of each claim object in the target search results are calculated, and the first matching threshold set for each trunk region is obtained. The first matching value of each trunk region is compared with the corresponding first matching threshold. When the comparison result shows that the first matching value exceeds the first matching threshold, one point is recorded. The total scores of each claim object after comparison are summarized. If there is a claim object with a total score greater than a preset threshold (for example, greater than 3), it means that the subject matter has a high degree of duplication with the claim object. At this time, the subject matter is determined to be a duplicate claim object.

[0141] In some optional implementations of this embodiment, the type determination module 406 may include a second calculation submodule, a second comparison submodule, a second determination submodule, and a third determination submodule.

[0142] A second calculation submodule, configured to calculate a second matching value between the background search result and the background area of ​​the image to be detected if the background search result is not a null value;

[0143] A second comparison submodule, used for comparing the second matching value with a preset second matching threshold;

[0144] a second determination submodule, configured to determine the fraud risk type of the repeated claim object as similar background fraud of the same subject matter if the second matching value exceeds the second matching threshold;

[0145] The third determination submodule is used to determine the fraud risk type of the duplicate claim object as background replacement fraud of the same subject matter if the second matching value does not exceed the second matching threshold.

[0146] Specifically, the second calculation submodule is used to determine the historical claims image corresponding to the background retrieval result if the background retrieval result is not a null value, that is, there are several historical claims images in the historical database whose background areas are similar to the background area of ​​the image to be detected, and obtain the background global features contained in the historical claims image; obtain the background global features corresponding to the background area of ​​the image to be detected; and calculate the second matching value between the background global features contained in the historical claims image and the background global features corresponding to the image to be detected.

[0147] Specifically, the second comparison submodule is used to obtain a preset second matching threshold, which is used to measure the similarity between the background area of ​​the image to be detected and the background area of ​​the historical claims image. The calculated second matching value is compared with the preset second matching threshold to obtain a comparison result.

[0148] Specifically, the second determination submodule is used to determine the fraud risk type of the repeated claim object (i.e., the subject matter) as similar background fraud of the same subject matter when the comparison result is that the second matching value exceeds the second matching threshold, indicating that there is a high similarity between the background area of ​​the image to be detected and the background area of ​​the historical claim image.

[0149] Specifically, the third determination submodule is used to determine the fraud risk type of the repeated claim object (i.e., the subject matter) as background replacement fraud of the same subject matter when the comparison result is that the second matching value does not exceed the second matching threshold, indicating that there is a large difference between the background area of ​​the image to be detected and the background area of ​​the historical claim image.

[0150] In some optional implementations of this embodiment, the type determination module 406 may further include a fourth determination submodule.

[0151] The fourth determination submodule is used to determine the fraud risk type of the repeated claim object as background replacement fraud of the same subject matter if the background search result is a null value.

[0152] Specifically, the fourth determination submodule is used to determine the fraud risk type of the duplicate claim object (i.e., the subject matter) as background replacement fraud of the same subject matter if the background retrieval result is a null value, that is, there is no historical claim image in the historical database whose background area is similar to the background area of ​​the image to be detected.

[0153] The type determination module 406 of the embodiment of the present application realizes intelligent risk judgment by calculating the second matching value between the background search result and the background area of ​​the image to be detected, and compares it with the preset second matching threshold, thereby reducing manual intervention and improving detection efficiency and accuracy. The embodiment of the present application covers a variety of fraud risk types, such as background replacement fraud of the same subject matter and similar background fraud of the same subject matter. This subdivision makes fraud detection more comprehensive and can identify more types of fraud risks.

[0154] To solve the above technical problems, the present application also provides a computer device. Figure 4 , Figure 4 This is a basic structural block diagram of the computer device in this embodiment.

[0155] The computer device 6 includes a memory 61, a processor 62, and a network interface 63 that are interconnected and communicated through a system bus. It should be noted that the figure only shows a computer device 6 with a memory 61, a processor 62, and a network interface 63, but it should be understood that it is not required to implement all the components shown, and more or fewer components can be implemented instead. Among them, those skilled in the art can understand that the computer device here is a device that can automatically perform numerical calculations and / or information processing according to pre-set or stored instructions, and its hardware includes but is not limited to microprocessors, application specific integrated circuits (Application Specific Integrated Circuit, ASIC), programmable gate arrays (Field-Programmable Gate Array, FPGA), digital processors (Digital Signal Processor, DSP), embedded devices, etc.

[0156] The computer device may be a computing device such as a desktop computer, a notebook, a PDA, a cloud server, etc. The computer device may interact with a user through a keyboard, a mouse, a remote controller, a touch pad, or a voice control device.

[0157] The memory 61 includes at least one type of readable storage medium, and the readable storage medium includes flash memory, hard disk, multimedia card, card-type memory (for example, SD or DX memory, etc.), random access memory (RAM), static random access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic memory, disk, optical disk, etc. In some embodiments, the memory 61 can be an internal storage unit of the computer device 6, such as a hard disk or memory of the computer device 6. In other embodiments, the memory 61 can also be an external storage device of the computer device 6, such as a plug-in hard disk equipped on the computer device 6, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (FlashCard), etc. Of course, the memory 61 can also include both the internal storage unit of the computer device 6 and its external storage device. In this embodiment, the memory 61 is generally used to store the operating system and various application software installed on the computer device 6, such as computer-readable instructions of the fraud risk identification method, etc. In addition, the memory 61 can also be used to temporarily store various types of data that have been output or are to be output.

[0158] The processor 62 may be a central processing unit (CPU), a controller, a microcontroller, a microprocessor or other data processing chip in some embodiments. The processor 62 is generally used to control the overall operation of the computer device 6. In this embodiment, the processor 62 is used to run the computer-readable instructions stored in the memory 61 or process data, such as running the computer-readable instructions of the fraud risk identification method.

[0159] The network interface 63 may include a wireless network interface or a wired network interface. The network interface 63 is generally used to establish a communication connection between the computer device 6 and other electronic devices.

[0160] The present application also provides another embodiment, namely, providing a computer-readable storage medium, wherein the computer-readable storage medium stores computer-readable instructions, and the computer-readable instructions can be executed by at least one processor to enable the at least one processor to perform the steps of the fraud risk identification method as described above.

[0161] The computer device, computer-readable storage medium and computer-readable instructions thereof provided in the embodiments of the present application perform target segmentation on the image to be detected through a processor, accurately distinguish the background area, the target area and the trunk area of ​​the target object, and ensure the accuracy and pertinence of subsequent feature extraction; by extracting the target global features and the background global features, it is helpful to quickly identify the overall properties of the target object and the image background, and by extracting the trunk local features, it is possible to more finely reflect the detail differences of the target object, and by combining the global and local feature extraction methods, it is possible to more comprehensively capture the key information in the image; by searching in a pre-constructed historical database according to the target global features, the trunk local features and the background global features, it is possible to compare the image features in the current image with those in the historical records, and quickly identify potential duplicate claims objects, thereby greatly shortening the time for fraud detection and improving work efficiency; by determining whether the target object is a duplicate claims object based on the target retrieval results; and further determining the fraud risk type based on the background retrieval results, a hierarchical decision-making strategy is used to make fraud detection more accurate, and different types of fraud risks can be more accurately identified. Based on this application plan, through refined segmentation and feature extraction, efficient retrieval and hierarchical decision-making strategies, the claims review process is digitized, which improves the efficiency and accuracy of fraud risk identification.

[0162] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present application, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes a number of instructions for a terminal device (which can be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in each embodiment of the present application.

[0163] Obviously, the embodiments described above are only some embodiments of the present application, rather than all embodiments. The preferred embodiments of the present application are given in the accompanying drawings, but they do not limit the patent scope of the present application. The present application can be implemented in many different forms. On the contrary, the purpose of providing these embodiments is to make the understanding of the disclosure of the present application more thorough and comprehensive. Although the present application is described in detail with reference to the aforementioned embodiments, for those skilled in the art, it is still possible to modify the technical solutions recorded in the aforementioned specific implementation methods, or to replace some of the technical features therein with equivalents. Any equivalent structure made using the contents of the specification and drawings of this application, directly or indirectly used in other related technical fields, is similarly within the scope of patent protection of this application.

[0164] The non-Company software tools or components appearing in the embodiments of this application are merely examples and do not represent actual use.

Claims

1. A fraud risk identification method, characterized in that: The steps include: Acquire an image to be detected, wherein the image to be detected includes at least one target object; Performing target segmentation on the image to be detected to obtain a background area, a target area of ​​the target object, and a torso area; Extracting the global features of the target from the target area, extracting the local features of the trunk from the trunk area, and extracting the global features of the background from the background area; According to the target global feature, the torso local feature and the background global feature, a search is performed in a pre-constructed historical database to obtain a target search result and a background search result; Determining whether the subject matter is a duplicate claim object according to the target search result; If the subject matter is a duplicate claim object, the fraud risk type is determined based on the background search results.

2. The fraud risk identification method according to claim 1, characterized in that: The step of performing target segmentation on the image to be detected to obtain a background area, a target area of ​​the target object, and a trunk area includes: Performing target detection on the image to be detected to obtain a background area, a target area of ​​the target object, and a type of the target object; Performing key point detection on the target area according to the type of the target object to obtain key point information of the target object; The target area is segmented according to the key point information to obtain a torso area of ​​the target object.

3. The fraud risk identification method according to claim 1, characterized in that: The step of searching in a pre-built historical database according to the target global feature, the torso local feature and the background global feature to obtain the target search result and the background search result comprises: Perform a global search in a historical database according to the target global feature to obtain a global search result; Acquire the local features of the claim object contained in the global search result from the historical database, search the local features of the claim object contained in the global search result according to the local features of the torso, and obtain the local search result as the target search result; The background global features corresponding to the target search result are obtained from the historical database, and the background global features corresponding to the target search result are searched according to the background global features to obtain the background search result.

4. The fraud risk identification method according to claim 3, characterized in that: The step of acquiring the local features of the claim object contained in the global search result from the historical database includes: Calculating the similarity value between the target global feature and the global search result; Sort the global search results from high to low according to the similarity values, and select a preset number of target global search results that are ranked top; The local features of the claim object included in the target global search result are obtained from the historical database.

5. The fraud risk identification method according to claim 1, characterized in that: The step of determining whether the subject matter is a duplicate claim object according to the target search result includes: Calculating a first matching value between the target search result and each torso region of the target object; Comparing the first matching value with a preset first matching threshold; If the first matching value exceeds the first matching threshold, the subject matter is determined as a duplicate claim object.

6. The fraud risk identification method according to claim 1, characterized in that: The step of determining the fraud risk type according to the background search result comprises: If the background search result is not a null value, calculating a second matching value between the background search result and the background area of ​​the image to be detected; Comparing the second matching value with a preset second matching threshold; If the second matching value exceeds the second matching threshold, the fraud risk type of the repeated claim object is determined as similar background fraud of the same subject matter; If the second matching value does not exceed the second matching threshold, the fraud risk type of the duplicate claim object is determined as background replacement fraud of the same subject matter.

7. The fraud risk identification method according to claim 1, characterized in that: The step of determining the fraud risk type according to the background search result comprises: If the background search result is a null value, the fraud risk type of the duplicate claim object is determined to be background replacement fraud of the same subject matter.

8. A fraud risk identification device, characterized in that: The device comprises: An image acquisition module, used to acquire an image to be detected, wherein the image to be detected includes at least one target object; A region segmentation module is used to perform target segmentation on the image to be detected to obtain a background region, a target region of the target object, and a trunk region; A feature extraction module, configured to extract a target global feature based on the target area, extract a torso local feature based on the torso area, and extract a background global feature based on the background area; A feature retrieval module, used to search in a pre-built historical database according to the target global feature, the torso local feature and the background global feature, to obtain a target retrieval result and a background retrieval result; An object determination module, used to determine whether the subject matter is a duplicate claim object according to the target search result; The type determination module is used to determine the fraud risk type based on the background search result if the subject matter is a repeated claim object.

9. A computer device, characterized in that: The method comprises a memory and a processor, wherein the memory stores computer-readable instructions, and the processor implements the steps of the fraud risk identification method according to any one of claims 1 to 7 when executing the computer-readable instructions.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-readable instructions, and when the computer-readable instructions are executed by a processor, the steps of the fraud risk identification method according to any one of claims 1 to 7 are implemented.

Images