A real scene three-dimensional surveying and mapping system based on big data

Abstract

The application discloses a kind of real scene three-dimensional surveying and mapping systems based on big data, and the application relates to three-dimensional surveying and mapping technical field, including acquisition module, for video data in the range area of camera acquisition equipment collection range area, the video data includes range area north side perspective data and range area upper perspective data, range area west side perspective data;The present application can capture the complete spatial information of the real object in the scene through the cooperative collection of north side, upper, west side multi-view video data, overcome the problem of single view shielding, through the processing of big data identification model, realize the rapid identification of each type of real object in the scene and the automatic extraction of three-dimensional characteristic parameters, improve the identification precision, reduce manual intervention, adopt dynamic three-dimensional model splicing and calibration method, according to actual identification result, the position, posture and size of object model are adjusted in real time, ensure that the generated three-dimensional map structure is accurate, boundary continuous, improve the technical level of automatic three-dimensional mapping.

Description

Technical Field

[0001] This invention relates to the field of 3D surveying and mapping technology, specifically to a real-scene 3D surveying and mapping system based on big data. Background Technology

[0002] Surveying and mapping technology is a key scientific and engineering means for acquiring, processing, and representing spatial information about the Earth's surface. Its core purpose is to provide high-precision spatial data support for numerous fields such as land resource management, urban and rural planning, engineering construction, and environmental monitoring. Traditional surveying and mapping technologies mainly rely on manual ground surveying, total stations, and GPS positioning. Although they can obtain relatively accurate two-dimensional or a small amount of three-dimensional spatial data, the operation cycle is long and the coverage is limited. Real-scene surveying and mapping technology refers to the use of multi-source sensors (such as cameras) to collect data on real-world scenes and combine computer vision, 3D modeling, and big data analysis to achieve three-dimensional digital reconstruction and representation of the scene's spatial structure, land feature attributes, and spatial relationships. Compared with traditional surveying and mapping methods, real-scene surveying and mapping technology can obtain more comprehensive, intuitive, and high-precision spatial information and is widely used in fields such as smart city construction, land spatial planning, digital twins, traffic management, and cultural heritage protection. Existing surveying and mapping systems rely on a single perspective or fixed process to model the collected data. They lack a cross-comparison mechanism for the arrangement of real objects from multiple perspectives. As a result, there is a significant lag in the detection of abnormal areas with mismatched categories or incorrect arrangement order. Manual review is required to locate and correct these abnormal areas. To address the aforementioned technical shortcomings, a solution is proposed. Summary of the Invention

[0003] To address the shortcomings of existing technologies, this invention provides a real-scene 3D mapping system based on big data.

[0004] To achieve the above objectives, the present invention provides the following technical solution: a real-scene 3D mapping system based on big data, comprising: The acquisition module is used by the camera acquisition device to acquire video data of the area. The video data includes the northern view data of the area, the upper view data of the area, and the western view data of the area. The recognition module uses a big data recognition model to identify real objects in the scene from the perspective data above the range area. The big data recognition model identifies the perspective data above the range area and marks the types of real objects in the scene. It then selects three-dimensional object models and combines them into a map of the range area. The height recognition module uses a big data recognition model to identify the height values ​​of real objects in the scene from the northern view of the range area, and modifies the 3D object models in the range area map according to the range area map. The verification module outputs a sequence of object models arranged from left to right based on the map of the range area. It uses a big data recognition model to identify the real object arrangement verification sequence of scene real objects in the western view data of the range area. It verifies the object model arrangement sequence based on the real object arrangement verification sequence to obtain the area to be verified. The area to be reviewed undergoes secondary data collection and secondary identification.

[0005] The process by which the big data recognition model identifies real objects in a scene from a perspective above a given area includes: S21. Input the collected view data above the range area into the big data recognition model. The model adopts a deep convolutional neural network structure to perform image normalization, edge enhancement and convolutional feature extraction on the input image to obtain an image feature map. S22. Using a semantic segmentation algorithm, candidate regions are generated and pixel-level divisions are performed on the image feature map, and multiple candidate boxes are output. The candidate boxes correspond to potential scene object regions in the image. S23. Apply a multi-type classification network to the candidate box regions respectively, identify and output the corresponding scene object type labels, wherein the scene objects include buildings, lanes, vegetation and rivers.

[0006] The steps to combine real objects in a scene into a map of a defined area include: S31. Based on the scene object type label, call the corresponding standard 3D object model template in the 3D model library, where the 3D model library is a collection of model templates; S32. Extract geographic planar coordinates (X,Y) from the view data above the range area; S33. Based on the pixel size of the real object in the view data above the range area, scale the selected 3D object model proportionally; and perform pose rotation on the model by referring to the direction in the view data above the range area. S34. Overlay the 3D object model onto the map model according to the geographic plane coordinates (X,Y), and process the boundaries between the 3D object models through an automatic stitching algorithm; S35. Perform texture mapping on the map model to obtain a 3D map of the area.

[0007] Extracting geographic plane coordinates (X,Y) from the view data above the range area includes: extracting the coordinates (u,v) of the two-dimensional projection center point of the scene object and the bounding box size information from the view data above the range area, and mapping the two-dimensional projection center point coordinates to geographic plane coordinates (X,Y).

[0008] Where u0,v0 are the image origin, (X0,Y0) are the geographic coordinate origin, Sx and Sy are the resolution of the distance from the pixel to the plane, and u,v are the coordinates of the center point of the two-dimensional projection.

[0009] The process of using a big data recognition model to identify the height values ​​of real objects in the scene from the northern viewpoint of the specified area includes: S41. Perform image preprocessing on the northern view data of the range area. The image preprocessing includes distortion correction, edge enhancement and illumination equalization. S42. Input the northern view data of the range into the big data recognition model, and use semantic segmentation and object detection algorithms to extract the target bounding box; S43. The height value of the target bounding box is calculated by applying a monocular image depth regression model set to the target bounding box. S44. Match the target bounding box with the scene object; S45. Adjust the height of the corresponding 3D object model according to the height value of the target bounding box.

[0010] The calculation is performed using a monocular image depth regression model set, including: projecting the center pixels of the upper and lower edges of the target bounding box onto the camera coordinate system based on the camera intrinsic parameter matrix, obtaining the vertical coordinates of the two points through calculation, and using the absolute value of the difference between the vertical coordinates of the two points as the height value of the target bounding box.

[0011] Output a sequence of object models arranged from left to right based on the area map, including: S501. Extract the coordinates of the two-dimensional projection center point of each three-dimensional object model from the three-dimensional map of the range area; S502. Sort the center point coordinates in ascending order of horizontal coordinate values ​​to form object model arrangement sequence L1. S503. Input the western view image of the range area into the big data recognition model. The big data recognition model performs target detection on the image and outputs the object arrangement verification sequence L2 from the western view.

[0012] The object arrangement verification sequence for identifying real objects in the scene from the western perspective of the defined area includes: S511. Compare sequence L1 and sequence L2 to determine the consistency of their positional order; S512. If there is a type mismatch or order mismatch in the comparison, the sequence L1 is marked as an abnormal sequence. S513. Define the spatial region corresponding to the abnormal sequence as the region to be reviewed.

[0013] The secondary acquisition and secondary recognition of the area to be reviewed includes: generating a secondary acquisition task based on the area to be reviewed; instructing the camera acquisition device to acquire image data of the area to be reviewed; using a big data recognition model to identify scene objects in the view data above the range area; selecting a 3D object model; updating the 3D object model to the range area map; outputting a sequence of object models arranged from left to right based on the range area map; using a big data recognition model to identify a verification sequence of scene object arrangement in the view data from the west side of the range area; and verifying the object model arrangement sequence based on the verification sequence of object arrangement.

[0014] This invention provides a real-scene 3D mapping system based on big data. Compared with existing technologies, it has the following advantages: This invention, through the collaborative acquisition of multi-view video data from the north, above, and west, can capture the complete spatial information of real objects within a scene, overcoming the problem of single-view occlusion. Through the processing of big data recognition models, it achieves rapid recognition of various types of real objects in the scene and automatic extraction of three-dimensional feature parameters, improving recognition accuracy and reducing manual intervention. By adopting a dynamic three-dimensional model stitching and calibration method, the position, posture, and size of the object model are adjusted in real time according to the actual recognition results, ensuring that the generated three-dimensional map structure is accurate and the boundaries are continuous, thus improving the technical level of automated three-dimensional mapping. This invention intelligently compares the initially generated object model arrangement sequence with the real object arrangement verification sequence collected from different perspectives. By detecting category mismatch and abnormal sequence problems, it discovers and locates potential abnormal areas, issues a secondary acquisition command, performs more detailed multi-view supplementary acquisition of abnormal areas, and combines a secondary recognition algorithm to re-analyze and correct abnormal areas. This improves the timeliness of problem discovery and processing, effectively reduces misjudgment and missed detection, and significantly improves the overall reliability and application accuracy of 3D mapping results. Attached Figure Description

[0015] Figure 1 This is a schematic diagram of the structural framework of the system of the present invention; Figure 2 This is a flowchart illustrating the system of the present invention. Detailed Implementation

[0016] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0017] Example 1:

[0018] Please see Figure 1 This application provides a real-scene 3D mapping system based on big data, including: The acquisition module is used by the camera acquisition device to acquire video data of the area. The video data includes the northern view data of the area, the upper view data of the area, and the western view data of the area. The recognition module uses a big data recognition model to identify real objects in the scene from the perspective data above the range area. The big data recognition model identifies the perspective data above the range area and marks the types of real objects in the scene. It then selects three-dimensional object models and combines them into a map of the range area. The height recognition module uses a big data recognition model to identify the height values ​​of real objects in the scene from the northern view of the range area, and modifies the 3D object models in the range area map according to the range area map. The verification module outputs a sequence of object models arranged from left to right based on the map of the range area. It uses a big data recognition model to identify the real object arrangement verification sequence of scene real objects in the western view data of the range area. It verifies the object model arrangement sequence based on the real object arrangement verification sequence to obtain the area to be verified. The area to be reviewed undergoes secondary data collection and secondary identification.

[0019] The process by which the big data recognition model identifies real objects in a scene from a perspective above a given area includes: S21. Input the collected view data above the range area into the big data recognition model. The model adopts a deep convolutional neural network structure to perform image normalization, edge enhancement and convolutional feature extraction on the input image to obtain an image feature map. S22. Using a semantic segmentation algorithm, candidate regions are generated and pixel-level divisions are performed on the image feature map, and multiple candidate boxes are output. The candidate boxes correspond to potential scene object regions in the image. S23. Apply a multi-type classification network to the candidate box regions respectively, identify and output the corresponding scene object type labels, wherein the scene objects include buildings, lanes, vegetation and rivers.

[0020] It is worth noting that the big data recognition model uses ResNet50 as its basic network structure. Specifically, the ResNet50 model is pre-trained on large-scale image datasets such as ImageNet to obtain well-initialized parameters. The system integrates mainstream deep learning frameworks (including but not limited to PyTorch, TensorFlow, and Keras), and can directly call the officially provided ResNet50 pre-trained weights. During application, users do not need to train the model from scratch; they only need to load the pre-trained ResNet50 model parameters to efficiently extract features or identify categories from newly acquired scene images, significantly reducing training costs and time, and improving the overall system's recognition efficiency and practicality.

[0021] The steps to combine real objects in a scene into a map of a defined area include: S31. Based on the scene object type label, call the corresponding standard 3D object model template in the 3D model library, where the 3D model library is a collection of model templates; S32. Extract geographic planar coordinates (X,Y) from the view data above the range area; S33. Based on the pixel size of the real object in the view data above the range area, scale the selected 3D object model proportionally; and perform pose rotation on the model by referring to the direction in the view data above the range area. S34. Overlay the 3D object model onto the map model according to the geographic plane coordinates (X,Y), and process the boundaries between the 3D object models through an automatic stitching algorithm; S35. Perform texture mapping on the map model to obtain a 3D map of the area.

[0022] The output type labels (such as "building", "road", "vegetation", "river") are automatically retrieved from the 3D model library and the corresponding standard 3D object model template is called. This 3D model library usually contains rich parametric model data, and each template not only contains the basic geometry, but also metadata that adapts to various scaling and rotation transformations.

[0023] The system employs various methods, such as image measurement, sensor positioning, or spatial reference points, to automatically extract the coordinates (u,v) of the two-dimensional projection center point of each real object in the scene from the image from the upper viewpoint. Combining camera parameters (such as intrinsic matrix and distortion coefficients) and known geographic benchmarks, the two-dimensional pixel coordinates are accurately converted into actual geographic plane coordinates (X,Y) through mapping functions (such as affine transformation or homography matrix).

[0024] After the map model is structurally stitched together, the system performs texture mapping on the surface of each 3D object model. Specifically, this involves projecting the acquired original image or high-resolution texture photograph onto the model surface to make its appearance details match the real scene. This process typically employs UV coordinate mapping or multi-view stitching techniques to achieve detail restoration and enhance visual realism.

[0025] Extracting geographic plane coordinates (X,Y) from the view data above the range area includes: extracting the coordinates (u,v) of the two-dimensional projection center point of the scene object and the bounding box size information from the view data above the range area, and mapping the two-dimensional projection center point coordinates to geographic plane coordinates (X,Y).

[0026] Where u0,v0 are the image origin, (X0,Y0) are the geographic coordinate origin, Sx and Sy are the resolution of the distance from the pixel to the plane, and u,v are the coordinates of the center point of the two-dimensional projection.

[0027] The process of using a big data recognition model to identify the height values ​​of real objects in the scene from the northern viewpoint of the specified area includes: S41. Perform image preprocessing on the northern view data of the range area. The image preprocessing includes distortion correction, edge enhancement and illumination equalization. S42. Input the northern view data of the range into the big data recognition model, and use semantic segmentation and object detection algorithms to extract the target bounding box; S43. The height value of the target bounding box is calculated by applying a monocular image depth regression model set to the target bounding box. S44. Match the target bounding box with the scene object; S45. Adjust the height of the corresponding 3D object model according to the height value of the target bounding box.

[0028] It is worth noting that the preprocessed image is input into the big data recognition model. The model uses semantic segmentation algorithms (such as DeepLabV3+) to divide the image into pixels and initially identify real objects in different areas of the scene. Then, object detection algorithms (such as YOLOv5, Faster R-CNN, etc.) are used to accurately extract the bounding box of each object and record the pixel coordinates and size of each bounding box.

[0029] The system automatically matches the spatial location, type label, and other information of each bounding box with the list of real objects in the original scene to ensure that the height information of each object is accurately associated with the corresponding 3D object model. The matching strategy uses a spatial overlap determination method.

[0030] The calculation is performed using a monocular image depth regression model set, including: projecting the center pixels of the upper and lower edges of the target bounding box onto the camera coordinate system based on the camera intrinsic parameter matrix, obtaining the vertical coordinates of the two points through calculation, and using the absolute value of the difference between the vertical coordinates of the two points as the height value of the target bounding box.

[0031] It's worth noting that the system acquires the intrinsic parameter matrix of the camera used to capture the image, which is used to transform between pixel coordinates and camera 3D coordinates. For each target bounding box, the system automatically determines the positions of its top and bottom center pixels, and then projects these two pixels onto the camera coordinate system using the camera intrinsic parameter matrix. Through a depth regression model, the system obtains the vertical coordinates of the projected points in the camera coordinate system, and further calculates the absolute value of the difference between the vertical coordinates of the top and bottom center points, which serves as an estimate of the actual height of the object corresponding to the target bounding box. To ensure the accuracy of the height estimation, the system can also perform scale correction and error correction by combining standard objects with known heights in the scene.

[0032] Beneficial effects: This invention, through the collaborative acquisition of video data from multiple perspectives (north, above, and west), can capture complete spatial information of real objects within a scene, overcoming the problem of single-view occlusion. Through the processing of big data recognition models, it achieves rapid identification of various types of real objects in the scene and automatic extraction of three-dimensional feature parameters, improving recognition accuracy and reducing manual intervention. By adopting a dynamic three-dimensional model stitching and calibration method, the position, posture, and size of the object model are adjusted in real time according to the actual recognition results, ensuring that the generated three-dimensional map structure is accurate and the boundaries are continuous, thus improving the technical level of automated three-dimensional mapping.

[0033] Example 2:

[0034] Output a sequence of object models arranged from left to right based on the area map, including: S501. Extract the coordinates of the two-dimensional projection center point of each three-dimensional object model from the three-dimensional map of the range area; S502. Sort the center point coordinates in ascending order of horizontal coordinate values ​​to form object model arrangement sequence L1. S503. Input the western view image of the range area into the big data recognition model. The big data recognition model performs target detection on the image and outputs the object arrangement verification sequence L2 from the western view.

[0035] The object arrangement verification sequence for identifying real objects in the scene from the western perspective of the defined area includes: S511. Compare sequence L1 and sequence L2 to determine the consistency of their positional order; S512. If there is a type mismatch or order mismatch in the comparison, the sequence L1 is marked as an abnormal sequence. S513. Define the spatial region corresponding to the abnormal sequence as the region to be reviewed.

[0036] It is worth noting that the geometric center or base centroid of each 3D object model in the 3D map of the area is calculated, and then precisely projected onto the 2D image plane using the camera extrinsic projection matrix to obtain pixel-level center points (u,v). For high-precision areas, the complete geometric center projection is used; for distant or low-precision areas, the bounding box centroid is used for simplified processing, thus balancing operational efficiency and positioning accuracy.

[0037] All center points are sorted in ascending order of their horizontal pixel coordinates u, forming a left-to-right model sequence L1. During the sorting process, a pixel tolerance δ_x (e.g., ±3px) is introduced to filter out minor measurement errors. If the difference in u between two models is less than the tolerance, they are considered to be in the same horizontal position and are then sorted a second time based on depth or category priority to ensure a stable and reliable order.

[0038] The original image from the western viewpoint is fed into a big data recognition model. First, YOLOv8 detects the target bounding boxes and assigns them to categories. Then, SAM is used to fine-tune the mask to obtain the precise center point. The system discards targets with a detection confidence score below 0.25 and performs non-maximum suppression on overlapping boxes (IoU>0.6), retaining only those with the highest confidence scores. Subsequently, the images are sorted according to the horizontal coordinates of the center points to obtain the western viewpoint verification sequence L2.

[0039] If a mismatch or overlapping arrangement is detected, the system marks the corresponding L1 segment as an abnormal sequence and classifies it into type abnormality, sequence abnormality, or compound abnormality according to the nature of the abnormality. If there are m (≥2) consecutive model abnormalities, they are automatically merged into one segment to prevent fragmented review and improve processing efficiency. Based on the three-dimensional bounding boxes of the first and last models of the abnormal sequence, the system generates a region to be reviewed by proportional buffering (e.g., 1.2 times) and records it in the task queue.

[0040] The secondary acquisition and secondary recognition of the area to be reviewed includes: generating a secondary acquisition task based on the area to be reviewed; instructing the camera acquisition device to acquire image data of the area to be reviewed; using a big data recognition model to identify scene objects in the view data above the range area; selecting a 3D object model; updating the 3D object model to the range area map; outputting a sequence of object models arranged from left to right based on the range area map; using a big data recognition model to identify a verification sequence of scene object arrangement in the view data from the west side of the range area; and verifying the object model arrangement sequence based on the verification sequence of object arrangement.

[0041] Once the system identifies the area to be reviewed, it first generates a secondary acquisition task, specifying the polygonal range of the area, the preferred acquisition angle (east or south), the target resolution and overlap rate requirements, and selects the optimal UAV or vehicle-mounted gimbal and flight path. The acquisition platform completes multi-angle image acquisition according to the planned path, and qualified images are then transmitted to the recognition module for re-recognition.

[0042] Beneficial effects: This invention intelligently compares the initially generated object model arrangement sequence with the real object arrangement verification sequence collected from different perspectives. By detecting category mismatch and sequence abnormalities, it discovers and locates potential abnormal areas, issues secondary acquisition commands, performs more detailed multi-view supplementary acquisition of abnormal areas, and combines secondary recognition algorithms to re-analyze and correct abnormal areas. This improves the timeliness of problem discovery and processing, effectively reduces misjudgment and missed detection, and significantly improves the overall reliability and application accuracy of 3D mapping results.

[0043] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of the present invention are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via a wired or wireless network. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more sets of available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. A semiconductor medium can be a solid-state drive.

[0044] Some of the data in the above formulas are numerical calculations with dimensions removed, and the contents not described in detail in this specification are all prior art known to those skilled in the art.

[0045] The above embodiments are only used to illustrate the technical methods of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical methods of the present invention without departing from the spirit and scope of the technical methods of the present invention.

Claims

1. A real-scene 3D mapping system based on big data, characterized in that, include: The acquisition module is used by the camera acquisition device to acquire video data of the area. The video data includes the northern view data of the area, the upper view data of the area, and the western view data of the area. The identification module uses a big data identification model to identify real objects in the scene from the perspective data above the range area. The big data identification model identifies the perspective data above the range area and marks the real objects in the scene as types. Select 3D object models and combine them into a map of the area. The height recognition module uses a big data recognition model to identify the height values ​​of real objects in the scene from the northern view of the range area, and modifies the 3D object models in the range area map according to the range area map. The verification module outputs a sequence of object models arranged from left to right based on the map of the range area. It uses a big data recognition model to identify the real object arrangement verification sequence of scene real objects in the western view data of the range area. It verifies the object model arrangement sequence based on the real object arrangement verification sequence to obtain the area to be verified. The area to be reviewed undergoes secondary data collection and secondary identification.

2. The real-scene 3D mapping system based on big data according to claim 1, characterized in that, The process by which the big data recognition model identifies real objects in a scene from a perspective above a given area includes: S21. Input the collected view data above the range area into the big data recognition model. The model adopts a deep convolutional neural network structure to perform image normalization, edge enhancement and convolutional feature extraction on the input image to obtain an image feature map. S22. Using a semantic segmentation algorithm, candidate regions are generated and pixel-level divisions are performed on the image feature map, and multiple candidate boxes are output. The candidate boxes correspond to potential scene object regions in the image. S23. Apply a multi-type classification network to the candidate box regions respectively, identify and output the corresponding scene object type labels, wherein the scene objects include buildings, lanes, vegetation and rivers.

3. The real-scene 3D mapping system based on big data according to claim 1, characterized in that, The steps to combine real objects in a scene into a map of a defined area include: S31. Based on the scene object type label, call the corresponding standard 3D object model template in the 3D model library, where the 3D model library is a collection of model templates; S32. Extract geographic planar coordinates (X,Y) from the view data above the range area; S33. Based on the pixel size of the real object in the view data above the range area, scale the selected 3D object model proportionally; and perform pose rotation on the model by referring to the direction in the view data above the range area. S34. Overlay the 3D object model onto the map model according to the geographic plane coordinates (X,Y), and process the boundaries between the 3D object models through an automatic stitching algorithm; S35. Perform texture mapping on the map model to obtain a 3D map of the area.

4. The real-scene 3D mapping system based on big data according to claim 3, characterized in that, Extracting geographic plane coordinates (X,Y) from the view data above the range area includes: extracting the coordinates (u,v) of the two-dimensional projection center point of the scene object and the bounding box size information from the view data above the range area, and mapping the two-dimensional projection center point coordinates to geographic plane coordinates (X,Y). Where u0,v0 are the image origin, (X0,Y0) are the geographic coordinate origin, Sx and Sy are the resolution of the distance from the pixel to the plane, and u,v are the coordinates of the center point of the two-dimensional projection.

5. The real-scene 3D mapping system based on big data according to claim 1, characterized in that, The process of using a big data recognition model to identify the height values ​​of real objects in the scene from the northern viewpoint of the specified area includes: S41. Perform image preprocessing on the northern view data of the range area. The image preprocessing includes distortion correction, edge enhancement and illumination equalization. S42. Input the northern view data of the range into the big data recognition model, and use semantic segmentation and object detection algorithms to extract the target bounding box; S43. The height value of the target bounding box is calculated by applying a monocular image depth regression model set to the target bounding box. S44. Match the target bounding box with the scene object; S45. Adjust the height of the corresponding 3D object model according to the height value of the target bounding box.

6. The real-scene 3D mapping system based on big data according to claim 5, characterized in that, The calculation is performed using a monocular image depth regression model set, including: projecting the center pixels of the upper and lower edges of the target bounding box onto the camera coordinate system based on the camera intrinsic parameter matrix, obtaining the vertical coordinates of the two points through calculation, and using the absolute value of the difference between the vertical coordinates of the two points as the height value of the target bounding box.

7. The real-scene 3D mapping system based on big data according to claim 1, characterized in that, Output a sequence of object models arranged from left to right based on the area map, including: S501. Extract the coordinates of the two-dimensional projection center point of each three-dimensional object model from the three-dimensional map of the range area; S502. Sort the center point coordinates in ascending order of horizontal coordinate values ​​to form object model arrangement sequence L1. S503. Input the western view image of the range area into the big data recognition model. The big data recognition model performs target detection on the image and outputs the object arrangement verification sequence L2 from the western view.

8. A real-scene 3D mapping system based on big data according to claim 7, characterized in that, The object arrangement verification sequence for identifying real objects in the scene from the western perspective of the defined area includes: S511. Compare sequence L1 and sequence L2 to determine the consistency of their positional order; S512. If there is a type mismatch or order mismatch in the comparison, the sequence L1 is marked as an abnormal sequence. S513. Define the spatial region corresponding to the abnormal sequence as the region to be reviewed.

9. A real-scene 3D mapping system based on big data according to claim 1, characterized in that, The secondary acquisition and secondary recognition of the area to be reviewed includes: generating a secondary acquisition task based on the area to be reviewed; instructing the camera acquisition device to acquire image data of the area to be reviewed; using a big data recognition model to identify scene objects in the view data above the range area; selecting a 3D object model; updating the 3D object model to the range area map; outputting a sequence of object models arranged from left to right based on the range area map; using a big data recognition model to identify a verification sequence of scene object arrangement in the view data from the west side of the range area; and verifying the object model arrangement sequence based on the verification sequence of object arrangement.

Images