Abnormity measurement system for external facade of urban building

Abstract

The invention discloses an anomaly measurement system for an urban building facade, which comprises an image information acquisition module, a three-dimensional model generation module, an anomaly identification processing module, a comprehensive verification analysis module and an identification information output module, relates to the technical field of facade detection, and solves the problem that only for identification detection of a single anomaly type, the anomaly detection efficiency is high. According to the method, through standardization of an acquisition path and special scene processing, image information integrity is ensured, acquisition efficiency is improved, unmanned aerial vehicle GPS coordinates and total station control point coordinates are fused, a coordinate system is unified through seven-parameter conversion, absolute deviation of a model is reduced, and the accuracy of the system is improved. According to different characteristics of cracks, hollowing and peeling, double recognition logic of appearance characteristics and three-dimensional geometric characteristics is designed, the misjudgment rate is reduced, multiple types of anomalies are covered, a comprehensive verification module is introduced, and the deviation of an anomaly recognition result is reduced through 30% sample repeated measurement and a manual recheck mechanism.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of facade detection, in particular to an abnormal measurement system for the facade of urban buildings. BACKGROUND

[0002] With the city construction entering a new stage, the demand for urban renewal and reconstruction is huge, and more and more people's livelihood projects such as urban renewal and beautification are involved, such as repairing leaks in existing building roof waterproofing, exterior wall falling off, mold updating, building exterior window damage replacement, municipal road damage in community (park), etc., vegetation or greenery nursery damage repair, elevated road beautification, public parking lot planning, etc.

[0003] According to the disclosure number CN119131591A, a city infrastructure maintenance method and system based on image recognition, the invention obtains the city infrastructure image; the city infrastructure image is identified by image recognition, and the defect image containing the defect is screened out; the defect image is extracted, and the defect feature is obtained; based on the defect feature, the risk level of the city infrastructure is determined; based on the risk level of the city infrastructure, the city infrastructure is arranged for maintenance. Solve the problems of low efficiency, insufficient accuracy and unreasonable resource allocation in the current city infrastructure maintenance, and promote the city management to a more intelligent, efficient and sustainable direction.

[0004] However, part of the existing abnormal measurement system only analyzes the facade, uses manual handheld tools for on-site measurement of low-rise buildings, estimates the abnormal range by visual inspection, needs to build scaffolding or use a basket for high-altitude work for medium and high-rise buildings, and calculates the overall data after sampling measurement. Based on image recognition technology, multiple focuses on a single abnormal type, and there are problems such as insufficient image acquisition accuracy, large deviation between three-dimensional modeling and actual size, abnormal identification is easily disturbed by light / shading, etc., which is difficult to meet the comprehensive needs of precision, efficiency and safety. SUMMARY

[0005] In view of the deficiencies of the prior art, the present application provides an abnormal measurement system for the facade of urban buildings, which solves the problem of only recognizing and detecting a single abnormal type, lacks targeted solutions, and has false positives in abnormal identification.

[0006] To achieve the above purpose, the present application realizes the following technical scheme: an abnormal measurement system for the facade of urban buildings, comprising: An image information acquisition module is used to acquire image information of the facade of urban buildings and transmit the image information to the three-dimensional model generation module; A three-dimensional model generation module is used to establish a corresponding three-dimensional real scene model according to the received image information, and transmit the three-dimensional real scene model to the abnormal identification processing module; An abnormality identification processing module is configured to identify cracks, detect hollowing, and identify peeling based on the three-dimensional real scene model, generate corresponding abnormality information, and transmit the abnormality information to an identification information output module. The identification information output module is configured to display the received abnormality information to a manager.

[0007] As a further scheme of the present application, the specific acquisition mode of the image information acquisition module includes: Quantitative indexes of the image information are determined, including spatial resolution, geometric accuracy, angle deviation, and integrity requirement, and corresponding acquisition paths and methods are designed based on building floor height.

[0008] As a further scheme of the present application, the specific mode of the corresponding acquisition paths and methods based on building floor height design includes: For low-rise buildings with a height of less than 6 floors, a ground handheld high-definition camera and a total station instrument are used for auxiliary positioning, and the image information is acquired along the circumference of the building in a clockwise direction; For medium and high-rise buildings with a height of 7 to 20 floors, a UAV is used to acquire image information in a Z-shaped flight path, and the GPS coordinates of each frame of image information are recorded synchronously; For super high-rise buildings with a height of more than 20 floors, a UAV and a magnetic wall climbing robot are used to acquire image information.

[0009] As a further scheme of the present application, the specific mode of the three-dimensional model generation module for establishing a three-dimensional real scene model includes: The image information is acquired, and the magnification blur and overlap rate are analyzed, if any one of them is abnormal, it indicates that the acquired image information is abnormal, and the corresponding abnormal data is obtained, and it is determined whether the image information can be repaired according to the abnormal data, if the image information can be repaired, the image information is repaired, if the image information cannot be repaired, the image information is reacquired, if both the magnification blur and the overlap rate are normal, it indicates that the acquired image information is normal; Then, the repaired image is acquired, the UAV GPS and total station instrument data are imported into the system BIM module, a unified plane coordinate system is established, an external facade three-dimensional frame model is generated, the repaired image is pasted to the three-dimensional frame through a system automatic alignment algorithm, a multi-angle image stitching technology is used for regions such as curved surfaces and corners, and a three-dimensional real scene model is generated.

[0010] As a further scheme of the present application, the crack identification mode of the abnormality identification processing module includes: Based on the repaired image, linear features are extracted through Canny edge detection, and the depth information of the three-dimensional real scene model is used to exclude shadow misjudgment, if there is a crack, crack information including the length, width, and position of the crack is generated and transmitted to the identification information output module, if there is no crack, a hollowing detection signal is generated and analyzed.

[0011] As a further scheme of the present application, the specific way of generating the hollowing detection signal and analyzing it is: The unmanned aerial vehicle is controlled to carry a high-precision infrared thermal imager to collect infrared thermal images of the outer facade according to a layered flight path, visible light images and LiDAR point cloud data are synchronously collected, the infrared thermal images and the visible light images are aligned through a SIFT algorithm based on the LiDAR point cloud, and temperature-texture-three-dimensional coordinates are associated; The temperature distribution of a normal area is analyzed, a hollowing temperature threshold is set, a low-temperature area is extracted through an Otsu method and a hollowing suspected area is marked, the surface flatness of the suspected area is calculated, and if the surface is flat, the suspected area is determined to be hollowing, and hollowing information including a position, an area and a maximum depth is generated; otherwise, a spalling identification signal is generated and analyzed.

[0012] As a further scheme of the present application, the specific way of generating the spalling identification signal and analyzing it is: The visible light images are converted through HSV conversion, the color difference between a pixel and a surrounding 5*5 area is calculated, a color abnormal area is screened, the texture is analyzed through a gray level co-occurrence matrix, a texture fracture area with an energy value drop of more than 50% is screened, and the texture fracture area is marked as a suspected area; The three-dimensional point cloud of the suspected area is extracted, the height difference with the surrounding is calculated, the edge curvature is analyzed, and if the conditions are met, the suspected area is determined to be spalling, and spalling information including a position, an area and a height difference is generated.

[0013] As a further scheme of the present application, a comprehensive verification and analysis module is further included, which is configured to receive abnormal information output by the abnormality identification processing module, randomly extract 30% samples for repeated measurement through a three-dimensional model, calculate the proportion of times of abnormality, mark the abnormality if the proportion exceeds a set threshold, otherwise, perform manual review, generate a review result and transmit the review result to the identification information output module.

[0014] The present application provides an abnormality measurement system for the outer facade of a city building. The present application ensures complete image information, improves collection efficiency, unifies the coordinate system through seven-parameter conversion by fusing the GPS coordinates of the unmanned aerial vehicle and the control point coordinates of the total station, reduces the absolute deviation of the model, designs a dual identification logic of appearance features and three-dimensional geometric features according to different characteristics of cracks, hollowing and spalling, reduces the misjudgment rate, covers multiple types of abnormalities, solves the limitations of single abnormality identification in the prior art, introduces a comprehensive verification module, reduces the deviation of abnormality identification results through 30% sample repeated measurement and manual review mechanism. BRIEF DESCRIPTION OF DRAWINGS

[0015] Figure 1 The present application is a system principle block diagram. DETAILED DESCRIPTION

[0016] With reference to the drawings of the embodiments of the present application, the technical solutions in the embodiments of the present application will be clearly and completely described, obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work belong to the scope of protection of the present application.

[0017] Embodiment one

[0018] Please refer to Figure 1 The present application provides an abnormal measurement system for the outer facade of urban buildings, comprising: an image information acquisition module, a three-dimensional model generation module, an abnormality identification processing module, a comprehensive verification and analysis module, and an identification information output module, combined with the accompanying Figure 1 It can be known that the information between the above functional modules is unidirectional transmission.

[0019] The image information acquisition module is used to acquire image information of urban buildings, and simultaneously transmit the acquired image information to the three-dimensional model generation module, and the specific acquisition method is as follows: The quantitative indicators of the image are determined, specifically including spatial resolution, geometric accuracy, angle deviation, and integrity requirements, and different acquisition paths and methods are involved based on different building heights, for low-rise buildings, specifically ≤6 floors, a ground handheld high-definition camera + total station auxiliary positioning is adopted, the acquisition path is clockwise around the building perimeter, and every 3m, 1 normal facade image is shot, and 45° oblique images are added at the corners, each 1; For medium and high-rise buildings, specifically 7-20 floors: an unmanned aerial vehicle is adopted, wherein the unmanned aerial vehicle is equipped with an RTK module + 50 million pixel camera, and the image is collected according to a “Z-shaped” flight path, the flight height is higher than the top of the building by 5-10m, the horizontal shooting interval is ≤2m, and specifically, in order to ensure that the image overlap rate is ≥70%, a longitudinal flight path is set every 3 floors, and the GPS coordinates of each frame of image are recorded synchronously; For super high-rise buildings, specifically >20 floors: unmanned aerial vehicles and wall climbing robots are fused, and magnetic attraction collection is adopted, the unmanned aerial vehicle is responsible for the outer facade above 10 floors and the top structure, and the wall climbing robot is responsible for close-range shooting for the complex area at the bottom, close-range shooting means a distance of 0.5-1m from the wall, and 1 image is shot every 0.5m along the vertical direction; Special scene supplement: for the glass curtain wall reflection area, a polarized lens is used for shooting; for the backlight area, a fill light is turned on to ensure uniform brightness; Each image is automatically associated with the collection time, equipment number, GPS coordinates, building building + floor + facade orientation label.

[0020] A three-dimensional model generation module is configured to establish a corresponding three-dimensional real scene model according to the acquired image information, and transmit the generated three-dimensional real scene model to the anomaly identification processing module, and the specific establishment manner is as follows: The image information is acquired, and the magnification blur condition and the overlap rate are analyzed. The magnification blur condition indicates that the image is magnified to 100%, and it is judged whether there is blur. The overlap rate is calculated by professional software, and compared with the corresponding judgment threshold value. The specific value of the judgment threshold value is set by the operator. If there is any one group of anomalies, it indicates that the collected image information is abnormal. At the same time, the corresponding abnormal data is acquired, and whether it can be repaired is judged according to the abnormal data. If it can be repaired, the image information is repaired, and the repair here specifically includes missing area repair and edge repair. If it cannot be repaired, it needs to be reacquired. If both do not exist, it indicates that the collected image information is normal. Then, the repaired image is acquired, the unmanned aerial vehicle GPS coordinates and the total station instrument measured control point coordinates are imported into the system BIM module, a unified plane coordinate system is established, the axis size of the architectural design drawing is used to generate an outer facade three-dimensional frame model in the BIM software, the repaired image is "stuck" to the three-dimensional frame through the system automatic alignment algorithm, the multi-angle image splicing technology is used for the curved surface and corner regions to ensure that the model surface texture is not stretched and misaligned after pasting, and a three-dimensional real scene model is generated.

[0021] The image is magnified to 100% display, and the degree of blur is quantified through a clarity evaluation algorithm such as a Tenengrad gradient function. The gradient value of the edge region in the image is calculated. If the average gradient is less than 20, the threshold value can be adjusted by the operator according to the scene, such as setting the threshold value of the complex texture region to 30, and determining it as a "blurred image"; Professional photogrammetry software such as Pix4D and ContextCapture is used to match and analyze the image sequence, calculate the overlap area ratio of adjacent images, and the heading overlap rate needs to be greater than or equal to 70%, and the lateral overlap rate needs to be greater than or equal to 60%. The operator can adjust according to the building height, such as setting the lateral overlap rate of super high-rise buildings to be greater than or equal to 80%. If the image has blur or the overlap rate does not meet the standard, which means any one abnormality, it is marked as "abnormal data" and repaired first. The repair content includes missing area repair and edge repair. The specific missing area repair method is as follows: For information missing areas such as tree shelter, shooting blind area, etc. The specific expression area ≤ 15%, based on the adjacent image of the fragment information, through the multi-view stereo matching algorithm to extract the geometric contour, such as wall lines, component boundaries, guide PatchMatch algorithm to fill the space structure of the missing area; Then call DeepFillv2 model, combined with the surrounding texture features, such as tile arrangement rules, paint color gradient, generate reasonable texture, ensure that the repaired area and the original image texture consistency ≥ 95%.

[0022] For edge repair, the image edge blur caused by shooting angle deviation, such as wall corner, decorative line edge, use EdgeConnect model to strengthen the edge features - through Canny operator to extract the edge contour, based on the contour continuity to generate complete edge lines, the edge definition after repair is improved to 0.5mm level of detail, such as tile seam width.

[0023] If the image is seriously blurred, specifically represented by the average gradient <10, or the missing area is too large, specifically represented by the area >15%, it is determined that it cannot be repaired, and the re-sampling mechanism needs to be triggered to inform the on-site personnel to shoot the image according to the original collection path; The GPS coordinates recorded synchronously when the unmanned aerial vehicle collects the image, with an accuracy of ±0.5m, and the control point coordinates (accuracy ≤1mm) measured by the ground total station are imported into the BIM module of the system, and the coordinate calibration is carried out through the seven-parameter coordinate conversion model based on at least three uniformly distributed control points; Based on the axis size of the architectural design drawing, such as column spacing, floor height, wall verticality, generate the three-dimensional structure frame of the facade in the BIM mapping software, such as Revit - the frame needs to include the geometric contour of key components such as columns, beams, walls, doors and windows, and the deviation of each component size from the design drawing is ≤5mm, such as the design window sill height is 90cm, the error in the frame model is ≤0.5cm; Use SIFT feature point matching algorithm to automatically align the repaired image with the three-dimensional frame - extract the feature points in the image that have unique characteristics, and the feature points include tile corner points, decorative line inflection points, and match the three-dimensional coordinates of the corresponding positions in the frame model. For areas that are difficult to completely cover by a single view of the unmanned aerial vehicle, such as curved curtain wall, wall corner groove, arc decoration, use multi-angle image stitching technology to obtain a three-dimensional real scene model, and the specific stitching method is as follows: Select 3-5 images taken at different angles, the image requirements include shooting angle interval ≤15°, overlap rate ≥80%, optimize the stitching parameters through BundleAdjustment, which is represented by bundle adjustment method, and process the color difference at the joint after stitching through texture fusion algorithm to ensure that the color transition of the spliced area is natural, and the color difference ΔE ≤5.

[0024] An abnormality identification processing module is configured to perform abnormality identification processing based on the obtained three-dimensional real scene model, and the abnormality identification processing specifically includes crack identification, hollow detection, and spalling identification. The specific identification methods are as follows: Linear features corresponding to the image information are extracted by Canny edge detection, and the image information here is the repaired image. The depth information of the three-dimensional model is used to exclude shadow misjudgment. For example, a crack is a continuous depression in three-dimensional space, while a shadow has no depth change. It is determined whether a crack abnormality exists. If a crack abnormality exists, crack information is generated, and the crack information includes crack length, crack width, and crack position. The crack information is transmitted to the identification information output module. If a crack abnormality does not exist, a hollow detection signal is generated. The generated hollow detection signal is analyzed. A high-precision infrared thermal imager is carried by the unmanned aerial vehicle. The resolution is ≥640×512 pixels, and the temperature measurement accuracy is ±0.5°C. The outer facade is photographed according to the layered flight route. The layered flight route specifically represents the photographing method corresponding to different floor heights of the building object. A high-definition visible light camera LiDAR laser radar is simultaneously carried by the unmanned aerial vehicle. Visible light images and three-dimensional point cloud data are collected by using a hybrid stereo photogrammetry method. The LiDAR point cloud model is used as a reference. The infrared thermal image and the visible light image are aligned by using a feature point matching algorithm, such as SIFT. The coordinate deviation of the same physical position in the two types of images is ensured to be ≤1 pixel, about 0.5 mm. The association of temperature information-texture information-three-dimensional coordinates is realized. The temperature distribution of the normal area is then analyzed. For example, the average temperature of the wall surface is 25°C, and the standard deviation is ±1°C. The temperature threshold for hollow identification is set. A threshold segmentation algorithm, such as the Otsu method, is used to extract the low-temperature area from the infrared thermal image. A pixel-level mask is output and the hollow suspected area is marked. The coordinates of the three-dimensional point cloud are combined to map the suspected area to the three-dimensional space and record its position. The surface flatness of the hollow suspected area is then calculated. The surface flatness is compared with the flatness threshold. If the surface flatness is greater than the flatness threshold, it indicates that the hollow suspected area is not hollow, and there is a misjudgment. A spalling identification signal is generated. If the surface flatness is less than the flatness threshold, it indicates that the hollow suspected area has a hollow. Hollow information is generated, and the hollow information includes the hollow position, area, maximum depth, and depth. The hollow information is transmitted to the identification information output module. The generated spalling identification signal is analyzed. The visible light image is obtained. The color difference between the pixel points and the surrounding intact area is calculated by HSV color space analysis. The color mutation area is screened out. The specific screening method is as follows. The high-definition visible light image is converted to HSV color space. The average color difference between each pixel and the surrounding 5×5 pixel area is calculated. If the color difference + + ≥80, and the continuous area is >0.005 m2 For example, about 2cmx2cm, filter point stains, wherein is a difference in hue, is a difference in saturation, is a difference in lightness, marked as a color anomaly area, while analyzing the continuity of the image texture with the gray level co-occurrence matrix GLCM, and calculating the energy value of the texture, if the energy value is lower than the comparison value, the corresponding color anomaly area is marked as a suspected area; Extract the three-dimensional point cloud data of the suspected area, calculate the average height difference with the intact area within a range of 10cm, if there is no height difference, there is no spalling, otherwise if there is a height difference, it indicates that there is spalling, at the same time, the edge of the suspected area is segmented and fitted, the curvature change of each segment is calculated, if more than 80% of the edge segment curvature change rate> 50°, it is determined as a spalling edge, and spalling information is generated, which is then transmitted to the identification information output module.

[0025] The identification information output module is used to display the obtained crack information, hollow drum information and spalling information to the corresponding management personnel.

[0026] Example two

[0027] As example two of the present application, on the basis of example one, and the difference from example one is as follows: The anomaly identification processing module transmits the generated crack information, hollow drum information and spalling information to the comprehensive verification analysis module, and performs verification analysis through the comprehensive verification analysis module, and the specific verification analysis method is as follows: After obtaining the abnormal information, 30% of the samples are randomly extracted, and repeated measurement is performed through the three-dimensional model, and the repeated measurement results are judged, the abnormal repeated measurement results are obtained, and the abnormal proportion times are calculated, the abnormal proportion times are compared with the proportion threshold value, and the specific value of the proportion threshold value is set by the operator, if the abnormal proportion times are greater than the proportion threshold value, it indicates that there is an abnormal identification, otherwise if the abnormal proportion times are less than the proportion threshold value, manual review is performed, the corresponding review result is generated, and the review result is transmitted to the identification information output module.

[0028] The identification information output module is used to display the obtained review result to the corresponding management personnel.

[0029] Example three

[0030] As example three of the present application, the focus is to combine the implementation process of example one and example two.

[0031] The data in the above formula are calculated by taking their values, and are not calculated by taking the units of the parameters. In addition, the contents not described in detail in the specification are all prior art known to those skilled in the art.

[0032] The above examples are only used to illustrate the technical method of the present application and not limit the present application. Although the present application is described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical method of the present application can be modified or replaced equivalently without departing from the spirit and scope of the technical method of the present application.

Claims

1. The abnormality measurement system of urban building facades is characterized by: include: An image information acquisition module is used to acquire image information of the facades of urban buildings and transmit the image information to the 3D model generation module; A three-dimensional model generation module is used to establish a corresponding three-dimensional real scene model according to the received image information, and transmit the three-dimensional real scene model to the abnormality recognition processing module; The abnormality identification and processing module is used to identify cracks, hollowing and spalling based on the 3D real-life model, generate corresponding abnormality information and transmit it to the identification information output module; The identification information output module is used to display the received abnormal information to the management personnel.

2. The abnormality measurement system for urban building facades according to claim 1 is characterized in that: The specific acquisition method of the image information acquisition module includes: Clarify the quantitative indicators of image information, including spatial resolution, geometric accuracy, angular deviation and integrity requirements, and design corresponding acquisition paths and methods based on building height.

3. The abnormality measurement system for urban building facades according to claim 2, characterized in that: The specific method of the acquisition path and method based on the building height design is as follows: For low-rise buildings with 6 floors or less, a ground-based handheld high-definition camera and a total station are used to assist in positioning, and data is collected in a clockwise direction along the perimeter of the building. For medium- and high-rise buildings with 7 to 20 floors, the drone collects images in a zigzag pattern and simultaneously records the GPS coordinates of each frame. For super high-rise buildings with more than 20 floors, drones and magnetic wall-climbing robots are integrated for data collection.

4. The abnormality measurement system for urban building facades according to claim 1, characterized in that: The specific method of the three-dimensional model generation module to establish the three-dimensional real scene model includes: Acquire image information and analyze its magnification blur and overlap rate. If any of the two is abnormal, it indicates that the acquired image information is abnormal. At the same time, obtain the corresponding abnormal data and determine whether it can be repaired based on the abnormal data. If it can be repaired, the image information is repaired. If it cannot be repaired, it needs to be re-acquired. If both are normal, it indicates that the acquired image information is normal. Then, the repaired image is obtained, and the drone GPS and total station data are imported into the system's BIM module to establish a unified plane coordinate system and generate a three-dimensional framework model of the facade. Through the system's automatic alignment algorithm, the repaired image is fitted into the three-dimensional framework, and multi-angle image stitching technology is used for curved surfaces, corners and other areas to generate a three-dimensional real-scene model.

5. The abnormality measurement system for urban building facades according to claim 1 is characterized in that: The crack identification method of the abnormality identification processing module includes: Based on the repaired image, linear features are extracted through Canny edge detection, and shadow misjudgment is eliminated by combining the depth information of the 3D real-scene model. If cracks are present, crack information including crack length, width, and position is generated and transmitted to the identification information output module. If no cracks are present, a hollow detection signal is generated and analyzed.

6. The abnormality measurement system for urban building facades according to claim 5, characterized in that: The specific method of generating the hollow drum detection signal and analyzing it is as follows: Control the drone equipped with a high-precision infrared thermal imager to collect infrared thermal images of the facade along a layered route, simultaneously collecting visible light images and LiDAR point cloud data. Using the LiDAR point cloud as a reference, the infrared thermal image and visible light image are aligned using the SIFT algorithm to correlate temperature, texture, and 3D coordinates. Analyze the temperature distribution in the normal area, set the hollowing temperature threshold, extract the low-temperature area using the Otsu method and mark the suspected hollowing area, calculate the surface flatness of the suspected area, and if it is flat, it is judged as a hollow, and generate hollowing information including position, area, and maximum depth; otherwise, generate a spalling recognition signal and analyze it.

7. The abnormality measurement system for urban building facades according to claim 6, characterized in that: The specific method of generating and analyzing the peeling identification signal is as follows: Perform HSV conversion on the visible light image, calculate the color difference between the pixel and the surrounding 5×5 area, screen out color abnormal areas, analyze the texture through the gray level co-occurrence matrix, screen out texture fracture areas with energy values ​​dropping by more than 50%, and mark them as suspicious areas; Extract the three-dimensional point cloud of the suspected area, calculate the height difference with the surrounding area, analyze the edge curvature, and determine it as spalling if the conditions are met. Generate spalling information including position, area, and height difference.

8. The abnormality measurement system for urban building facades according to claim 1, characterized in that: It also includes a comprehensive verification and analysis module, which is used to receive the abnormal information output by the abnormality identification and processing module, randomly select 30% of the samples and repeat the measurement through the three-dimensional model, calculate the number of abnormal proportions, and if the proportion exceeds the set threshold, it is marked as an identified abnormality. Otherwise, manual review is performed to generate the review results and transmit them to the identification information output module.

Images