Risk identification method and device based on image characters, equipment and storage medium

Abstract

The invention belongs to the technical field of artificial intelligence, and discloses a risk identification method and device based on image characters, equipment and a storage medium, and the method comprises the steps: receiving a vehicle leasing request initiated by a user, and generating leasing information according to the vehicle leasing request; acquiring an image set shot when a third-party company installs GPS equipment on the vehicle, and establishing an association relationship between the image set and the renting information; identifying image characters of each image in the image set by adopting a preset image identification technology, and matching the image characters with a preset risk lexicon and a preset sensitive lexicon to obtain a matching result; when the matching result is that the text segmented words hit a preset sensitive word bank, code printing processing is conducted on the text segmented words, and when the matching result is that the text segmented words hit a preset risk word bank, an early warning label is given to the leasing information according to the incidence relation, and the leasing information with the early warning label is uploaded to a preset risk control management platform; the accuracy and efficiency of risk identification can be improved.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence, and in particular to a risk identification method, apparatus, device and storage medium based on image text. Background Art

[0002] At present, rental companies generally install GPS devices on vehicles when they rent them. This is used to monitor the operation of the vehicle and ensure its safe and compliant use. This measure not only helps rental companies understand the dynamics of the vehicle, but also allows them to quickly locate the vehicle and take corresponding measures when any abnormality occurs.

[0003] During the process of installing GPS equipment on vehicles, the construction personnel of the GPS installation supplier will record the construction process in detail, including taking photos of the construction site environment, photos of people and vehicles, channel stores, etc., and upload all the photos taken to the leasing company's system.

[0004] However, there are some problems with the existing photo uploading and management process. On the one hand, since the photos contain a large amount of on-site information, such as vehicle information, which may involve sensitive content such as privacy and business secrets, the photos may be non-compliant; on the other hand, due to the large number of photos, the workload when reviewing the photos is huge, and it is difficult to conduct a comprehensive and detailed review, resulting in potential risks being ignored.

[0005] In summary, the accuracy and efficiency of risk identification of images in the existing technology are low, and the images are prone to non-compliance. Summary of the Invention

[0006] The purpose of the embodiments of the present application is to propose a risk identification method, device, equipment and storage medium based on image text, the main purpose of which is to improve the accuracy and efficiency of risk identification and avoid image non-compliance.

[0007] First, to solve the above technical problems, the present application provides an image-based risk identification method, which adopts the following technical solutions:

[0008] Receive a vehicle rental request initiated by a user, and generate rental information according to the vehicle rental request;

[0009] Obtaining a set of images taken by a third-party company when installing a GPS device on the vehicle, and associating the set of images with the rental start information;

[0010] Using a preset image recognition technology to identify the image text of each image in the image set, matching the image text with a preset risk word library and a preset sensitive word library to obtain a matching result;

[0011] When the matching result is that the text segmentation hits the preset sensitive word library, the text segmentation is coded; when the matching result is that the text segmentation hits the preset risk word library, a warning label is assigned to the rental start information based on the association relationship, and the rental start information with the warning label is uploaded to the preset risk control management platform.

[0012] Secondly, in order to solve the above technical problems, the present application also provides an image-based risk identification device, which adopts the following technical solutions:

[0013] A rental information generation module is configured to receive a vehicle rental request initiated by a user and generate rental information based on the vehicle rental request;

[0014] An image set acquisition module is used to acquire an image set taken by a third-party company when installing a GPS device on a vehicle, and to associate the image set with the rental start information;

[0015] An image text recognition module, configured to use a preset image recognition technology to identify the image text of each image in the image set, and match the image text with a preset risk word library and a preset sensitive word library to obtain a matching result;

[0016] The image and text processing module is used to, when the matching result is that the text segmentation hits the preset sensitive word library, perform coding processing on the text segmentation; when the matching result is that the text segmentation hits the preset risk word library, assign a warning label to the rental start information based on the association relationship, and upload the rental start information with the warning label to the preset risk control management platform.

[0017] On the third aspect, in order to solve the above-mentioned technical problems, an embodiment of the present application also provides a computer device, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the image-text-based risk identification method as described above.

[0018] Fourthly, in order to solve the above-mentioned technical problems, an embodiment of the present application further provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the above-mentioned image-text-based risk identification method.

[0019] Compared with the prior art, the embodiments of the present application have the following beneficial effects:

[0020] Through automated extraction, the information fields in the vehicle rental request can be accurately extracted, avoiding the inefficiency and errors caused by manual operations, improving the processing efficiency of the vehicle rental business, and thus improving the efficiency of the subsequent risk identification process.

[0021] By obtaining a set of images taken by a third-party company when installing a GPS device on a vehicle, and establishing an association between the image set and the rental information, the risk of the rental request can be identified based on the association and the image set, thereby improving the accuracy of identifying risky rental requests; and also improving the compliance of the rental business.

[0022] By adopting the preset image recognition technology to identify the image text of each image in the image set, the image text in each image can be accurately identified, and then subsequent risk identification can be performed based on the text in the image, thereby improving the accuracy and efficiency of risk identification; and the image text is matched with the preset risk vocabulary and the preset sensitive vocabulary to obtain a matching result, and subsequent processing can be performed based on the matching result, thereby improving the compliance of the image and improving the accuracy of risk identification.

[0023] By coding the text in the image that matches the preset sensitive word library, the compliance of the image in the photo management process is ensured; by assigning warning labels to the rental information based on the text in the image that matches the preset risk word library, risk identification can be performed based on the image, improving the accuracy and efficiency of business risk identification. BRIEF DESCRIPTION OF THE DRAWINGS

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

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

[0026] Figure 2 A flowchart of an embodiment of the risk identification method based on image text according to the present application;

[0027] Figure 3 This is a schematic structural diagram of an embodiment of an image-based risk identification device according to the present application;

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

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

[0030] References to "embodiments" herein mean that a particular feature, structure, or characteristic described in connection with the embodiments may be included in at least one embodiment of the present application. The appearance of this phrase in various places in the specification does not necessarily refer to the same embodiment, nor does it constitute an independent or alternative embodiment that is mutually exclusive of other embodiments. It is understood, both explicitly and implicitly, 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, system architecture 100 may include a terminal device 101, a network 102, and a server 103. Terminal device 101 may be a laptop computer 1011, a tablet computer 1012, or a mobile phone 1013. Network 102 is a medium for providing a communication link between terminal device 101 and server 103. Network 102 may include various connection types, such as wired or wireless communication links or fiber optic cables.

[0033] The user can use the terminal device 101 to interact with the server 103 via 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 various electronic devices with a display screen and supporting web browsing. In addition to the laptop computer 1011, tablet computer 1012 or 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) player, a laptop computer and 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 web pages displayed on the terminal device 101 .

[0036] It should be noted that the image-text-based risk identification method provided in the embodiment of the present application is generally executed by a server / terminal device, and accordingly, the image-text-based 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 merely illustrative. Any number of terminal devices, networks and servers may be provided as required.

[0038] Continue to refer Figure 2 , shows a flow chart of an embodiment of the image-text-based risk identification method according to the present application. According to different needs, the order of the steps in the flow chart can be changed, and some steps can be omitted. The image-text-based risk identification method provided in the embodiment of the present application can be applied to any scenario that requires medical search, and the image-text-based risk identification method can be applied to products in these scenarios. The image-text-based risk identification method includes the following steps:

[0039] Step S201: Receive a vehicle rental request initiated by a user, and generate rental information according to the vehicle rental request.

[0040] In this embodiment, the image-based risk identification method is executed on the electronic device (eg Figure 1The server / terminal device shown in the figure) can receive the vehicle rental request initiated by the user through a wired connection or a wireless connection. It should be noted that the above-mentioned wireless connection methods may include but are not limited to 3G / 4G / 5G connections, WiFi connections, Bluetooth connections, WiMAX connections, Zigbee connections, UWB (Ultra Wide Band) connections, and other wireless connection methods currently known or developed in the future.

[0041] In this embodiment, the vehicle rental request refers to a vehicle rental request initiated by a user. For example, the user needs to rent a car of a certain brand and initiates the vehicle rental request through multiple channels, including but not limited to: the official website of the rental company, the rental APP, the rental company mini-program, the customer service hotline, and on-site registration; the rental information includes but is not limited to the contract number (unique identifier), user basic information (such as name, ID number, contact information, etc.), rental start time, rental end time, rental purpose, rental vehicle information, etc.

[0042] In this embodiment, a vehicle rental request initiated by a user through at least one application channel is received, and information fields contained in the vehicle rental request are extracted from the vehicle rental request, wherein the information fields include but are not limited to basic customer information (such as name, contact information, ID number, etc.), information of the desired rental vehicle (vehicle year, vehicle model, vehicle color, etc.), rental time range (lease start time, lease end time), and rental purpose, etc. Based on the extracted information fields, a rental vehicle corresponding to the vehicle rental request is selected from a rental vehicle database, and information about the rental vehicle is obtained to obtain associated information. Rental information corresponding to the rental request is generated based on the information fields and the associated information, and the generated rental information is stored in a preset database, and a corresponding index is established to avoid subsequent query and retrieval.

[0043] In one embodiment, the specific steps of receiving a vehicle rental request initiated by a user and generating rental information according to the vehicle rental request include:

[0044] receiving the vehicle rental request initiated by the user, performing field extraction on the vehicle rental request to obtain information fields;

[0045] Verifying the information field according to a preset verification rule;

[0046] When the verification passes, determining the rental vehicle corresponding to the information field, and obtaining information corresponding to the rental vehicle to obtain associated information;

[0047] The information field and the associated information are integrated and assigned a unique identifier to obtain the rental start information.

[0048] In this embodiment, a vehicle rental request initiated by a user is received and parsed to extract specific information from key fields to obtain information fields (such as user basic information, desired rental vehicle, rental time range, etc.). Request format verification is then performed on the information fields according to preset verification rules. The preset verification rules include field value verification (i.e., verifying whether key information is missing from the information field) and field format verification (i.e., verifying whether the information field has format errors, etc.). For example, if an ID card number is omitted from the upload or the format of the uploaded ID card number is incorrect, the verification will fail. If the verification fails, an error message is generated based on the specific circumstances of the failure and fed back to the user to guide the user to resubmit a correct vehicle rental request. If the verification passes, a rental vehicle is determined from a rental vehicle database based on the desired rental vehicle in the information field, and information about the rental vehicle is obtained to obtain associated information, including but not limited to the vehicle frame number, license plate number, vehicle wear and tear locations, and maintenance records. The information fields and associated information are integrated and assigned a unique identifier to obtain rental information corresponding to the vehicle rental request.

[0049] In this embodiment, by verifying the information fields of the vehicle rental request, it is possible to avoid the failure of subsequent rental information generation due to non-compliant information upload, thereby improving the efficiency of rental information generation and further improving the overall efficiency of the risk identification process.

[0050] In an implementable example A, customer Zhang San opens the rental APP, fills in relevant information on the vehicle rental start page and initiates a vehicle rental start request. This request includes user basic information (Name: Zhang San, ID number: 110101199001010001, Contact information: 13800138000), the expected rental vehicle (selects a Toyota Corolla), the rental time range (April 9, 2025 - April 15, 2025), etc. After the background system of the rental APP receives this rental start request, it extracts the above information fields, and then validates the extracted information fields through validation rules, checks whether there are any key information omissions in each information field, and checks whether the format of the information fields is correct. For example, it verifies whether the format of the ID number "110101199001010001" conforms to the standard format of 18 digits, and whether the rental time range "October 10, 2024 - October 15, 2024" conforms to the date format requirements; after verification, the formats of the ID number and the rental time range are both correct. At this time, it is determined that the verification is passed, and the rental vehicle corresponding to the information fields in the rental vehicle database is determined to be the Toyota Corolla selected by Mr. Zhang, and the associated information of the Toyota Corolla is obtained. The associated information includes the vehicle identification number "LFMK440F9R3000001", license plate number "Yue B12345", vehicle wear location (there is a slight scratch on the left front door), maintenance records (the last maintenance time was September 1, 2024, and the maintenance items were oil and oil filter replacement), etc. The extracted information fields (user basic information, expected rental vehicle, rental time range, etc.) and the obtained associated information (vehicle identification number, license plate number, vehicle wear location, maintenance records, etc.) are integrated. At the same time, a unique identifier is assigned to this integrated information, such as contract number PA - 2123, to obtain the integrated rental start information.

[0051] In this embodiment, through the automated extraction method, the information fields in the vehicle rental start request can be accurately extracted, avoiding the inefficiency and errors caused by manual operations, improving the processing efficiency of the vehicle rental start business, and thus improving the efficiency of the subsequent risk identification process.

[0052] Step S202: Obtain the image set taken by the third - party company when installing the GPS device on the vehicle, and establish an association relationship between the image set and the rental start information;

[0053] In this embodiment, after the start-up rental information is generated, the GPS device installation system sends an order to the corresponding GPS device installation supplier (third-party company) to generate order information, and the order information includes the work order number, construction location, unique identifier (contract number), etc. The work order number refers to the order number used to identify this installation task; the construction location is the specific construction address determined according to the location specified by the customer; the unique identifier refers to the contract number contained in the above-mentioned start-up rental information; when the third-party company receives the order information, it calls the construction personnel of the third-party company to arrive at the construction site according to the order information, and the construction personnel takes an image set when installing the GPS device and uploads it, and obtains the image set taken by the third-party company when installing the GPS device on the vehicle, wherein the image set includes but is not limited to the construction site environment, photos of people and vehicles, construction site store information, etc. According to the unique identifier (contract number), the obtained image set is associated with the start-up rental information.

[0054] In one embodiment, before associating the image set with the rental information, the method further includes:

[0055] Identify the type probability of each image in the image set by using a preset image type recognition model;

[0056] determining an image type of each image in the image set according to the type probability, and assigning a corresponding type label according to the image type;

[0057] Determine whether the type labels are repeated. If repeated, retain images with a higher probability of the type based on the same type label to obtain the deduplicated image set.

[0058] Performing missing judgment on the deduplicated image set according to preset business rules and the type label corresponding to each image in the deduplicated image set;

[0059] When there are missing images in the deduplicated image set, the labels corresponding to the missing images are extracted, missing prompt information is generated according to the labels corresponding to the missing images, and the missing prompt information is fed back to the third-party company.

[0060] In this embodiment, the above-mentioned image type recognition model refers to a pre-trained convolutional neural network. Through this pre-selected trained convolutional neural network, the type probability of each image can be output (such as the probability of a person-car photo is 90%, the probability of a construction site environment is 50%, etc.).

[0061] In this embodiment, the image set is extracted and an image preprocessing operation is performed on each image in the image set, wherein the image preprocessing operation includes but is not limited to traversing each image in the image set, converting each image into an input format that can be processed by the model, and performing image denoising, image enhancement, etc. The preprocessed images are batch-inputted into the image type recognition model, and the type probability of each image belonging to each image type is output. The image type corresponding to each image is determined by a preset probability threshold, and a corresponding type label is assigned based on each image type. When each image in the image set has a corresponding type label, the number of occurrences of each type label is counted, and type labels with a number of occurrences greater than 1 are marked as duplicates. For type labels with duplicates, their type probabilities are compared, and the image with the largest type probability is selected for retention, and other images with the same type label and a smaller probability are retained. The images are deleted to obtain a deduplicated image set, wherein in the deduplicated image set, there is only one image of each type label; the preset business rules are loaded, wherein the preset business rules refer to the image types that must be included in the image set, for example, the construction site environment, photos of people and vehicles, etc., and the deduplicated image set is judged for missing images based on the preset business rules and the type labels corresponding to each image in the deduplicated image set. When there is no missing image, the subsequent association relationship establishment operation is completed. When there is a missing image, all missing type labels are extracted, and missing prompt information is generated based on each missing type label, wherein the missing prompt information includes the representation information of the rental business, the missing image type, etc., and the missing prompt information is fed back to the third-party company to prompt the construction personnel of the third-party company to fill in the missing images until the missing image is judged.

[0062] In this embodiment, the preset image type recognition model can accurately identify the type probability of each image, thereby improving the accuracy of image type recognition; the corresponding type label is assigned according to the identified type probability to achieve automatic labeling of the image, which is convenient for the subsequent judgment of business rules; by judging whether there are duplications in the type label, the redundancy of the image can be reduced, the computing resources are reduced, and the integrity and timeliness of the image set can be improved through business rule verification.

[0063] Continuing with the above-mentioned feasible example A, after receiving the lease start information, the GPS supplier generates a dispatch information, which includes the work order number 10000, the construction location (a certain auto repair shop) and the unique identifier (contract number PA-2123); the third-party platform notifies the relevant construction personnel to prepare to perform the installation task. The construction personnel arrive at the site to install the GPS equipment according to the construction location in the dispatch information. During the installation process, the construction personnel take a set of images including the construction site environment, photos of people and vehicles, construction site shop information, etc. in accordance with the standard operating procedures, and upload these images. After receiving the image set uploaded by the supplier, the image set is automatically associated with the corresponding lease start information based on the unique identifier (contract number FA-2123) in the image set.

[0064] In this embodiment, by obtaining a set of images taken when a third-party company installs a GPS device on a vehicle, and establishing an association between the image set and the rental information, the risk of the rental request can be identified based on the association and the image set, thereby improving the accuracy of identifying risky rental requests; and also improving the compliance of the rental business.

[0065] Step S203: using a preset image recognition technology to identify the image text of each image in the image set, and matching the image text with a preset risk word library and a preset sensitive word library to obtain a matching result.

[0066] In this embodiment, the above-mentioned preset risk word library refers to a database storing multiple risk words, including but not limited to the names of blacklisted channel merchants, as well as "second deposit", "0 yuan purchase", "free", "subsidy", etc.; the sensitive words refer to words that violate laws and regulations and privacy words, etc.

[0067] In this embodiment, a preprocessing operation is performed on each image in the image set, and the preprocessing operation includes image denoising (such as Gaussian blur, median filtering, etc.), grayscale processing, image rotation correction, and a preprocessed image set is obtained. Text recognition is performed on each preprocessed image through a preset image recognition technology to obtain a text recognition result. The text recognition result is compared with a preset risk word library and a preset sensitive word library respectively to determine whether the text recognition result hits a risk word in the risk word library or hits a sensitive word in the sensitive word library.

[0068] In one embodiment, the specific step of using a preset image recognition technology to identify the image text of each image in the image set includes:

[0069] Mapping each image in the image set to a two-dimensional coordinate system, locating the text area of each image using a preset text detection algorithm, and obtaining the coordinates of the text area corresponding to each image;

[0070] According to the coordinates of the text area corresponding to each image, the preset image recognition technology is used to perform text recognition on each image to obtain the initial image text and the confidence score corresponding to the initial image text;

[0071] When the confidence score of the initial image text is greater than a preset confidence threshold, determining that the initial image text is the image text of each of the images;

[0072] When the confidence score of the initial image text is less than or equal to the preset confidence threshold, a review mark is given to the initial image text.

[0073] In this embodiment, the pixels of each image are mapped to a two-dimensional index coordinate system, with the coordinate origin set to the lower left corner of the image, the x-axis pointing right, and the y-axis pointing upward. A preset text detection algorithm, such as a target detection algorithm based on deep learning (such as YOLO, Faster R-CNN, etc.) or an edge detection algorithm, is used to generate a candidate frame of the text area. According to the aspect ratio, area, length-to-width ratio and other features of the candidate frame, the non-text area (such as larger than the maximum threshold of the candidate frame area or smaller than the minimum threshold of the candidate frame area) is filtered out to obtain the final coordinates of the text area. The corresponding text area coordinates are generated for each image, such as the coordinates (x1, x2, y1, y2) corresponding to image 1 and the coordinates (x3, x4, y3, y4) corresponding to image 1. The preset image recognition technology is used to perform text recognition on each image with text area coordinates. Recognition is performed to obtain the corresponding initial image text and the confidence score corresponding to the initial image text. If the confidence score of the initial image text is greater than the preset confidence threshold (such as 0.8), the initial image text is directly determined to be the final recognition result, that is, the image text of the image; if the confidence score of the initial image text is less than or equal to the preset confidence threshold, a review mark (such as a pending review mark) is added to the initial image text, and the initial image text with the review mark is submitted to the manual review queue or another model (such as OCR engine integration) is called for secondary recognition.

[0074] In this embodiment, by mapping the image to a two-dimensional coordinate system and using a preset text detection algorithm to locate the text area in each image, accurate detection and positioning of the text in the image is achieved, thereby improving the accuracy of subsequent image text recognition; by setting a preset confidence threshold, the recognition results are filtered to ensure that only the initial image text with a confidence score greater than the threshold is determined as the final image text, thereby improving the reliability of the text recognition results; and the automated processing from image text recognition to confidence score screening reduces the need for manual intervention and improves recognition efficiency.

[0075] In another embodiment, the specific steps of performing text recognition on each image using the preset image recognition technology based on the coordinates of the text area corresponding to each image to obtain the initial image text and the confidence score corresponding to the initial image text include:

[0076] According to the coordinates of the text area corresponding to each of the images, a corresponding sub-image is cropped from each of the images to obtain a sub-image set corresponding to each of the images;

[0077] Extracting features from each sub-image in the sub-image set using the preset image recognition technology to obtain a text feature atlas;

[0078] Performing sequence modeling on each character feature graph in the character feature graph set to generate a character sequence;

[0079] Decoding the character sequence to obtain image text corresponding to each sub-image;

[0080] Calculating the confidence score of each character in the image text corresponding to the sub-image, and calculating the average of the confidence scores of each character to obtain the confidence score of the image text corresponding to the sub-image;

[0081] The image text corresponding to all the sub-images and the confidence score corresponding to the image text corresponding to each sub-image are integrated to obtain the initial image text corresponding to each image.

[0082] In this embodiment, each image in the image set is traversed, and a corresponding sub-image is cropped from each image according to the coordinates of the text area. The cropped sub-images are preprocessed (such as resizing, normalizing pixel values to the range of [0, 1], converting to grayscale or RGB channel standardization, etc.) to obtain a sub-image set corresponding to each image, wherein each sub-image contains an independent text area; a preset image recognition technology (such as a convolutional neural network CNN, such as ResNet, CRNN, etc.) is used to extract features from each sub-image in the sub-image set, and the extracted feature map is post-processed (such as normalization, channel compression, dimensionality reduction, etc.) to reduce computational complexity and enhance feature expression capabilities to obtain a text feature map set corresponding to each sub-image; each feature map in the text feature map set is serialized to generate a character sequence, specifically: the text feature map is expanded into a sequence by row (or column), a fixed-size sliding window is used to move on the feature map to generate a local feature sequence, and the serialized features are input into a sequence modeling model (such as a recurrent neural network RNN, a long short-term memory network LSTM, a gated recurrent unit GRU or Tr ansformer, etc.), outputting a character sequence corresponding to each text feature map; decoding the character sequence to obtain the image text corresponding to each sub-image; calculating the confidence score of each character in the image text corresponding to the sub-image, and after obtaining the confidence score of each character, calculating the average of the confidence scores of each character to obtain the confidence score of the image text corresponding to the sub-image; integrating the image text corresponding to all the sub-images and the confidence scores corresponding to the image text corresponding to each sub-image to obtain the initial image text corresponding to each image.

[0083] In this embodiment, cropping an image into multiple sub-images using the coordinates of the image can reduce computing resources, increase computing speed, and avoid the influence of irrelevant factors, thereby improving the accuracy and efficiency of image recognition and reducing computing costs.

[0084] In one embodiment, after identifying the image text of each image in the image set using a preset image recognition technology, the method further includes the following specific steps:

[0085] Obtaining a confidence score for each character in the image text of each image, and marking the character as a potential erroneous character when the confidence score of the character is lower than a preset character confidence score threshold;

[0086] Performing semantic analysis on the potential erroneous characters using a preset error correction model to generate a candidate correction character set;

[0087] A target correction character is selected from a candidate correction character set according to a preset correction strategy, and the potential error character is replaced by the target correction character to obtain the corrected image text.

[0088] In this embodiment, the confidence score of each character in the image text of each of the images mentioned above is obtained, and the confidence score of each character is compared with a preset character confidence score threshold (such as 0.6). When the confidence score of a character is lower than the preset character confidence score threshold, the character is marked as a potential error character; the marked potential error characters are processed by a preset error correction model (such as a pre-trained language model such as BERT, GTP, etc.), specifically: the image text is input into the error correction model, wherein the image text at this time is marked with potential error characters, and semantic analysis and context are performed. Text association processing: The error correction model generates multiple possible correct characters as a candidate correction character set based on context semantics, grammatical rules, vocabulary collocation and other information. For each character in the candidate correction character set, the matching score between the character and the context is calculated, and the candidate characters are sorted from high to low according to the matching score. The target correction character is selected from the candidate correction character set according to the preset correction strategy, wherein the preset correction strategy refers to selecting the character with the highest matching score in the candidate correction character set as the target correction character, and using the target correction character to replace the potential error character to obtain the corrected image text.

[0089] In this embodiment, the confidence score of each character in the image text is used to determine whether it is a potential erroneous character, and then the character is modified through the error correction model, which can improve the accuracy of the image text and further improve the accuracy of subsequent text matching.

[0090] In one embodiment, the specific steps of matching the image text with a preset risk word library and a preset sensitive word library to obtain a matching result include:

[0091] Segmenting the image text using a forward maximum matching segmentation method to obtain a first segmentation set;

[0092] Segmenting the image text using a reverse maximum matching segmentation method to obtain a second segmentation set;

[0093] When the first word segmentation set and the second word segmentation set are consistent, a text word segmentation set is obtained;

[0094] When the first word segmentation set and the second word segmentation set are inconsistent, determining the first word segmentation set or the second word segmentation set as the optimal word segmentation using a preset segmentation rule to obtain the text word segmentation set;

[0095] Each word in the text word set is matched with the preset risk word library and the preset sensitive word library to obtain a matching result.

[0096] In this embodiment, the maximum segmentation word length of the forward maximum matching segmentation method is set, starting from the left side of the image text, according to the preset dictionary, substrings of the maximum segmentation word length are taken from the image text in sequence. If a matching term is found, the word is segmented and the matching continues from the remaining text. If not found, the segmentation word length is shortened (for example, minus 1) based on the previous match, and the search is repeated until the segmentation word length is 1, then the matching is stopped, and the substrings obtained by the above segmentation are collected to obtain the first segmentation set; the maximum segmentation word length of the reverse maximum matching segmentation method is set, starting from the right side of the image text, according to the preset dictionary, substrings of the maximum segmentation word length are taken from the image text in sequence. If a matching term is found, the word is segmented and the matching continues from the remaining text. If not found, the segmentation word length is shortened based on the previous match. Long length (for example, minus 1), and repeat the search until the segmentation word length is 1, stop matching, collect the substrings obtained by the above segmentation, and obtain the second segmentation set; then compare whether the first segmentation set and the second segmentation set are the same. When the first segmentation set and the second segmentation set are exactly the same, any one of the segmentation sets of the first segmentation set and the second segmentation set is used as the text segmentation set; when the first segmentation set and the second segmentation set are different, determine the first segmentation set or the second segmentation set as the optimal segmentation according to the preset segmentation rule to obtain the text segmentation set, wherein the preset segmentation rule refers to the principle of the least number of words: give priority to the result with the fewer times in the segmentation set; and the principle of the longest word length: when the number of times of the first segmentation set and the second segmentation set is the same, determine the segmentation set with the longer word length as the text segmentation set.

[0097] In this embodiment, through dual verification of forward maximum matching and reverse maximum matching, the complementarity of the two methods is utilized to reduce the error of a single word segmentation algorithm. When the two word segmentation results are inconsistent, the optimal word segmentation scheme is automatically selected through preset rules to avoid manual intervention, improve the degree of automation, and thus improve the efficiency and accuracy of the risk identification process.

[0098] Continuing with the above-mentioned feasible example A, after receiving the image set uploaded by the third-party platform, the text "Vehicle mortgage, quick loan, please contact 138xxxx8888" contained in each image in the image set is identified through image text recognition technology, and the identified text is segmented through the forward maximum matching method and the reverse maximum matching method, and the final text segmentation set "Vehicle mortgage, quick, loan, contact, 138xxxx8888" is determined through the preset segmentation rules, and the segments in the text segmentation set are matched with the preset risk vocabulary and the preset sensitive vocabulary respectively to obtain matching results.

[0099] In this embodiment, by adopting a preset image recognition technology to identify the image text of each image in the image set, the image text in each image can be accurately identified, and then subsequent risk identification can be performed based on the text in the image, thereby improving the accuracy and efficiency of risk identification; and the image text is matched with a preset risk vocabulary and a preset sensitive vocabulary to obtain a matching result, and subsequent processing can be performed based on the matching result, thereby improving the compliance of the image and improving the accuracy of risk identification.

[0100] Step S204: When the matching result is that the text segmentation hits the preset sensitive word library, the text segmentation is coded; when the matching result is that the text segmentation hits the preset risk word library, a warning label is assigned to the rental start information based on the association relationship, and the rental start information with the warning label is uploaded to the preset risk control management platform.

[0101] In this embodiment, when the matching result is that the text segmentation is consistent with a sensitive word in a preset sensitive word library, the text segmentation is determined to be a sensitive word, the image where the segmentation is located is obtained, and the image is coded.

[0102] Specifically, the coding processing methods include but are not limited to pixel blur, mosaic, color blocking, patch replacement and intelligent coding processing;

[0103] Among them, pixel blur (PixelBlur) refers to reducing the resolution of the word in the local area of ​​the image and mixing the pixels to make the word unrecognizable; Mosaic (Mosaic) refers to dividing the area of ​​the word in the image into small blocks, and each small block is filled with a single color to achieve a mosaic effect; Color Blocking (Color Blocking) refers to filling the area of ​​the word in the image with a solid color block; Patch Replacement (Patch Replacement) refers to replacing the area of ​​the word in the image with other image content (such as blurred background, random texture, etc.); Intelligent coding refers to coding the marked area of ​​the word in the image area through a deep learning model.

[0104] Continuing with the above-mentioned feasible example A, when "138xxxx8888" is identified as a mobile phone number and is found in the preset sensitive library, "138xxxx8888" is coded in the area where the image is located.

[0105] In this embodiment, when the matching result is that the text segmentation hits the preset risk vocabulary, a warning label is assigned to the rental information based on the association between the image and the rental information, and the rental information with the warning label is uploaded to the preset risk control management platform to prompt the staff to follow up.

[0106] In this embodiment, by coding the text in the image according to the preset sensitive word library, the compliance of the image in the photo management process is ensured; by assigning a warning label to the rental information according to the text in the image that hits the preset risk word library, risk identification can be performed based on the image, thereby improving the accuracy and efficiency of business risk identification.

[0107] It should be emphasized that in order to further ensure the privacy and security of the above-mentioned rental information, the above-mentioned rental information can also be stored in a node of a blockchain.

[0108] The blockchain referred to in this application refers to a new application model for computer technologies such as distributed data storage, peer-to-peer transmission, consensus mechanisms, and encryption algorithms. Blockchain is essentially a decentralized database, a series of data blocks generated using cryptographic methods. Each data block contains information about a batch of network transactions, which is used to verify the validity of the information (to prevent counterfeiting) and generate the next block. Blockchain can include the blockchain underlying platform, the platform product service layer, and the application service layer.

[0109] The embodiments of the present application can acquire and process relevant data based on artificial intelligence technology. Artificial Intelligence (AI) is the theory, method, technology, and application system that uses digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use knowledge to achieve optimal results.

[0110] Fundamental AI technologies generally include sensors, dedicated AI chips, cloud computing, distributed storage, big data processing, operating / interaction systems, and mechatronics. AI software technologies primarily encompass computer vision, robotics, biometrics, speech processing, natural language processing, and machine learning / deep learning.

[0111] Those skilled in the art will appreciate that all or part of the processes in the above-described method embodiments can be implemented by instructing related hardware via computer-readable instructions. The computer-readable instructions can be stored in a computer-readable storage medium, and when the program is executed, it can include the processes in the above-described method embodiments. The aforementioned storage medium can be a non-volatile storage medium such as a magnetic disk, an optical disk, a read-only memory (ROM), or a random access memory (RAM).

[0112] It should be understood that although the steps in the flowcharts of the accompanying drawings are shown in sequence as indicated by the arrows, these steps are not necessarily executed 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 some of the steps in the flowcharts 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 portion of the sub-steps or stages of other steps.

[0113] Further references Figure 3 , as a response to the above Figure 2 The present application provides an embodiment of a risk identification device based on image text. Figure 2 Corresponding to the method embodiment shown, the apparatus can be specifically applied to various computer devices.

[0114] like Figure 3 As shown, the image-text-based risk identification device 300 of this embodiment includes: a rental information module 301, an image set acquisition module 302, an image-text recognition module 303, and an image-text processing module 304.

[0115] The rental information generation module 301 is configured to receive a vehicle rental request initiated by a user and generate rental information according to the vehicle rental request;

[0116] In one embodiment, the rental information generation module includes:

[0117] A field extraction submodule is configured to receive the vehicle rental request initiated by the user, perform field extraction on the vehicle rental request, and obtain information fields;

[0118] A verification submodule, configured to verify the information field according to a preset verification rule;

[0119] A vehicle information acquisition submodule is configured to determine the rental vehicle corresponding to the information field when the verification is passed, and to acquire information corresponding to the rental vehicle to obtain associated information;

[0120] The integration submodule is used to integrate the information field and the associated information, and assign a unique identifier to obtain the rental start information.

[0121] An image set acquisition module 302 is used to acquire an image set taken by a third-party company when installing a GPS device on a vehicle, and to associate the image set with the rental start information;

[0122] In one embodiment, the apparatus further comprises:

[0123] A type probability recognition module, configured to recognize the type probability of each image in the image set using a preset image type recognition model;

[0124] a type label assigning module, configured to determine the image type of each image in the image set according to the type probability, and assign a corresponding type label according to the image type;

[0125] a deduplication module, configured to determine whether the type labels are repeated, and when the type labels are repeated, retain images with a higher probability of the type based on the same type label, to obtain the deduplicated image set;

[0126] A missing judgment module is used to judge missing images in the deduplicated image set according to preset business rules and the type label corresponding to each image in the deduplicated image set;

[0127] The prompt information feedback module is used to extract the labels corresponding to the missing images when there are missing images in the deduplicated image set, generate missing prompt information according to the labels corresponding to the missing images, and feed back the missing prompt information to the third-party company.

[0128] The image text recognition module 303 is configured to use a preset image recognition technology to recognize the image text of each image in the image set, and match the image text with a preset risk word library and a preset sensitive word library to obtain a matching result.

[0129] In one embodiment, the image text recognition module includes:

[0130] A mapping submodule, configured to map each image in the image set to a two-dimensional coordinate system, locate the text area of each image using a preset text detection algorithm, and obtain the coordinates of the text area corresponding to each image;

[0131] a text recognition submodule, configured to perform text recognition on each image using the preset image recognition technology according to the coordinates of the text area corresponding to each image, and obtain the initial image text and the confidence score corresponding to the initial image text;

[0132] an image text determination submodule, configured to determine that the initial image text is the image text of each image when the confidence score of the initial image text is greater than a preset confidence threshold;

[0133] The review mark determination submodule is configured to give a review mark to the initial image text when the confidence score of the initial image text is less than or equal to the preset confidence threshold.

[0134] In another embodiment, the text recognition submodule includes:

[0135] a cropping sub-unit, configured to crop a corresponding sub-image from each image according to the coordinates of the text area corresponding to each image, to obtain a sub-image set corresponding to each image;

[0136] a feature extraction subunit, configured to extract features from each sub-image in the sub-image set using the preset image recognition technology to obtain a text feature atlas;

[0137] A sequence modeling subunit, configured to perform sequence modeling on each character feature graph in the character feature graph set to generate a character sequence;

[0138] A decoding subunit, configured to decode the character sequence to obtain image text corresponding to each sub-image;

[0139] a calculation subunit, configured to calculate a confidence score of each character in the image text corresponding to the sub-image, and calculate an average of the confidence scores of each character to obtain a confidence score of the image text corresponding to the sub-image;

[0140] The integration subunit is used to integrate the image text corresponding to all the sub-images and the confidence score corresponding to the image text corresponding to each sub-image to obtain the initial image text corresponding to each image.

[0141] In one embodiment, the apparatus further comprises:

[0142] an erroneous character recognition device, configured to obtain a confidence score of each character in the image text of each image, and mark the character as a potential erroneous character when the confidence score of the character is lower than a preset character confidence score threshold;

[0143] A correction character generating device, configured to perform semantic analysis on the potential erroneous characters using a preset error correction model to generate a candidate correction character set;

[0144] The character replacement device is used to select a target correction character from a candidate correction character set according to a preset correction strategy, and use the target correction character to replace the potential error character to obtain the corrected image text.

[0145] In one embodiment, the image text recognition module includes:

[0146] A first word segmentation submodule is used to segment the image text using a forward maximum matching segmentation method to obtain a first word segmentation set;

[0147] A second word segmentation submodule is used to segment the image text using a reverse maximum matching segmentation method to obtain a second word segmentation set;

[0148] A first comparison submodule, configured to obtain a text segmentation set when the first segmentation set and the second segmentation set are consistent;

[0149] a second comparison submodule, configured to, when the first segmentation set and the second segmentation set are inconsistent, determine, using a preset segmentation rule, that the first segmentation set or the second segmentation set is the optimal segmentation set, thereby obtaining the text segmentation set;

[0150] The matching submodule is used to match each word in the text word set with the preset risk word library and the preset sensitive word library to obtain a matching result.

[0151] The image and text processing module 304 is used to, when the matching result is that the text segmentation hits the preset sensitive word library, perform coding processing on the text segmentation; when the matching result is that the text segmentation hits the preset risk word library, assign a warning label to the rental start information based on the association relationship, and upload the rental start information with the warning label to the preset risk control management platform.

[0152] In this embodiment, through automated extraction, the information fields in the vehicle rental request can be accurately extracted, avoiding inefficiency and errors caused by manual operation, improving the processing efficiency of the vehicle rental business, and further improving the efficiency of the subsequent risk identification process.

[0153] By obtaining a set of images taken by a third-party company when installing a GPS device on a vehicle, and establishing an association between the image set and the rental information, the risk of the rental request can be identified based on the association and the image set, thereby improving the accuracy of identifying risky rental requests; and also improving the compliance of the rental business.

[0154] By adopting the preset image recognition technology to identify the image text of each image in the image set, the image text in each image can be accurately identified, and then subsequent risk identification can be performed based on the text in the image, thereby improving the accuracy and efficiency of risk identification; and the image text is matched with the preset risk vocabulary and the preset sensitive vocabulary to obtain a matching result, and subsequent processing can be performed based on the matching result, thereby improving the compliance of the image and improving the accuracy of risk identification.

[0155] By coding the text in the image that matches the preset sensitive word library, the compliance of the image in the photo management process is ensured; by assigning warning labels to the rental information based on the text in the image that matches the preset risk word library, risk identification can be performed based on the image, improving the accuracy and efficiency of business risk identification.

[0156] In order to solve the above technical problems, the embodiment of the present application also provides a device (computer device). Figure 4 , Figure 4 This is a basic structural block diagram of the computer device in this embodiment.

[0157] The computer device 4 includes a memory 41, a processor 42, and a network interface 43 that are interconnected through a system bus. It should be noted that the figure only shows a computer device 4 with a memory 41, a processor 42, and a network interface 43, 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 a microprocessor, an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), a digital signal processor (DSP), an embedded device, etc.

[0158] The computer device may be a desktop computer, notebook computer, PDA, cloud server, etc. The computer device may interact with the user via a keyboard, mouse, remote control, touchpad, or voice control device.

[0159] The memory 41 includes at least one type of readable storage medium, including flash memory, hard disk, multimedia card, card-type memory (e.g., SD or DX memory), 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, magnetic disk, optical disk, etc. In some embodiments, the memory 41 may be an internal storage unit of the computer device 4, such as the hard disk or memory of the computer device 4. In other embodiments, the memory 41 may also be an external storage device of the computer device 4, such as a plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash memory card, etc. equipped on the computer device 4. Of course, the memory 41 may also include both the internal storage unit of the computer device 4 and its external storage device. In this embodiment, the memory 41 is generally used to store the operating system and various application software installed on the computer device 4, such as computer-readable instructions for the image-based risk identification method. In addition, the memory 41 can also be used to temporarily store various types of data that have been output or are to be output.

[0160] In some embodiments, the processor 42 may be a central processing unit (CPU), a controller, a microcontroller, a microprocessor, or other data processing chip. The processor 42 is typically used to control the overall operation of the computer device 4. In this embodiment, the processor 42 is used to execute computer-readable instructions or process data stored in the memory 41, such as computer-readable instructions for executing the image-based risk identification method.

[0161] The network interface 43 may include a wireless network interface or a wired network interface. The network interface 43 is generally used to establish a communication connection between the computer device 4 and other electronic devices.

[0162] During the implementation process of the electronic device of the present application, the information fields in the vehicle rental request can be accurately extracted through automated extraction, thereby avoiding the inefficiency and errors caused by manual operation, improving the processing efficiency of the vehicle rental business, and further improving the efficiency of the subsequent risk identification process.

[0163] By obtaining a set of images taken by a third-party company when installing a GPS device on a vehicle, and establishing an association between the image set and the rental information, the risk of the rental request can be identified based on the association and the image set, thereby improving the accuracy of identifying risky rental requests; and also improving the compliance of the rental business.

[0164] By adopting the preset image recognition technology to identify the image text of each image in the image set, the image text in each image can be accurately identified, and then subsequent risk identification can be performed based on the text in the image, thereby improving the accuracy and efficiency of risk identification; and the image text is matched with the preset risk vocabulary and the preset sensitive vocabulary to obtain a matching result, and subsequent processing can be performed based on the matching result, thereby improving the compliance of the image and improving the accuracy of risk identification.

[0165] By coding the text in the image that matches the preset sensitive word library, the compliance of the image in the photo management process is ensured; by assigning warning labels to the rental information based on the text in the image that matches the preset risk word library, risk identification can be performed based on the image, improving the accuracy and efficiency of business risk identification.

[0166] The present application also provides another embodiment, namely, providing a storage medium (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 above-mentioned image-text-based risk identification method.

[0167] During implementation, the computer-readable storage medium of the present application can accurately extract information fields from vehicle rental requests through automated extraction, thereby avoiding inefficiencies and errors caused by manual operations, improving the processing efficiency of vehicle rental services, and thereby improving the efficiency of subsequent risk identification processes.

[0168] By obtaining a set of images taken by a third-party company when installing a GPS device on a vehicle, and establishing an association between the image set and the rental information, the risk of the rental request can be identified based on the association and the image set, thereby improving the accuracy of identifying risky rental requests; and also improving the compliance of the rental business.

[0169] By adopting the preset image recognition technology to identify the image text of each image in the image set, the image text in each image can be accurately identified, and then subsequent risk identification can be performed based on the text in the image, thereby improving the accuracy and efficiency of risk identification; and the image text is matched with the preset risk vocabulary and the preset sensitive vocabulary to obtain a matching result, and subsequent processing can be performed based on the matching result, thereby improving the compliance of the image and improving the accuracy of risk identification.

[0170] By coding the text in the image that matches the preset sensitive word library, the compliance of the image in the photo management process is ensured; by assigning warning labels to the rental information based on the text in the image that matches the preset risk word library, risk identification can be performed based on the image, improving the accuracy and efficiency of business risk identification.

[0171] The non-Company software tools or components appearing in the embodiments of this application are merely examples and do not represent actual use.

[0172] 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 the necessary general hardware platform, and of course can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this 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 enabling 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.

[0173] Obviously, the embodiments described above are only some of the embodiments of the present application, rather than all of the 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 has been described in detail with reference to the aforementioned embodiments, for those skilled in the art, it is still possible to modify the technical solutions described in the aforementioned specific embodiments, or to make equivalent replacements for some of the technical features therein. Any equivalent structure made using the contents of the present application specification and the accompanying drawings, directly or indirectly used in other related technical fields, is also within the scope of patent protection of the present application.

Claims

1. A risk identification method based on image text, characterized in that: The method comprises: Receive a vehicle rental request initiated by a user, and generate rental information according to the vehicle rental request; Obtaining a set of images taken by a third-party company when installing a GPS device on the vehicle, and associating the set of images with the rental start information; Using a preset image recognition technology to identify the image text of each image in the image set, matching the image text with a preset risk word library and a preset sensitive word library to obtain a matching result; When the matching result is that the text segmentation hits the preset sensitive word library, the text segmentation is coded; when the matching result is that the text segmentation hits the preset risk word library, a warning label is assigned to the rental start information based on the association relationship, and the rental start information with the warning label is uploaded to the preset risk control management platform.

2. The image-based risk identification method according to claim 1, wherein: The specific steps of receiving a vehicle rental request initiated by a user and generating rental information according to the vehicle rental request include: receiving the vehicle rental request initiated by the user, performing field extraction on the vehicle rental request to obtain information fields; Verifying the information field according to a preset verification rule; When the verification passes, determining the rental vehicle corresponding to the information field, and obtaining information corresponding to the rental vehicle to obtain associated information; The information field and the associated information are integrated and assigned a unique identifier to obtain the rental start information.

3. The image-based risk identification method according to claim 1, wherein: Before establishing an association relationship between the image set and the rental information, the method further includes: Identify the type probability of each image in the image set by using a preset image type recognition model; determining an image type of each image in the image set according to the type probability, and assigning a corresponding type label according to the image type; Determine whether the type labels are repeated. If repeated, retain images with a higher probability of the type based on the same type label to obtain the deduplicated image set. Performing missing judgment on the deduplicated image set according to preset business rules and the type label corresponding to each image in the deduplicated image set; When there are missing images in the deduplicated image set, the labels corresponding to the missing images are extracted, missing prompt information is generated according to the labels corresponding to the missing images, and the missing prompt information is fed back to the third-party company.

4. The image-based risk identification method according to claim 1, wherein: The specific steps of using a preset image recognition technology to identify the image text of each image in the image set include: Mapping each image in the image set to a two-dimensional coordinate system, locating the text area of each image using a preset text detection algorithm, and obtaining the coordinates of the text area corresponding to each image; According to the coordinates of the text area corresponding to each image, the preset image recognition technology is used to perform text recognition on each image to obtain the initial image text and the confidence score corresponding to the initial image text; When the confidence score of the initial image text is greater than a preset confidence threshold, determining that the initial image text is the image text of each of the images; When the confidence score of the initial image text is less than or equal to the preset confidence threshold, a review mark is given to the initial image text.

5. The image-based risk identification method according to claim 4, wherein: The specific steps of performing text recognition on each image using the preset image recognition technology according to the coordinates of the text area corresponding to each image to obtain the initial image text and the confidence score corresponding to the initial image text include: According to the coordinates of the text area corresponding to each of the images, a corresponding sub-image is cropped from each of the images to obtain a sub-image set corresponding to each of the images; Extracting features from each sub-image in the sub-image set using the preset image recognition technology to obtain a text feature atlas; Performing sequence modeling on each character feature graph in the character feature graph set to generate a character sequence; Decoding the character sequence to obtain image text corresponding to each sub-image; Calculating the confidence score of each character in the image text corresponding to the sub-image, and calculating the average of the confidence scores of each character to obtain the confidence score of the image text corresponding to the sub-image; The image text corresponding to all the sub-images and the confidence score corresponding to the image text corresponding to each sub-image are integrated to obtain the initial image text corresponding to each image.

6. The image-based risk identification method according to claim 1, wherein: After the preset image recognition technology is used to identify the image text of each image in the image set, the method further includes the following specific steps: Obtaining a confidence score for each character in the image text of each of the images, and marking the character as a potential erroneous character when the confidence score of the character is lower than a preset character confidence score threshold; Performing semantic analysis on the potential erroneous characters using a preset error correction model to generate a candidate correction character set; A target correction character is selected from a candidate correction character set according to a preset correction strategy, and the potential error character is replaced by the target correction character to obtain the corrected image text.

7. The image-based risk identification method according to claim 1, wherein: The specific steps of matching the image text with a preset risk word library and a preset sensitive word library to obtain a matching result include: Segmenting the image text using a forward maximum matching segmentation method to obtain a first segmentation set; Segmenting the image text using a reverse maximum matching segmentation method to obtain a second segmentation set; When the first word segmentation set and the second word segmentation set are consistent, a text word segmentation set is obtained; When the first word segmentation set and the second word segmentation set are inconsistent, determining the first word segmentation set or the second word segmentation set as the optimal word segmentation using a preset segmentation rule to obtain the text word segmentation set; Each word in the text word set is matched with the preset risk word library and the preset sensitive word library to obtain a matching result.

8. A risk identification device based on image text, characterized in that: The device comprises: A rental information generation module is configured to receive a vehicle rental request initiated by a user and generate rental information based on the vehicle rental request; An image set acquisition module is used to acquire an image set taken by a third-party company when installing a GPS device on a vehicle, and to associate the image set with the rental start information; An image text recognition module, configured to use a preset image recognition technology to identify the image text of each image in the image set, and match the image text with a preset risk word library and a preset sensitive word library to obtain a matching result; The image and text processing module is used to, when the matching result is that the text segmentation hits the preset sensitive word library, perform coding processing on the text segmentation; when the matching result is that the text segmentation hits the preset risk word library, assign a warning label to the rental start information based on the association relationship, and upload the rental start information with the warning label to the preset risk control management platform.

9. A computer device, characterized in that: The computer device comprises: at least one processor; and, a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the image-text-based risk identification method according to any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the image-text-based risk identification method according to any one of claims 1 to 7 is implemented.

Images