Financial pledge survey method, device and equipment based on computer vision and medium

Abstract

The invention discloses a financial pledge survey method and device based on computer vision, equipment and a medium, and relates to the technical field of target recognition, and the method comprises the steps: obtaining a target video of a collected target house pledge, and determining a target key frame of the target video; determining a key frame set corresponding to a target house layout map and a room type of the target house collateral from the target key frames to obtain a target key frame album, and determining a target financial service template of the target house collateral; based on the target financial service template and a preset real-time target detection algorithm, performing target identification on the target key frame in the target key frame shadow set to obtain a target identification result; and according to the target identification result, the target key frame album and a preset evaluation strategy, generating a survey analysis result of the target house guarantee, and based on a preset artificial intelligence technology, a preset cue word and the survey analysis result, generating a target survey report of the target house guarantee. According to the invention, automatic investigation of the house collateral can be realized.

Description

Technical Field

[0001] The present invention relates to the field of target recognition technology, and in particular to a computer vision-based financial collateral detection method, device, equipment, and medium. Background Art

[0002] With the rapid development of computer vision technology in the field of artificial intelligence, traditional image processing techniques have evolved to a more advanced stage capable of deeply understanding and analyzing video content. In the financial sector, collateral assessment is a key step in risk management and loan approval processes. Its core goal is to accurately value the collateralized assets and control risks.

[0003] Currently, commercial banks' home renovation loan services require relationship managers to personally visit the client's renovation site to conduct on-site surveys, including data collection, photography, and video recording. They then manually compile survey reports to verify the authenticity of the loan proceeds. Given the typically large loan amounts, banks disburse loans in phases based on renovation progress. This requires relationship managers to make multiple site visits and manually compare renovation progress to ensure the authenticity of the funds used and the alignment of loan disbursement and construction progress. This service places a high demand on commercial banks' human resources, resulting in low efficiency, time-consuming, and labor-intensive work for relationship managers. Furthermore, manual judgments can be inaccurate and subjective.

[0004] In summary, how to achieve automated surveying of financial collateral and improve the accuracy and objectivity of value assessment is a technical problem that needs to be solved urgently. Summary of the Invention

[0005] In view of this, the present invention aims to provide a computer vision-based financial collateral survey method, apparatus, device, and medium that can automate the survey of financial collateral and improve the accuracy and objectivity of value assessment. The specific solution is as follows:

[0006] In a first aspect, the present application provides a financial collateral detection method based on computer vision, comprising:

[0007] Acquire the collected target video of the target house collateral, and decode the target video to determine a number of target key frames of the target video;

[0008] Determining a set of key frames corresponding to a target house layout and preset room types of the target house collateral from the plurality of target key frames to obtain a target key frame album, and determining a target financial business template corresponding to the target house collateral;

[0009] performing target recognition on the target key frames in the target key frame collection based on the target financial business template and a preset real-time target detection algorithm to obtain a corresponding target recognition result;

[0010] A survey and analysis result of the target housing collateral is generated according to the target recognition result, the target key frame album and a preset evaluation strategy, and a target survey report of the target housing collateral is generated based on a preset artificial intelligence technology, preset prompt words and the survey and analysis result.

[0011] Optionally, before acquiring the collected target video of the target house collateral, the method further includes:

[0012] Selecting the apartment structure of the target housing mortgage based on the target selection instruction to obtain a reference housing layout corresponding to the apartment structure;

[0013] Using a target drag command, and based on the orientation structure of each room in the target house collateral, drag the room elements in the reference house layout diagram to generate the target house layout diagram corresponding to the target house collateral;

[0014] A target survey route corresponding to the target house collateral is generated based on the target house layout diagram, so as to capture the target video of the target house collateral based on the target survey route.

[0015] Optionally, determining a number of target key frames of the target video includes:

[0016] storing the image data obtained after decoding the target video into a preset distributed storage system;

[0017] A plurality of target key frames of the target video are determined from the image data.

[0018] Optionally, storing the image data obtained after decoding the target video in a preset distributed storage system includes:

[0019] Configuring target decoding parameters corresponding to the target video; the target decoding parameters include target frame rate and target resolution;

[0020] Decoding the target video based on the target decoding parameters to obtain initial image data corresponding to the target video, and storing the initial image data in a memory;

[0021] The initial image data is subjected to image preprocessing according to a preset image preprocessing method to obtain target image data, and the target image data in the memory is stored in the preset distributed storage system based on a preset data redundancy strategy.

[0022] Optionally, before decoding the target video based on the target decoding parameters, the method further includes:

[0023] Determining the total number of frames of the target video and setting a corresponding target counter in the target decoding program;

[0024] Correspondingly, the process of decoding the target video based on the target decoding parameters further includes:

[0025] Determining the number of decoded frames corresponding to the target video in real time, and updating the value of the target counter in real time based on the number of decoded frames; there is a positive correlation between the number of decoded frames and the value of the target counter;

[0026] determining in real time a ratio of a value of the target counter to a total number of frames of the target video to monitor a decoding progress of the target video;

[0027] monitoring target resource utilization during decoding of the target video, and identifying error information based on the decoding progress and the target resource utilization;

[0028] If the error information is identified, the error information is recorded in a target log file, and the error information is processed based on a preset error correction mechanism.

[0029] Optionally, performing image preprocessing on the initial image data according to a preset image preprocessing method to obtain target image data includes:

[0030] The initial image data is preprocessed according to a preset image noise reduction method and a preset contrast enhancement method, and the size of the preprocessed initial image data is adjusted to a target size based on a preset image size condition to obtain the target image data, so as to determine several target key frames of the target video from the target image data.

[0031] Optionally, determining the plurality of target key frames of the target video from the target image data includes:

[0032] Performing a two-dimensional Fourier transform on each of the target image data, and determining a target high-frequency area in each of the transformed target image data;

[0033] Determining the energy proportion corresponding to each target high-frequency area, and determining the energy proportion below a preset energy proportion threshold as the target energy proportion;

[0034] determining the target image data corresponding to the target energy proportion as a blurred image, and filtering the blurred image from the target image data;

[0035] The filtered target image data is determined as a clear image, and a plurality of target key frames of the target video are determined from the target image data based on the clear image.

[0036] Optionally, determining a set of key frames corresponding to the target house layout and preset room types of the target house collateral from the plurality of target key frames includes:

[0037] Determining the target survey room in the target survey route corresponding to the target house collateral based on the target house layout and the preset room type, and matching each target key frame with each survey room to determine the target survey room corresponding to each target key frame;

[0038] The key frame set corresponding to each target key frame is determined based on the target survey room corresponding to each target key frame.

[0039] Optionally, determining a set of key frames corresponding to the target house layout and preset room types of the target house collateral from the plurality of target key frames includes:

[0040] Constructing a target three-dimensional spatial model corresponding to the target house collateral based on the preset real-time positioning and map construction technology and the target house layout map;

[0041] Establishing a corresponding target three-dimensional rectangular coordinate system in the target three-dimensional spatial model; the target three-dimensional rectangular coordinate system is a spatial coordinate system constructed with the front of the entrance of the target house mortgage as the X-axis direction, the vertically upward direction as the positive Z-axis direction, and the Y-axis direction determined according to the right-hand rule;

[0042] Determining target three-dimensional coordinate data corresponding to each target key frame based on the target three-dimensional rectangular coordinate system, and determining a target survey room corresponding to each target key frame based on the target three-dimensional coordinate data, the target house layout diagram, and the preset room type;

[0043] The key frame set corresponding to each target key frame is determined based on the target survey room corresponding to each target key frame.

[0044] Optionally, performing target recognition on the target key frame in the target key frame album based on the target financial business template and a preset real-time target detection algorithm to obtain a corresponding target recognition result includes:

[0045] Using the preset real-time target detection algorithm, and dividing each target key frame in each target key frame album based on a preset grid size condition to obtain grids corresponding to the target key frames; the grids are used to detect target objects in the target key frames;

[0046] Generating a preset number of initial bounding boxes corresponding to the grid using the preset real-time object detection algorithm, and determining initial confidences corresponding to the initial bounding boxes and class probabilities corresponding to the initial bounding boxes;

[0047] Target recognition is performed on the target key frame according to the initial bounding box, the initial confidence corresponding to the initial bounding box, the category probability corresponding to the initial bounding box, a preset non-maximum suppression algorithm and the target financial business template to obtain the target recognition result.

[0048] Optionally, the performing target recognition on the target key frame according to the initial bounding box, the initial confidence corresponding to the initial bounding box, the category probability corresponding to the initial bounding box, a preset non-maximum suppression algorithm, and the target financial business template to obtain the target recognition result includes:

[0049] Determining an object confidence corresponding to the initial bounding box based on a product of an initial confidence corresponding to the initial bounding box and a category probability corresponding to the initial bounding box;

[0050] Target recognition is performed on the target key frame based on the target confidence, the preset non-maximum suppression algorithm and the target financial business template to obtain the target recognition result.

[0051] Optionally, the performing target recognition on the target key frame based on the target confidence, the preset non-maximum suppression algorithm, and the target financial business template to obtain the target recognition result includes:

[0052] Sorting the initial bounding boxes based on the target confidence to obtain a first bounding box list, and initializing a target result set;

[0053] Determining an initial bounding box with the highest target confidence from the first bounding box list, and determining the initial bounding box with the highest target confidence as the target bounding box;

[0054] adding the target bounding box to the target result set and removing the target bounding box from the first bounding box list;

[0055] determining an initial IoU between the target bounding box and each of the remaining initial bounding boxes in the first bounding box list, and determining a target IoU greater than a preset IoU threshold from the initial IoU;

[0056] The initial bounding box corresponding to the target intersection-over-union ratio is removed from the first bounding box list, and the process proceeds to the step of determining the initial bounding box with the highest target confidence as the target bounding box until the first bounding box list is empty, and the target recognition result corresponding to the target key frame is generated based on the target financial business template and the target result set.

[0057] Optionally, generating a survey and analysis result of the target housing collateral according to the target recognition result, the target key frame album, and a preset evaluation strategy includes:

[0058] generating a target survey summary of the target housing collateral according to the target identification result;

[0059] Deduplication processing is performed on the target recognition result based on the target key frame album, and a value assessment result corresponding to the target house collateral is generated based on the deduplication target recognition result;

[0060] generating an integrity assessment result corresponding to the target housing collateral based on the target recognition result after deduplication and the target key frame album, and performing a risk assessment on the target housing collateral according to the target recognition result after deduplication and a preset risk assessment indicator to obtain a corresponding risk assessment result;

[0061] The survey and analysis result of the target housing collateral is generated based on the target survey summary, the value assessment result, the integrity assessment result, and the risk assessment result.

[0062] Optionally, performing risk assessment on the target housing collateral according to the target identification result after deduplication and a preset risk assessment index to obtain a corresponding risk assessment result includes:

[0063] Evaluating the rationality of the valuation of the target housing collateral based on the value evaluation result to obtain a valuation rationality evaluation result, and performing an abnormal risk evaluation on the target housing collateral based on the target identification result after deduplication to obtain an abnormal risk evaluation result;

[0064] A risk assessment result corresponding to the target housing collateral is generated based on the valuation rationality assessment result, the integrity assessment result, the abnormal risk assessment result, and the overall risk assessment result corresponding to the target housing collateral.

[0065] Optionally, the generating of the target survey report of the target housing collateral based on the preset artificial intelligence technology, preset prompt words and the survey analysis results includes:

[0066] Performing data structuring processing on the initial survey data corresponding to the survey and analysis result based on a preset data structure, and determining target survey data corresponding to the initial survey data after the data structuring processing according to a preset text format;

[0067] The target survey report of the target house mortgage is generated according to the preset generative artificial intelligence technology, the preset prompt words and the target survey data.

[0068] In a second aspect, the present application provides a financial collateral detection device based on computer vision, comprising:

[0069] a target key frame determination module, configured to obtain a captured target video of the target housing collateral and decode the target video to determine a number of target key frames of the target video;

[0070] a target key frame album determining module, configured to determine, from the plurality of target key frames, a set of key frames corresponding to the target house layout and preset room types of the target house collateral, to obtain a target key frame album, and determine a target financial business template corresponding to the target house collateral;

[0071] a target recognition result determination module, configured to perform target recognition on the target key frames in the target key frame collection based on the target financial business template and a preset real-time target detection algorithm to obtain a corresponding target recognition result;

[0072] A target survey report generation module is used to generate a survey and analysis result of the target housing collateral based on the target recognition result, the target key frame album and a preset evaluation strategy, and to generate a target survey report of the target housing collateral based on a preset artificial intelligence technology, preset prompt words and the survey and analysis result.

[0073] In a third aspect, the present application provides an electronic device, comprising:

[0074] Memory, used to store computer programs;

[0075] A processor is configured to execute the computer program to implement the aforementioned financial collateral survey method based on computer vision.

[0076] In a fourth aspect, the present application provides a computer-readable storage medium for storing a computer program; wherein, when the computer program is executed by a processor, the aforementioned computer vision-based financial collateral survey method is implemented.

[0077] In this application, a target video of a target mortgaged property is first acquired and decoded to determine a number of target key frames of the target video. A set of key frames corresponding to the target house layout and preset room types of the target mortgaged property is then determined from the target key frames to obtain a target key frame album, and a target financial business template corresponding to the target mortgaged property is determined. Target recognition is then performed on the target key frames in the target key frame album based on the target financial business template and a preset real-time target detection algorithm to obtain a corresponding target recognition result. Finally, a survey and analysis result of the target mortgaged property is generated based on the target recognition result, the target key frame album, and a preset evaluation strategy. A target survey report of the target mortgaged property is generated based on a preset artificial intelligence technology, preset prompt words, and the survey and analysis result. As can be seen from the above, in this application, a target video of the target mortgaged property is first acquired and decoded to determine a number of target key frames therein. A set of key frames corresponding to the target key frames is then determined based on the target house layout and preset room types of the target mortgaged property to obtain a target key frame album. Then, based on the target financial business template corresponding to the target housing mortgage and the preset real-time target detection algorithm, the target keyframe is identified to obtain the corresponding target recognition result. Finally, the survey and analysis results are generated based on the target recognition results, the target keyframe album and the preset evaluation strategy, and then the target survey report is generated using the preset artificial intelligence technology, the preset prompt words and the survey and analysis results. In this way, the present application can achieve a significant reduction in human resource costs and improve business processing efficiency by obtaining the target video of the target housing mortgage and directly performing target recognition on the target video; the target video of the target housing mortgage can be automatically recognized by using the target financial business template and the preset real-time target detection algorithm; the target survey report of the target housing mortgage can be automatically output based on the target recognition result by using the preset artificial intelligence technology and the preset prompt words. In this way, the present application can realize the automated survey of the target housing mortgage, improve the accuracy and objectivity of the value assessment, and reduce the operating costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0078] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are merely embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying any creative work.

[0079] Figure 1 A flowchart of a computer vision-based financial collateral detection method provided in this application;

[0080] Figure 2A specific video capture flow chart of the target house collateral provided for this application;

[0081] Figure 3 A schematic diagram of the layout of a specific target housing mortgage provided for this application;

[0082] Figure 4 A schematic diagram of video capture of a specific target house mortgage provided for this application;

[0083] Figure 5 A target identification flow chart of a specific target housing collateral provided for this application;

[0084] Figure 6 A schematic diagram of the identification results of a specific target housing mortgage provided for this application;

[0085] Figure 7 A schematic diagram of the structure of a financial collateral detection device based on computer vision provided in this application;

[0086] Figure 8 This is a structural diagram of an electronic device provided in this application. DETAILED DESCRIPTION

[0087] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0088] With the rapid development of computer vision technology in the field of artificial intelligence, traditional image processing technology has evolved to an advanced stage capable of deeply understanding and analyzing video content. In the financial sector, collateral surveying is a critical step in risk management and loan approval processes. Its core goal is to accurately value and control the risk of collateralized assets. Currently, in commercial banks' home renovation loan business, relationship managers must personally visit the client's renovation site to conduct on-site surveys, including data collection, photography, and video recording. They then manually prepare survey reports to verify the authenticity of the loan funds' use. Given that loan amounts are typically large, banks must disburse loans in phases based on the progress of renovations. Relationship managers must revisit the site multiple times and manually compare renovation progress to ensure the authenticity of the funds' use and the alignment of loan disbursement progress with construction progress. This business requires a high level of human resources from commercial banks, resulting in low efficiency, time-consuming, and labor-intensive work for relationship managers. Furthermore, manual judgment can be inaccurate and subjective. Therefore, this application provides a computer vision-based financial collateral survey solution that automates the survey of financial collateral and improves the accuracy and objectivity of value assessments.

[0089] See also Figure 1 As shown, an embodiment of the present invention discloses a financial collateral survey method based on computer vision, which may include:

[0090] Step S11: Acquire the collected target video of the target house mortgage, and decode the target video to determine a number of target key frames of the target video.

[0091] In this embodiment, customers can use their personal mobile devices to complete the video collection of the target house collateral by themselves, and upload the collected target video to the system server. Figure 2As shown, the specific operation process is as follows: the customer first accesses the H5 website (HTML, HyperText Markup Language, i.e. Hypertext Markup Language) address link provided in this embodiment through a mobile device to perform authorized login and real-name authentication to ensure the security and legality of the operation. After completing the login authentication, the customer needs to complete the association with a specific financial business in the system interface, such as the housing renovation loan business of a commercial bank. Before the above-mentioned acquisition of the collected target video of the target house mortgage, it may also include: first, based on the target check instruction, the apartment structure of the target house mortgage is selected to obtain a reference house layout corresponding to the apartment structure; then, through the target drag instruction, and based on the orientation structure of each room in the target house mortgage, the room elements in the reference house layout are dragged to generate the target house layout corresponding to the target house mortgage; finally, based on the target house layout, a target survey route corresponding to the target house mortgage is generated, so as to collect the target video of the target house mortgage based on the target survey route. See Figure 3 As shown, in the system interface, the customer first selects the floor plan structure of the target house mortgage, for example: three bedrooms and two living rooms, and the system generates the corresponding reference house layout diagram; then, the customer can drag the interface image to the position of the room in the reference house layout diagram, configure the floor plan and orientation structure of the target house mortgage, and generate a detailed target house layout diagram corresponding to the target house mortgage. Figure 4 As shown, based on the target house layout diagram of the target house collateral, the system can automatically generate a target survey route and guide the customer to record the on-site video along the target survey route. During the video recording process, the system can provide real-time route prompts to ensure the comprehensiveness and accuracy of the video collection.

[0092] It is understandable that the above-mentioned determination of several target key frames of the target video may include: first storing the image data obtained after decoding the target video in a preset distributed storage system; and then determining several target key frames of the target video from the image data. Specifically, the above-mentioned storage of the image data obtained after decoding the target video in a preset distributed storage system may include: first configuring the target decoding parameters corresponding to the target video; the target decoding parameters include the target frame rate and the target resolution; then decoding the target video based on the target decoding parameters to obtain the initial image data corresponding to the target video, and storing the initial image data in the memory; finally, performing image preprocessing on the initial image data according to a preset image preprocessing method to obtain the target image data, and storing the target image data in the memory in the preset distributed storage system based on the preset data redundancy strategy. In this way, the problem of data loss can be solved by storing the target image data in the memory in the preset distributed storage system through the data redundancy strategy.

[0093] It should be noted that the above-mentioned image preprocessing of the initial image data according to the preset image preprocessing method to obtain the target image data may include: performing image preprocessing on the initial image data according to a preset image noise reduction method and a preset contrast enhancement method, and adjusting the size of the preprocessed initial image data to the target size based on a preset image size condition to obtain the target image data, so as to determine the target key frames of the target video from the target image data. Specifically, in order to reduce the noise of the image, the current pixel value may be replaced by the mean of the pixels in the pixel neighborhood to smooth the noise; the current pixel value may be replaced by the median of the pixel neighborhood to remove impulse noise such as salt and pepper noise; the pixel neighborhood may be weighted averaged according to the Gaussian distribution to effectively remove Gaussian noise; the image may be decomposed into different frequency sub-bands using wavelet transform, and the noise may be removed by thresholding the high-frequency sub-bands while retaining important details of the image. To enhance image contrast, the image's histogram can be transformed to make the image's grayscale distribution more uniform, thereby enhancing overall contrast. Histogram equalization can be performed within local regions to better meet the contrast requirements of different image regions and avoid local over- or under-enhancement that may result from global histogram equalization. In this embodiment, to reduce computational complexity and meet the image size constraints of the algorithm input, a target image size can be determined, and the preprocessed initial image data can be resized to the target size to obtain the target image data.

[0094] Specifically, after video capture is complete, the system automatically decodes the uploaded target video and converts the video file into an analyzable data format. The system then efficiently stores the decoded image and audio data and performs a series of preprocessing operations, including but not limited to image denoising, contrast enhancement, and resizing, to optimize image data quality and lay the foundation for subsequent feature extraction and analysis. The logic for decoding and storing the target video is as follows:

[0095] A[Start]-->B[Customer uses acquisition equipment to capture video]

[0096] B-->C[The capture device saves the video in a specific format]

[0097] C-->D[Customer uploads the video to the system server]

[0098] D-->E [Server receives video and performs preliminary storage]

[0099] E-->F[The system decodes the video]

[0100] F-->G [The decoded image sequence and audio signal are stored in memory]

[0101] G-->H[Preprocess the image]

[0102] H-->I[Store the pre-processed image in the distributed storage system]

[0103] I-->J[Data processing completed, image data can be used for subsequent analysis]

[0104] J-->K[end]

[0105] The image preprocessing includes:

[0106] G-->G1[Image denoising]

[0107] G1->G2[Image contrast enhancement]

[0108] G2-->G3 [Image Size Adjustment]

[0109] In this embodiment, to avoid errors caused by randomly determining key frames, the image clarity can be determined first to avoid identifying blurred image data as key frames. Specifically, image clarity can be determined using methods such as FFT (Fast Fourier Transform) frequency domain analysis and Laplacian (Laplace Operator) gradient methods.

[0110] In one specific embodiment, when determining image clarity through FFT frequency domain analysis, each target image data can first be subjected to a two-dimensional Fourier transform, and a target high-frequency region within each transformed target image data can be determined. The energy fraction corresponding to each target high-frequency region can then be determined, and the energy fraction below a preset energy fraction threshold can be determined as the target energy fraction. The target image data corresponding to the target energy fraction can then be determined as a blurred image, and the blurred image can be filtered from the target image data. Finally, the filtered target image data can be determined as a clear image. It should be noted that the high-frequency region for each image data requires specific definition. For low-texture images, such as solid color areas, the high-frequency region radius can be reduced, while for high-texture images, such as furniture details, the high-frequency region radius can be increased. Because high-frequency distributions vary significantly across different image contents, such as complex textures and large areas of solid color, to avoid the inability of a single preset threshold to adapt to the energy distribution differences across different scenes, image noise can first be estimated using spatial variance or frequency-domain low-frequency energy, and the preset energy fraction threshold can be dynamically adjusted. Spatial gradient features and frequency-domain energy features can then be calculated in parallel, and weighted fusion, such as through SVM (Support Vector Machine) scoring, can be used to improve judgment accuracy. Furthermore, to avoid filtering out images with locally clear areas, such as frames in a video focused on furniture with a blurred background, images with slightly blurred global images but high-frequency energy that does not fall below a threshold are retained. The image can be divided into overlapping sub-blocks, such as an 8×8 pixel window, and high-frequency energy is independently calculated for each sub-block. Only images with sub-block blur exceeding a certain percentage are filtered out. Edge detection algorithms, such as the Sobel operator, are then used to calculate the mean gradient magnitude of edge pixels to compensate for the lack of high-frequency energy.

[0111] In another specific embodiment, in the process of determining the clarity of an image by the Laplacian gradient method, a Laplacian operator, such as a 3×3 convolution kernel, can be used to calculate the second-order derivative of the image, and then the absolute value or variance of the Laplacian response is counted; the larger the absolute value or variance, the clearer the image.

[0112] It should be noted that after obtaining a clear image, several target key frames of the target video can be determined from the target image data based on the clear image. In one specific embodiment, the target key frames can be determined from the clear image based on time intervals. Specifically, the sampling interval can be dynamically adjusted based on the rate of change of the image content. When the image content changes rapidly, the sampling interval can be shortened; when the image content changes slowly, the sampling interval can be lengthened. The rate of change of the image content can be determined by calculating differences between adjacent frames, such as pixel differences or color histogram differences. In another specific embodiment, motion detection algorithms, such as background subtraction and frame difference methods, can be used to detect moving objects in the video. When a moving object is detected or a significant change in motion is detected, the current image frame is designated as a key frame. Specifically, the optical flow field can be calculated between adjacent images, and the magnitude and direction of the optical flow vector can be used to determine the degree of dynamic change in the image. When the optical flow change exceeds a certain threshold, it indicates that an object has moved or the perspective has changed significantly, and the image frame can be retained as a key frame. For example, when the photographer moves through different rooms or discovers a new object, the corresponding image may contain valuable information. In addition, the shot switching in the video can be detected. The first frame after the shot switching is often the beginning of a new scene or a new perspective. Using the first frame after the shot switching as a key frame can effectively segment different content segments. In a third specific implementation, the stability can be evaluated by calculating the degree of jitter of the image sequence, and the image motion estimation parameters, such as translation, rotation, and scaling parameters, can be used to determine the degree of image jitter. Images with severe jitter may be difficult to accurately identify due to unstable shooting and should be eliminated. Images with stable shooting, clear pictures, and no large shaking should be retained as key frames to ensure the accuracy of subsequent recognition.

[0113] Step S12: determining a set of key frames corresponding to the target house layout and preset room types of the target house collateral from the plurality of target key frames to obtain a target key frame album, and determining a target financial business template corresponding to the target house collateral.

[0114] In this embodiment, the features of the target keyframes are first extracted, converting the image data in the video into numerical feature vectors that can represent the content. These feature vectors can be used for subsequent target detection, recognition, and behavior analysis, forming the basis for automated surveying. It is understood that multiple sets of financial business templates can be preset in this embodiment, such as a renovation construction progress survey tree. Financial business templates refer to a series of standard models predefined based on the needs of different financial businesses. For example, in the housing renovation loan business, the financial business templates can include renovation progress standards for different stages, layout of key areas, and so on.

[0115] See also Figure 5As shown in the figure, to improve target recognition accuracy and reduce the model's computational burden, target keyframes can be matched and reordered based on the target survey route and preset room types. Target keyframes with successful route matching are assigned to the corresponding keyframe set, resulting in a target keyframe album. The survey route includes various surveyed rooms, such as the master bedroom, secondary bathroom #1, secondary bathroom #2, and the entryway; each room type is considered an album of keyframes.

[0116] In a specific embodiment, the above-mentioned determination of the key frame set corresponding to the target house layout and preset room types of the target house collateral from the plurality of target key frames may include: firstly determining the survey rooms in the target survey route corresponding to the target house collateral based on the target house layout and the preset room types, and matching each target key frame with each survey room to determine the target survey room corresponding to each target key frame; and then determining the key frame set corresponding to each target key frame based on the target survey room corresponding to each target key frame. Specifically, the target survey route can be stored in the form of structured data, such as using a list or tree structure to clarify the order and hierarchical relationship of each room, for example [entrance, master bedroom, secondary bathroom #1, secondary bathroom #2]. A corresponding key frame album is established for each room type, which can be stored using a dictionary or hash table, with the key being the room type name and the value being the container for storing the key frames. During the target video capture process, metadata can be added to each frame using methods such as GPS (Global Positioning System) positioning, inertial sensors, or manual annotation to record the shooting location and room information corresponding to each frame. If GPS positioning is used, the mapping of geographic coordinates to rooms must be performed using a layout diagram of the target house. If manual annotation is used, the camera operator must mark the current room during the capture process. The keyframe metadata is then matched against the room information in the survey route. The shooting location or room annotation information of each keyframe is checked one by one and compared with the room names in the survey route. If the names match, the keyframe is considered to be a successful match with the corresponding room. The successfully matched keyframes are then rearranged according to the order of the rooms in the survey route. Keyframes corresponding to the rooms appearing first are sorted first, and keyframes corresponding to the rooms appearing later are sorted later. During the sorting process, data structure sorting algorithms, such as bubble sort and quick sort, can be used to sort the target keyframes according to their rooms' positions in the survey route. Afterwards, the reordered target key frames are stored in corresponding target key frame albums according to the room types.

[0117] In another specific embodiment, determining a set of keyframes corresponding to the target house layout and preset room types of the target house property from the plurality of target keyframes may include: first constructing a target three-dimensional spatial model corresponding to the target house property based on a preset real-time positioning and mapping technology and the target house layout; then establishing a corresponding target three-dimensional rectangular coordinate system in the target three-dimensional spatial model; the target three-dimensional rectangular coordinate system being a spatial coordinate system constructed with the X-axis directly in front of the entrance of the target house property, the Z-axis vertically upward, and the Y-axis direction determined according to the right-hand rule; then determining target three-dimensional coordinate data corresponding to each target keyframe based on the target three-dimensional coordinate data, the target house layout, and the preset room types; and finally determining the set of keyframes corresponding to each target keyframe based on the target survey room corresponding to each target keyframe. Specifically, the customer may select a suitable camera, such as a camera with a depth sensor or a device equipped with both a high-definition camera and an IMU (Inertial Measurement Unit). The depth sensor directly captures scene depth information, while the IMU records the device's motion information, such as rotation and acceleration, providing data support for the SLAM (Simultaneous Localization and Mapping) algorithm. When the camera begins operating, the SLAM system starts and initializes. Using the device's first captured image frame and sensor data, the system determines the initial camera position and pose, which serve as a reference point for subsequent calculations. The SLAM algorithm extracts features from each frame. A feature matching algorithm compares feature points between adjacent frames to identify identical feature point pairs, thereby determining the relative motion of the cameras at different moments in time. The IMU motion information and feature point matching results are then combined to calculate the camera's spatial trajectory, including translation and rotation. Through continuous iterative optimization, the camera's position and pose at each moment are accurately estimated. Based on the camera's position and image information, a 3D spatial model of the target housing collateral is gradually constructed. A 3D rectangular coordinate system is then established with the house entrance as the origin. Typically, the positive X-axis direction is defined as directly in front of the building entrance, the positive Z-axis direction is defined as vertically upward, and the Y-axis direction is determined using the right-hand rule. Next, the camera pose information corresponding to the target keyframe is determined—that is, the camera's position and orientation in 3D space when the target keyframe was captured. Based on the camera pose information of the target keyframe, the X, Y, and Z offsets of the target keyframe relative to the building entrance, i.e., the coordinate origin, are calculated. Using spatial geometric transformations and coordinate conversion formulas, the camera's position information is converted into 3D coordinates in the set coordinate system, resulting in the target 3D coordinate data corresponding to the target keyframe.The target key frame is located in the target three-dimensional space model of the target house collateral, and the room type corresponding to the target key frame is determined. The target key frame is stored in the corresponding target key frame album according to the room type.

[0118] It should be noted that each target key frame album can be stored in a file system, with folders named after the room type, and the target key frames stored in the corresponding folders in the form of images; it can also be stored in a database to establish a table structure containing target key frame data and metadata, such as shooting time, room type, etc.

[0119] Step S13: performing target recognition on the target key frames in the target key frame album based on the target financial business template and a preset real-time target detection algorithm to obtain corresponding target recognition results.

[0120] In this embodiment, the target recognition of the target keyframes in the target keyframe collection based on the target financial business template and a preset real-time target detection algorithm to obtain corresponding target recognition results may include: first, using the preset real-time target detection algorithm and dividing each target keyframe in the target keyframe collection based on a preset grid size condition to obtain grids corresponding to the target keyframes; the grids are used to detect target objects in the target keyframes; then, using the preset real-time target detection algorithm to generate a preset number of initial bounding boxes corresponding to the grids, and determining initial confidences and class probabilities corresponding to the initial bounding boxes; finally, performing target recognition on the target keyframes based on the initial bounding boxes, the initial confidences and class probabilities corresponding to the initial bounding boxes, a preset non-maximum suppression algorithm, and the target financial business template to obtain the target recognition results. In one specific embodiment, the system performs grid division on each target keyframe based on the YOLO (You Only Look Once) algorithm, dividing the target keyframes into 7×7 grids. Each grid is responsible for detecting target objects whose center points fall within the grid. Each grid can predict B initial bounding boxes, each of which contains location information (x, y, w, h) and confidence (i.e., Confidence Score). The system predicts the initial bounding box, which includes the location of the initial bounding box (x, y, w, h) and the confidence of the initial bounding box. Among them, (x, y) refers to the offset of the center point of the initial bounding box relative to the upper left corner of the grid; (w, h) refers to the width and height of the initial bounding box; the confidence of the initial bounding box can be measured by IOU (Intersection over Union), which indicates the probability of the target being contained in the initial bounding box and the accuracy of the bounding box. At the same time, the system predicts the probability distribution of the category to which each initial bounding box belongs, and obtains the category probability corresponding to the initial bounding box. In the current housing mortgage survey scenario, categories include "wall", "ground", "indoor facilities", etc. The confidence calculation formula is as follows:

[0121] ;

[0122] in, Indicates the probability that the initial bounding box contains the target object; Represents the intersection-over-union ratio between the predicted initial bounding box and the true bounding box.

[0123] Subsequently, target recognition is performed on the target keyframe based on the initial bounding box confidence and the class probability of the initial bounding box to determine the specific object identified and its location. This process of performing target recognition on the target keyframe based on the initial bounding box, the initial confidence corresponding to the initial bounding box, the class probability corresponding to the initial bounding box, a preset non-maximum suppression algorithm, and the target financial business template to obtain the target recognition result may include: first, determining the target confidence corresponding to the initial bounding box based on the product of the initial confidence corresponding to the initial bounding box and the class probability corresponding to the initial bounding box; and finally, performing target recognition on the target keyframe based on the target confidence, the preset non-maximum suppression algorithm, and the target financial business template to obtain the target recognition result. That is, in this embodiment, the product of the initial confidence corresponding to the initial bounding box and the class probability corresponding to the initial bounding box may be used to determine the target confidence corresponding to the initial bounding box. Subsequently, to ensure that each target object is detected only once, in this embodiment, redundant bounding boxes within the initial bounding box may be removed through non-maximum suppression, retaining the recognition result with the highest confidence and minimal overlap with other bounding boxes.

[0124] It should be noted that the above-mentioned target recognition of the target key frame based on the target confidence, the preset non-maximum suppression algorithm, and the target financial business template to obtain the target recognition result may include: first, sorting each of the initial bounding boxes based on the target confidence to obtain a first bounding box list, and initializing a target result set; then, determining the initial bounding box with the highest target confidence from the first bounding box list, and determining the initial bounding box with the highest target confidence as the target bounding box; then, adding the target bounding box to the target result set, and removing the target bounding box from the first bounding box list; then, determining an initial intersection-over-union (IoU) between the target bounding box and each of the remaining initial bounding boxes in the first bounding box list, and determining a target IoU greater than a preset IoU threshold from the initial IoU ratios; finally, removing the initial bounding box corresponding to the target IoU ratio from the first bounding box list, and jumping to the step of determining the initial bounding box with the highest target confidence as the target bounding box, until the first bounding box list is empty, and generating the target recognition result corresponding to the target key frame based on the target financial business template and the target result set. Specifically, this embodiment can combine the location information of the initial bounding box, the house type corresponding to the target key frame album, and the target confidence of the initial bounding box to ultimately determine the detected target object and its location. The specific process is as follows:

[0125] (1) Sort by target confidence of the initial bounding box: Sort all the initial bounding boxes by their target confidence from high to low, and generate the first bounding box list based on the sorted initial bounding boxes, while initializing the target result set;

[0126] (2) Select the initial bounding box with the highest target confidence: select the initial bounding box with the highest confidence in the first bounding box list as the target bounding box, add the target bounding box to the target result set, and remove the target bounding box from the first bounding box list;

[0127] (3) Calculate the target IOU: Calculate the target IOU between the target bounding box and each of the remaining initial bounding boxes in the first bounding box list; the IOU calculation formula is: IOU = intersection area / union area;

[0128] (4) Removing overlapping initial bounding boxes: If the target IOU between the initial bounding box and the target bounding box is greater than a preset intersection-over-union ratio threshold, such as 0.5, then remove it from the first bounding box list;

[0129] (5) Repeat steps (2)-(4): until all initial bounding boxes in the first bounding box list are processed and the target result set is obtained.

[0130] See also Figure 6 As shown, in this embodiment, through the YOLO algorithm and the preset route mechanism (ie, The Mechanism of Predefined Survey Routes), the system can identify the target object and the status of the target object in the target key frame.

[0131] Step S14: generating a survey and analysis result of the target housing collateral according to the target recognition result, the target key frame album and a preset evaluation strategy, and generating a target survey report of the target housing collateral based on a preset artificial intelligence technology, preset prompt words and the survey and analysis result.

[0132] In this embodiment, the system can automatically evaluate the value of the target house mortgage based on the target recognition results and generate a survey and analysis result of the target house mortgage. Afterwards, the system can automatically generate a target survey report of the target house mortgage based on the survey and analysis result of the target house mortgage according to a preset prompt word, i.e., a preset Prompt, and using generative artificial intelligence technology.

[0133] As can be seen from the above, this embodiment first acquires a target video of the target mortgaged property and decodes the target video to determine several target keyframes in the target video. Then, from the target keyframes, a set of keyframes corresponding to the target house layout and preset room types of the target mortgaged property is determined to obtain a target keyframe album, and a target financial business template corresponding to the target mortgaged property is determined. Then, based on the target financial business template and a preset real-time target detection algorithm, target recognition is performed on the target keyframes in the target keyframe album to obtain corresponding target recognition results. Finally, based on the target recognition results, the target keyframe album, and a preset evaluation strategy, a survey and analysis result of the target mortgaged property is generated. A target survey report for the target mortgaged property is generated based on preset artificial intelligence technology, preset prompt words, and the survey and analysis result. As can be seen from the above, this embodiment first acquires a target video of the target mortgaged property and decodes the video to determine several target keyframes therein. Then, based on the target house layout and preset room types of the target mortgaged property, a set of keyframes corresponding to the target keyframes is determined to obtain a target keyframe album. Then, based on the target financial business template corresponding to the target housing collateral and a preset real-time target detection algorithm, the target keyframes are identified to obtain the corresponding target recognition results. Finally, a survey and analysis result is generated based on the target recognition results, the target keyframe album, and the preset evaluation strategy. A target survey report is then generated using the preset artificial intelligence technology, preset prompt words, and the survey and analysis results. In this way, by acquiring a target video of the target housing collateral and directly performing target recognition on the target video, this embodiment can significantly reduce human resource costs and improve business processing efficiency. The target financial business template and the preset real-time target detection algorithm can automatically perform target recognition on the target video of the target housing collateral. The preset artificial intelligence technology and preset prompt words can automatically output a target survey report for the target housing collateral based on the target recognition results. In this way, this embodiment can achieve automated surveying of the target housing collateral, improve the accuracy and objectivity of value assessment, and reduce operating costs.

[0134] Based on the previous embodiment, it can be seen that the present application can obtain a target video of a target house mortgage, decode the target video, determine the target key frame, and then perform target recognition on the target key frame to obtain a target recognition result. Next, this embodiment will explain in detail how to handle error information during the video decoding process.

[0135] In this embodiment, to promptly handle errors during the target video decoding process, before decoding the target video based on the target decoding parameters, the process may further include determining the total number of frames of the target video and setting a corresponding target counter in the target decoding program. Specifically, during the decoding process, to monitor the decoding progress of the target video in real time, the total number of frames of the target video may be obtained when the target video decoding task is initiated, and used as a total indicator of the decoding progress.

[0136] Accordingly, the above-mentioned process of decoding a target video based on target decoding parameters may further include: first, determining the number of decoded frames corresponding to the target video in real time, and updating the value of the target counter in real time based on the number of decoded frames; the number of decoded frames and the value of the target counter are positively correlated; then, determining in real time the ratio of the value of the target counter to the total number of frames of the target video to monitor the decoding progress of the target video; monitoring the target resource utilization corresponding to the decoding process of the target video, and identifying error information based on the decoding progress and the target resource utilization; if the error information is identified, recording the error information to a target log file and handling the error information based on a preset error correction mechanism. Specifically, in this embodiment, a counter can be set in the decoding program, and the counter value is incremented by one each time a frame of the target video is successfully decoded. By calculating the ratio of the number of decoded frames to the total number of frames, the decoding progress can be intuitively displayed, for example, as a percentage. Furthermore, a visual interface can be used to provide real-time feedback on the decoding progress to system administrators or relevant operators, so that the progress of the decoding task can be kept informed. If decoding progress stagnates for an extended period, such as exceeding a preset time threshold, such as 10 minutes without change, the system automatically triggers an early warning mechanism and sends an alert to relevant personnel, alerting them to a potential decoding bottleneck. Furthermore, the system can monitor resource utilization during the decoding of the target video. In one embodiment, CPU resource utilization can be monitored through metrics such as CPU usage and the number of occupied cores. If CPU usage remains excessively high, such as exceeding 80%, for an extended period, it can lead to decreased decoding efficiency and even slow system response. In this case, the system can implement dynamic adjustment strategies, such as lowering the priority of the decoding task or pausing other non-critical tasks to free up some CPU resources to ensure smooth decoding. In another embodiment, parameters such as the memory size and memory usage growth rate of the decoding process can be monitored in real time. When memory usage reaches a preset threshold, such as 80% of physical memory, the system can initiate a memory optimization mechanism to clear invalid cached data and free up memory space. If memory usage remains high, non-essential processes can be terminated to ensure sufficient memory resources are available for the decoding task. In a third embodiment, metrics such as the frequency of disk read and write operations and the amount of data read and written can be monitored in real time during the decoding process.

[0137] It should be noted that when the system recognizes error information, it can record key data such as the time when the error occurred, specific error information, current decoding progress, and store the error information in the corresponding log file. At the same time, the system can process error information based on a preset error correction mechanism. In a specific embodiment, if the current error information is a decoding anomaly caused by incomplete data transmission due to network fluctuations, the system can attempt to automatically re-decode. Before re-decoding, the target video can first be checked for integrity to ensure that the target video is not further damaged. When re-decoding, the decoding parameters can be adjusted, such as reducing the decoding frame rate, reducing the resolution, etc., to improve the decoding success rate. In another specific embodiment, when encountering a video frame that cannot be decoded, frame interpolation technology can be used to generate approximate image frames through an algorithm based on the information of the previous and next frames, skip the error frame and continue decoding to ensure the continuity of video decoding. At the same time, the skipped video frames are marked for subsequent manual inspection or special processing.

[0138] As can be seen above, this embodiment determines the decoding progress of the target video in real time and monitors the target resource utilization during the decoding process in real time. This allows for the identification of errors based on the decoding progress and resource utilization of the target video and for error correction to be handled using a pre-set error correction mechanism. This ensures the data quality of key frames and improves the accuracy of key frame recognition, ensuring stable operation of the entire process from video decoding to key frame recognition, and preventing the accumulation of underlying issues that could impact the final recognition results.

[0139] Based on the above examples, it can be seen that this application can obtain a target video of a target housing collateral and directly identify the target keyframes corresponding to the target video using a target financial business template and a preset real-time target detection algorithm to obtain a target recognition result. Next, this example will explain in detail how to determine the target survey report for the target housing collateral.

[0140] In this embodiment, the system can utilize pre-collected external data sources such as e-commerce platform data, industry association data, professional valuation agency data, and home furnishing market research reports, combined with a comprehensive assessment model based on market comparison methods to determine the integrity score of the target recognition result and the valuation of the target housing collateral. Subsequently, a survey and analysis result for the target housing collateral can be generated based on the target recognition result, the target keyframe album, and a preset assessment strategy. Specifically, the system may include: first, generating a target survey summary for the target housing collateral based on the target recognition result; then, deduplicating the target recognition result based on the target keyframe album, and generating a value assessment result corresponding to the target housing collateral based on the deduplicated target recognition result; then, generating an integrity assessment result corresponding to the target housing collateral based on the deduplicated target recognition result and the target keyframe album, and performing a risk assessment on the target housing collateral based on the deduplicated target recognition result and preset risk assessment indicators to obtain a corresponding risk assessment result; and finally, generating the survey and analysis result for the target housing collateral based on the target survey summary, the value assessment result, the integrity assessment result, and the risk assessment result. Specifically, the system first generates a summary of the basic information of the survey and analysis results, including (1) report number; (2) collateral type, such as: housing decoration, commercial equipment, etc.; (3) customer name, such as: customer name, company name, etc.; (4) business type, such as: housing decoration loan, equipment mortgage loan, etc.; (5) survey date; (6) surveyor, such as: customer self-service, account manager, or others. In this embodiment, multiple key frames may be selected for the same room, which may lead to repeated identification. Therefore, the identified objects in the same target key frame album can be deduplicated to obtain the deduplicated target identification results. The system summarizes and sorts the objects identified in the survey according to the target key frame album to generate a value assessment result. The system generates a master table and multiple sub-tables, as shown in Table 1. The sub-table refers to the collection of a key frame album; the master table refers to the collection of multiple key frame albums.

[0141] Table 1 Value Assessment Table

[0142]

[0143] In addition, as shown in Table 2, the system will also conduct a completeness evaluation of the overall survey of the target housing mortgage and generate a completeness evaluation result corresponding to the target housing mortgage, with a completeness score of 100.

[0144] Table 2 Completeness Assessment Table

[0145]

[0146] The system can then perform a risk assessment on the target housing collateral based on the deduplicated target identification results and preset risk assessment indicators, obtaining a corresponding risk assessment result. The specific process may include: first, evaluating the valuation rationality of the target housing collateral based on the value assessment results to obtain a valuation rationality assessment result; and then, performing an abnormality risk assessment on the target housing collateral based on the deduplicated target identification results to obtain an abnormality risk assessment result; and then, generating a risk assessment result corresponding to the target housing collateral based on the valuation rationality assessment result, the integrity assessment result, the abnormality risk assessment result, and the overall risk assessment result corresponding to the target housing collateral. Specifically, the preset risk assessment indicators include valuation rationality, integrity, abnormality risk, and overall risk assessment. That is, the system can determine the valuation rationality assessment result of the target housing collateral based on the corresponding combined value assessment result of the target housing collateral, and perform an abnormality risk assessment on the target housing collateral. Finally, as shown in Table 3, the system can generate a risk assessment result corresponding to the target housing collateral based on the valuation rationality assessment result, the integrity assessment result, the abnormality risk assessment result, and the overall risk assessment result of the target housing collateral.

[0147] Table 3 Risk Assessment Table

[0148]

[0149] Finally, as shown in Table 4, the system will automatically generate the survey and analysis results of the target housing collateral by combining the above-mentioned value assessment results, integrity evaluation results, comprehensive risk assessment results, and the target survey summary of the target housing collateral.

[0150] Table 4 Survey Analysis Table

[0151]

[0152] It should be noted that generating a target survey report for a target housing mortgage based on a preset artificial intelligence technology, preset prompts, and survey and analysis results may include: first, structuring the initial survey data corresponding to the survey and analysis results based on a preset data structure, and determining target survey data corresponding to the initial survey data after data structuring according to a preset text format; then, generating the target survey report for the target housing mortgage based on a preset generative artificial intelligence technology, the preset prompts, and the target survey data. Specifically, the system first structures the initial survey data corresponding to the survey and analysis results, converting the initial survey data into key-value pairs (i.e., key-value pairs) or a table format. Then, a generative model is determined based on the preset generative artificial intelligence technology, and the structured initial survey data is converted into a text format understandable by the generative model to obtain the target survey data, for example, converting JSON data into a natural language description. The generative model is then used to generate the target survey report for the target housing mortgage, in combination with the target survey data and a preset prompt, i.e., a preset prompt. The preset prompt includes:

[0153] ① Basic information;

[0154] ② Collateral valuation assessment;

[0155] ③Integrity evaluation;

[0156] ④ Abnormal circumstances;

[0157] ⑤ Comprehensive risk assessment;

[0158] ⑥Conclusion and suggestions;

[0159] ⑦Review information.

[0160] The system can call the generative model through the API (Application Programming Interface) and input the preset prompt as a request parameter to the generative model. Finally, the content of the generated initial survey report needs to be further formatted to ensure that it meets the requirements of the financial institution. The system extracts key information from the text of the generated initial survey report and converts it into a table or paragraph form. For example, regular expressions or natural language processing tools can be used to convert the key information extracted from the text of the generated initial survey report into tabular data. After that, the extracted content is formatted according to a preset format, such as PDF or Word. Specifically, you can use Python's ReportLab to generate PDF or use the python-docx library to generate Word to complete the formatting and obtain the target survey report of the target housing mortgage. Some code references are shown below:

[0161] # Create a Word document

[0162] doc = Document()

[0163] doc.add_heading('Financial Collateral Survey and Evaluation Report', level=1)

[0164] # Add basic information

[0165] doc.add_heading('I. Basic Information', level=2)

[0166] doc.add_paragraph('Report number: 20250120-001')

[0167] doc.add_paragraph('Pledge type: housing renovation')

[0168] doc.add_paragraph('Customer Name: Zhang San')

[0169] doc.add_paragraph('Business Type: Housing Renovation Loan')

[0170] doc.add_paragraph('Survey date: 2025-01-20')

[0171] doc.add_paragraph('Surveyor: Automatically generated by the system')

[0172] # Add collateral valuation assessment

[0173] doc.add_heading('II. Collateral Valuation Assessment', level=2)

[0174] table = doc.add_table(rows=1, cols=5)

[0175] table.cell(0, 0).text = 'Project'

[0176] table.cell(0, 1).text = 'Description'

[0177] table.cell(0, 2).text = 'Quantity'

[0178] table.cell(0, 3).text = 'Status'

[0179] table.cell(0, 4).text = 'Estimated value (unit: yuan)'

[0180] # Add data rows

[0181] data = [

[0182] ["Wall decoration", "Completed", "100 square meters", "Completed", "5000"],

[0183] ["Floor decoration", "Completed", "80 square meters", "Completed", "8000"],

[0184] ["Indoor facilities", "5 lamps, 3 sets of sanitary ware", "Completed", "Completed", "3000"],

[0185] ["Total Valuation", "", "", "", "16000"] ]

[0187] for row_data in data:

[0188] row = table.add_row().cells

[0189] for i, cell_text in enumerate(row_data):

[0190] row[i].text = cell_text

[0191] # Save the document

[0192] doc.save('Survey and Assessment Report.docx')

[0193] As can be seen above, this embodiment can generate a survey and analysis result for the target housing collateral based on the target recognition results of the target key frames in the target video of the target housing collateral and a preset assessment strategy. It then generates a target survey report for the target housing collateral based on the preset artificial intelligence technology, preset prompt words, and the survey and analysis results. In this way, through generative artificial intelligence technology, this embodiment can output a target survey report for the target housing collateral, automating the survey of financial collateral, significantly reducing the human resource investment of financial institutions, improving business processing efficiency, and reducing financial risks.

[0194] Accordingly, see Figure 7 As shown, the embodiment of the present application further provides a financial collateral detection device based on computer vision, which may include:

[0195] The target key frame determination module 11 is used to obtain the collected target video of the target house mortgage and decode the target video to determine a number of target key frames of the target video;

[0196] a target key frame set determining module 12, configured to determine a set of key frames corresponding to a target house layout and preset room types of the target house collateral from the plurality of target key frames, to obtain a target key frame set, and to determine a target financial business template corresponding to the target house collateral;

[0197] A target recognition result determination module 13 is configured to perform target recognition on the target key frames in the target key frame collection based on the target financial business template and a preset real-time target detection algorithm to obtain a corresponding target recognition result;

[0198] The target survey report generation module 14 is used to generate a survey and analysis result of the target housing collateral based on the target recognition result, the target key frame album and the preset evaluation strategy, and to generate a target survey report of the target housing collateral based on the preset artificial intelligence technology, preset prompt words and the survey and analysis result.

[0199] As can be seen from the above, in this application, a target video of a target mortgaged property is first acquired and decoded to determine a number of target key frames of the target video. Then, a set of key frames corresponding to the target house layout and preset room types of the target mortgaged property is determined from the target key frames to obtain a target key frame album, and a target financial business template corresponding to the target mortgaged property is determined. Then, based on the target financial business template and a preset real-time target detection algorithm, target recognition is performed on the target key frames in the target key frame album to obtain a corresponding target recognition result. Finally, a survey and analysis result of the target mortgaged property is generated based on the target recognition result, the target key frame album, and a preset evaluation strategy, and a target survey report of the target mortgaged property is generated based on a preset artificial intelligence technology, preset prompt words, and the survey and analysis result. As can be seen from the above, in this application, a target video of the target mortgaged property is first acquired and decoded to determine a number of target key frames therein. Then, a set of key frames corresponding to the target key frames is determined based on the target house layout and preset room types of the target mortgaged property to obtain a target key frame album. Then, based on the target financial business template corresponding to the target housing mortgage and the preset real-time target detection algorithm, the target keyframe is identified to obtain the corresponding target recognition result. Finally, the survey and analysis results are generated based on the target recognition results, the target keyframe album and the preset evaluation strategy, and then the target survey report is generated using the preset artificial intelligence technology, the preset prompt words and the survey and analysis results. In this way, the present application can achieve a significant reduction in human resource costs and improve business processing efficiency by obtaining the target video of the target housing mortgage and directly performing target recognition on the target video; the target video of the target housing mortgage can be automatically recognized by using the target financial business template and the preset real-time target detection algorithm; the target survey report of the target housing mortgage can be automatically output based on the target recognition result by using the preset artificial intelligence technology and the preset prompt words. In this way, the present application can realize the automated survey of the target housing mortgage, improve the accuracy and objectivity of the value assessment, and reduce the operating costs.

[0200] In some specific embodiments, the computer vision-based financial collateral detection device may further include:

[0201] A reference house layout determining module is configured to select the unit structure of the target house mortgage based on the target selection instruction, so as to obtain a reference house layout corresponding to the unit structure;

[0202] a target house layout generating module, configured to generate the target house layout corresponding to the target house collateral by dragging the room elements in the reference house layout based on the orientation structure of each room in the target house collateral through a target drag command;

[0203] The target survey route generation module is used to generate a target survey route corresponding to the target house collateral based on the target house layout diagram, so as to capture the target video of the target house collateral based on the target survey route.

[0204] In some specific implementations, the target key frame determination module 11 may include:

[0205] An image data storage submodule is used to store the image data obtained after decoding the target video into a preset distributed storage system;

[0206] The target key frame determination submodule is used to determine a plurality of target key frames of the target video from the image data.

[0207] In some specific implementations, the image data storage submodule may include:

[0208] A decoding parameter configuration unit, configured to configure target decoding parameters corresponding to the target video; the target decoding parameters include a target frame rate and a target resolution;

[0209] a target video decoding unit, configured to decode the target video based on the target decoding parameters to obtain initial image data corresponding to the target video, and store the initial image data in a memory;

[0210] The image data storage unit is used to perform image preprocessing on the initial image data according to a preset image preprocessing method to obtain target image data, and store the target image data in the memory to the preset distributed storage system based on a preset data redundancy strategy.

[0211] In some specific embodiments, the computer vision-based financial collateral detection device may further include:

[0212] A total frame number determination module is used to determine the total frame number of the target video and set a corresponding target counter in the target decoding program;

[0213] Correspondingly, the target video decoding unit is specifically used to determine the number of decoded frames corresponding to the target video in real time, and update the value of the target counter in real time based on the number of decoded frames; there is a positive correlation between the number of decoded frames and the value of the target counter; the ratio of the value of the target counter to the total number of frames of the target video is determined in real time to monitor the decoding progress of the target video; the corresponding target resource utilization during the decoding process of the target video is monitored, and error information is identified based on the decoding progress and the target resource utilization; if the error information is identified, the error information is recorded in the target log file, and the error information is processed based on a preset error correction mechanism.

[0214] In some specific embodiments, the image data storage unit is specifically used to perform image preprocessing on the initial image data according to a preset image noise reduction method and a preset contrast enhancement method, and adjust the size of the preprocessed initial image data to a target size based on a preset image size condition to obtain the target image data, so as to determine several target key frames of the target video from the target image data.

[0215] In some specific embodiments, the image data storage unit is specifically used to perform a two-dimensional Fourier transform on each target image data, and determine the target high-frequency area in each transformed target image data; determine the energy proportion corresponding to each target high-frequency area, and determine the energy proportion below a preset energy proportion threshold as the target energy proportion; determine the target image data corresponding to the target energy proportion as a blurred image, and filter the blurred image from the target image data; determine the filtered target image data as a clear image, and determine several target key frames of the target video from the target image data based on the clear image.

[0216] In some specific implementations, the target key frame album determination module 12 may include:

[0217] a first target survey room determination unit, configured to determine a survey room in a target survey route corresponding to the target house collateral based on the target house layout and the preset room type, and match each target key frame with each survey room to determine a target survey room corresponding to each target key frame;

[0218] The first target key frame set determining unit is configured to determine the key frame set corresponding to each target key frame based on the target survey room corresponding to each target key frame.

[0219] In some specific implementations, the target key frame album determination module 12 may include:

[0220] A target three-dimensional space model construction unit, configured to construct a target three-dimensional space model corresponding to the target house collateral based on a preset real-time positioning and map construction technology and the target house layout diagram;

[0221] a target three-dimensional rectangular coordinate system establishing unit, configured to establish a corresponding target three-dimensional rectangular coordinate system in the target three-dimensional spatial model; the target three-dimensional rectangular coordinate system is a spatial coordinate system constructed with the front of the entrance of the target housing mortgage as the X-axis direction, the vertically upward direction as the positive Z-axis direction, and the Y-axis direction determined according to the right-hand rule;

[0222] a second target survey room determination unit, configured to determine target three-dimensional coordinate data corresponding to each target key frame based on the target three-dimensional rectangular coordinate system, and determine a target survey room corresponding to each target key frame based on the target three-dimensional coordinate data, the target house layout diagram, and the preset room type;

[0223] The second target key frame set determining unit is configured to determine the key frame set corresponding to each target key frame based on the target survey room corresponding to each target key frame.

[0224] In some specific implementations, the target recognition result determination module 13 may include:

[0225] a target key frame division submodule, configured to divide each target key frame in each target key frame album using the preset real-time target detection algorithm and based on a preset grid size condition to obtain grids corresponding to the target key frames; the grids are used to detect target objects in the target key frames;

[0226] an initial bounding box generation submodule, configured to generate a preset number of initial bounding boxes corresponding to the grid using the preset real-time object detection algorithm, and determine initial confidences corresponding to the initial bounding boxes and class probabilities corresponding to the initial bounding boxes;

[0227] The target recognition result determination submodule is used to perform target recognition on the target key frame based on the initial bounding box, the initial confidence corresponding to the initial bounding box, the category probability corresponding to the initial bounding box, the preset non-maximum suppression algorithm and the target financial business template to obtain the target recognition result.

[0228] In some specific implementations, the target recognition result determination submodule may include:

[0229] an object confidence determination unit, configured to determine an object confidence corresponding to the initial bounding box based on a product of an initial confidence corresponding to the initial bounding box and a category probability corresponding to the initial bounding box;

[0230] A target recognition result determination unit is used to perform target recognition on the target key frame based on the target confidence, the preset non-maximum suppression algorithm and the target financial business template to obtain the target recognition result.

[0231] In some specific embodiments, the target recognition result determination unit is specifically used to sort each of the initial bounding boxes based on the target confidence to obtain a first bounding box list and initialize a target result set; determine the initial bounding box with the highest target confidence from the first bounding box list, and determine the initial bounding box with the highest target confidence as the target bounding box; add the target bounding box to the target result set and remove the target bounding box from the first bounding box list; determine an initial intersection-and-union (IoU) between the target bounding box and each of the remaining initial bounding boxes in the first bounding box list, and determine a target IoU greater than a preset IoU threshold from the initial IoU; remove the initial bounding box corresponding to the target IoU from the first bounding box list, and jump to the step of determining the initial bounding box with the highest target confidence as the target bounding box, until the first bounding box list is empty, and generate the target recognition result corresponding to the target key frame based on the target financial business template and the target result set.

[0232] In some specific implementations, the target survey report generating module 14 may include:

[0233] A target survey summary generation submodule, configured to generate a target survey summary of the target housing collateral according to the target recognition result;

[0234] a value assessment result generating submodule, configured to perform deduplication processing on the target recognition result based on the target key frame album, and generate a value assessment result corresponding to the target housing collateral based on the deduplication target recognition result;

[0235] a risk assessment result determination submodule, configured to generate an integrity assessment result corresponding to the target housing collateral based on the target recognition result after deduplication and the target key frame album, and perform a risk assessment on the target housing collateral based on the target recognition result after deduplication and a preset risk assessment indicator to obtain a corresponding risk assessment result;

[0236] The survey and analysis result generating submodule is configured to generate the survey and analysis result of the target housing collateral based on the target survey summary, the value assessment result, the integrity assessment result, and the risk assessment result.

[0237] In some specific embodiments, the risk assessment result determination submodule may include:

[0238] an abnormal risk assessment result determining unit, configured to evaluate the valuation rationality of the target housing collateral based on the value assessment result to obtain a valuation rationality assessment result, and to perform an abnormal risk assessment on the target housing collateral based on the target identification result after deduplication to obtain an abnormal risk assessment result;

[0239] A risk assessment result determination unit is used to generate a risk assessment result corresponding to the target housing collateral based on the valuation rationality assessment result, the integrity assessment result, the abnormal risk assessment result and the overall risk assessment result corresponding to the target housing collateral.

[0240] In some specific implementations, the target survey report generating module 14 may include:

[0241] a target survey data determining unit, configured to perform data structuring processing on the initial survey data corresponding to the survey and analysis result based on a preset data structure, and determine target survey data corresponding to the initial survey data after the data structuring processing according to a preset text format;

[0242] A target survey report generating unit is used to generate the target survey report of the target house mortgage according to a preset generative artificial intelligence technology, the preset prompt words and the target survey data.

[0243] Furthermore, the embodiment of the present application also discloses an electronic device, Figure 8 This is a structural diagram of an electronic device 20 according to an exemplary embodiment. The content in this diagram should not be construed as limiting the scope of application of this application. The electronic device 20 may specifically include: at least one processor 21, at least one memory 22, a power supply 23, a communication interface 24, an input / output interface 25, and a communication bus 26. The memory 22 is used to store a computer program, which is loaded and executed by the processor 21 to implement the relevant steps of the computer vision-based financial collateral detection method disclosed in any of the aforementioned embodiments. Furthermore, the electronic device 20 in this embodiment may specifically be an electronic computer.

[0244] In this embodiment, the power supply 23 is used to provide operating voltage for each hardware device on the electronic device 20; the communication interface 24 can create a data transmission channel between the electronic device 20 and the external device. The communication protocol it follows is any communication protocol that can be applied to the technical solution of this application and is not specifically limited here; the input and output interface 25 is used to obtain external input data or output data to the outside world. Its specific interface type can be selected according to specific application needs and is not specifically limited here.

[0245] In addition, the memory 22, as a carrier for resource storage, can be a read-only memory, random access memory, disk or CD, etc. The resources stored thereon can include an operating system 221, a computer program 222, etc., and the storage method can be temporary storage or permanent storage.

[0246] The operating system 221 is used to manage and control the hardware devices on the electronic device 20 and the computer program 222. It can be Windows Server, NetWare, Unix, Linux, etc. In addition to including computer programs capable of implementing the computer vision-based financial collateral detection method performed by the electronic device 20 as disclosed in any of the aforementioned embodiments, the computer program 222 may further include computer programs capable of performing other specific tasks.

[0247] Furthermore, this application discloses a computer-readable storage medium for storing a computer program. When executed by a processor, the computer program implements the aforementioned computer vision-based financial collateral detection method. The specific steps of this method can be found in the aforementioned embodiments and will not be further elaborated here.

[0248] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from the other embodiments. Reference can be made to the descriptions of the identical or similar parts between the various embodiments. For the devices disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the descriptions are relatively simple, and the relevant parts can be referred to the descriptions of the methods.

[0249] Professionals may further appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the above description has generally described the components and steps of each example according to their functions. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professionals and technicians may use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0250] The steps of the methods or algorithms described in conjunction with the embodiments disclosed herein may be implemented directly using hardware, a software module executed by a processor, or a combination of the two. The software module may be placed in random access memory (RAM), internal memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, a hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art.

[0251] Finally, it should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or device comprising a series of elements includes not only those elements, but also other elements not explicitly listed, or elements inherent to such process, method, article, or device. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of additional identical elements in the process, method, article, or device comprising the element.

[0252] The above is a detailed introduction to the technical solution provided by the present application. Specific examples are used herein to illustrate the principles and implementation methods of the present application. The description of the above embodiments is only used to help understand the method of the present application and its core idea. At the same time, for those skilled in the art, according to the ideas of the present application, there may be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as a limitation on the present application.

Claims

1. A financial collateral detection method based on computer vision, characterized in that: include: Acquire the collected target video of the target house collateral, and decode the target video to determine a number of target key frames of the target video; Determining a set of key frames corresponding to a target house layout and preset room types of the target house collateral from the plurality of target key frames to obtain a target key frame album, and determining a target financial business template corresponding to the target house collateral; performing target recognition on the target key frames in the target key frame collection based on the target financial business template and a preset real-time target detection algorithm to obtain a corresponding target recognition result; A survey and analysis result of the target housing collateral is generated according to the target recognition result, the target key frame album and a preset evaluation strategy, and a target survey report of the target housing collateral is generated based on a preset artificial intelligence technology, preset prompt words and the survey and analysis result.

2. The computer vision-based financial collateral detection method according to claim 1, characterized in that: Before acquiring the collected target video of the target house collateral, the method further includes: Selecting the apartment structure of the target housing mortgage based on the target selection instruction to obtain a reference housing layout corresponding to the apartment structure; Using a target drag command, and based on the orientation structure of each room in the target house collateral, drag the room elements in the reference house layout diagram to generate the target house layout diagram corresponding to the target house collateral; A target survey route corresponding to the target house collateral is generated based on the target house layout diagram, so as to capture the target video of the target house collateral based on the target survey route.

3. The computer vision-based financial collateral detection method according to claim 1, characterized in that: The determining of a plurality of target key frames of the target video includes: storing the image data obtained after decoding the target video into a preset distributed storage system; A plurality of target key frames of the target video are determined from the image data.

4. The computer vision-based financial collateral detection method according to claim 3, characterized in that: The storing of the image data obtained after decoding the target video into a preset distributed storage system includes: Configuring target decoding parameters corresponding to the target video; the target decoding parameters include target frame rate and target resolution; Decoding the target video based on the target decoding parameters to obtain initial image data corresponding to the target video, and storing the initial image data in a memory; The initial image data is subjected to image preprocessing according to a preset image preprocessing method to obtain target image data, and the target image data in the memory is stored in the preset distributed storage system based on a preset data redundancy strategy.

5. The computer vision-based financial collateral detection method according to claim 4, characterized in that: Before decoding the target video based on the target decoding parameters, the method further includes: Determining the total number of frames of the target video and setting a corresponding target counter in the target decoding program; Correspondingly, the process of decoding the target video based on the target decoding parameters further includes: Determining the number of decoded frames corresponding to the target video in real time, and updating the value of the target counter in real time based on the number of decoded frames; there is a positive correlation between the number of decoded frames and the value of the target counter; determining in real time a ratio of a value of the target counter to a total number of frames of the target video to monitor a decoding progress of the target video; monitoring target resource utilization during decoding of the target video, and identifying error information based on the decoding progress and the target resource utilization; If the error information is identified, the error information is recorded in a target log file, and the error information is processed based on a preset error correction mechanism.

6. The computer vision-based financial collateral detection method according to claim 4, characterized in that: The performing image preprocessing on the initial image data according to a preset image preprocessing method to obtain target image data includes: The initial image data is preprocessed according to a preset image noise reduction method and a preset contrast enhancement method, and the size of the preprocessed initial image data is adjusted to a target size based on a preset image size condition to obtain the target image data, so as to determine several target key frames of the target video from the target image data.

7. The computer vision-based financial collateral detection method according to claim 6, characterized in that: Determining the plurality of target key frames of the target video from the target image data includes: Performing a two-dimensional Fourier transform on each of the target image data, and determining a target high-frequency area in each of the transformed target image data; Determining the energy proportion corresponding to each target high-frequency area, and determining the energy proportion below a preset energy proportion threshold as the target energy proportion; determining the target image data corresponding to the target energy proportion as a blurred image, and filtering the blurred image from the target image data; The filtered target image data is determined as a clear image, and a plurality of target key frames of the target video are determined from the target image data based on the clear image.

8. The computer vision-based financial collateral detection method according to claim 1, characterized in that: The step of determining a set of key frames corresponding to the target house layout and preset room types of the target house collateral from the plurality of target key frames includes: Determining the target survey room in the target survey route corresponding to the target house collateral based on the target house layout and the preset room type, and matching each target key frame with each survey room to determine the target survey room corresponding to each target key frame; The key frame set corresponding to each target key frame is determined based on the target survey room corresponding to each target key frame.

9. The computer vision-based financial collateral detection method according to claim 1, characterized in that: The step of determining a set of key frames corresponding to the target house layout and preset room types of the target house collateral from the plurality of target key frames includes: Constructing a target three-dimensional spatial model corresponding to the target house collateral based on the preset real-time positioning and map construction technology and the target house layout map; Establishing a corresponding target three-dimensional rectangular coordinate system in the target three-dimensional spatial model; the target three-dimensional rectangular coordinate system is a spatial coordinate system constructed with the front of the entrance of the target house mortgage as the X-axis direction, the vertically upward direction as the positive Z-axis direction, and the Y-axis direction determined according to the right-hand rule; Determining target three-dimensional coordinate data corresponding to each target key frame based on the target three-dimensional rectangular coordinate system, and determining a target survey room corresponding to each target key frame based on the target three-dimensional coordinate data, the target house layout diagram, and the preset room type; The key frame set corresponding to each target key frame is determined based on the target survey room corresponding to each target key frame.

10. The computer vision-based financial collateral detection method according to claim 1, characterized in that: The performing target recognition on the target key frame in the target key frame album based on the target financial business template and a preset real-time target detection algorithm to obtain a corresponding target recognition result includes: Using the preset real-time target detection algorithm, and dividing each target key frame in each target key frame album based on a preset grid size condition to obtain grids corresponding to the target key frames; the grids are used to detect target objects in the target key frames; Generating a preset number of initial bounding boxes corresponding to the grid using the preset real-time object detection algorithm, and determining initial confidences corresponding to the initial bounding boxes and class probabilities corresponding to the initial bounding boxes; Target recognition is performed on the target key frame according to the initial bounding box, the initial confidence corresponding to the initial bounding box, the category probability corresponding to the initial bounding box, a preset non-maximum suppression algorithm and the target financial business template to obtain the target recognition result.

11. The computer vision-based financial collateral detection method according to claim 10, characterized in that: The performing target recognition on the target key frame according to the initial bounding box, the initial confidence corresponding to the initial bounding box, the category probability corresponding to the initial bounding box, the preset non-maximum suppression algorithm, and the target financial business template to obtain the target recognition result includes: Determining an object confidence corresponding to the initial bounding box based on a product of an initial confidence corresponding to the initial bounding box and a category probability corresponding to the initial bounding box; Target recognition is performed on the target key frame based on the target confidence, the preset non-maximum suppression algorithm and the target financial business template to obtain the target recognition result.

12. The computer vision-based financial collateral detection method according to claim 11, characterized in that: The performing target recognition on the target key frame based on the target confidence, the preset non-maximum suppression algorithm, and the target financial business template to obtain the target recognition result includes: Sorting the initial bounding boxes based on the target confidence to obtain a first bounding box list, and initializing a target result set; Determining an initial bounding box with the highest target confidence from the first bounding box list, and determining the initial bounding box with the highest target confidence as the target bounding box; adding the target bounding box to the target result set and removing the target bounding box from the first bounding box list; determining an initial IoU between the target bounding box and each of the remaining initial bounding boxes in the first bounding box list, and determining a target IoU greater than a preset IoU threshold from the initial IoU; The initial bounding box corresponding to the target intersection-over-union ratio is removed from the first bounding box list, and the process proceeds to the step of determining the initial bounding box with the highest target confidence as the target bounding box until the first bounding box list is empty, and the target recognition result corresponding to the target key frame is generated based on the target financial business template and the target result set.

13. The computer vision-based financial collateral detection method according to claim 1, characterized in that: Generating the survey and analysis results of the target housing collateral according to the target recognition result, the target key frame album and a preset evaluation strategy includes: generating a target survey summary of the target housing collateral according to the target identification result; Deduplication processing is performed on the target recognition result based on the target key frame album, and a value assessment result corresponding to the target house collateral is generated based on the deduplication target recognition result; generating an integrity assessment result corresponding to the target housing collateral based on the target recognition result after deduplication and the target key frame album, and performing a risk assessment on the target housing collateral according to the target recognition result after deduplication and a preset risk assessment indicator to obtain a corresponding risk assessment result; The survey and analysis result of the target housing collateral is generated based on the target survey summary, the value assessment result, the integrity assessment result, and the risk assessment result.

14. The computer vision-based financial collateral detection method according to claim 13, characterized in that: The risk assessment of the target housing collateral is performed based on the target identification result after deduplication and the preset risk assessment index to obtain a corresponding risk assessment result, including: Evaluating the rationality of the valuation of the target housing collateral based on the value evaluation result to obtain a valuation rationality evaluation result, and performing an abnormal risk evaluation on the target housing collateral based on the target identification result after deduplication to obtain an abnormal risk evaluation result; A risk assessment result corresponding to the target housing collateral is generated based on the valuation rationality assessment result, the integrity assessment result, the abnormal risk assessment result, and the overall risk assessment result corresponding to the target housing collateral.

15. The financial collateral detection method based on computer vision according to any one of claims 1 to 14, characterized in that: The generating of the target survey report of the target housing collateral based on the preset artificial intelligence technology, the preset prompt words and the survey analysis results includes: Performing data structuring processing on the initial survey data corresponding to the survey and analysis result based on a preset data structure, and determining target survey data corresponding to the initial survey data after the data structuring processing according to a preset text format; The target survey report of the target house mortgage is generated according to the preset generative artificial intelligence technology, the preset prompt words and the target survey data.

16. A financial collateral detection device based on computer vision, characterized in that: include: a target key frame determination module, configured to obtain a captured target video of the target housing collateral and decode the target video to determine a number of target key frames of the target video; a target key frame album determining module, configured to determine, from the plurality of target key frames, a set of key frames corresponding to the target house layout and preset room types of the target house collateral, to obtain a target key frame album, and determine a target financial business template corresponding to the target house collateral; a target recognition result determination module, configured to perform target recognition on the target key frames in the target key frame collection based on the target financial business template and a preset real-time target detection algorithm to obtain a corresponding target recognition result; A target survey report generation module is used to generate a survey and analysis result of the target housing collateral based on the target recognition result, the target key frame album and a preset evaluation strategy, and to generate a target survey report of the target housing collateral based on a preset artificial intelligence technology, preset prompt words and the survey and analysis result.

17. An electronic device, characterized in that: The electronic device includes a processor and a memory; wherein the memory is used to store a computer program, and the computer program is loaded and executed by the processor to implement the financial collateral survey method based on computer vision according to any one of claims 1 to 15.

18. A computer-readable storage medium, characterized in that Used to store a computer program, which, when executed by a processor, implements the financial collateral survey method based on computer vision according to any one of claims 1 to 15.

Images