Intelligently acquired live-action three-dimensional map comprehensive modeling method

Abstract

The invention discloses an intelligent acquisition live-action three-dimensional map comprehensive modeling method, which comprises the following steps of: 1, acquiring relevant information of a target place, and processing the relevant information of the target place to obtain acquisition place evaluation information; 2, acquiring place evaluation information comprises a first-class acquiring place, a second-class acquiring place and a third-class acquiring place; step 3, when the acquisition place is a first-class acquisition place, performing live-action three-dimensional construction information acquisition of the acquisition place by using a first acquisition mode; step 4, when the acquisition place is a second-class acquisition place, performing actual three-dimensional construction information acquisition of the acquisition place by using a second acquisition mode; and step 5, when the acquisition places are three types of acquisition places, performing actual three-dimensional construction information acquisition of the acquisition places by using a third acquisition mode. According to the invention, more intelligent live-action three-dimensional map comprehensive acquisition can be carried out.

Description

Technical Field

[0001] The present invention relates to the field of modeling methods, and particularly to a comprehensive modeling method for real - scene three - dimensional maps with intelligent acquisition. Background Art

[0002] With the continuous development of geographic information technology, computer technology, and sensor technology, real - scene three - dimensional maps have been widely used in many fields, such as urban planning, intelligent transportation, disaster assessment, cultural heritage protection, etc. Traditional real - scene three - dimensional map modeling methods have certain limitations and are difficult to meet the increasing requirements for high precision, high efficiency, and intelligence.

[0003] In actual application scenarios, the geographical environment, building distribution, and vegetation coverage in different regions vary greatly. A single acquisition method cannot be applied to all types of regions. For example, for regions with large areas, complex building and vegetation distributions, traditional acquisition means may not be able to obtain data comprehensively and accurately; while for some small regions with high - detail requirements, conventional methods may be too cumbersome and inefficient.

[0004] In addition, with the development of intelligent technology, various industries have put forward higher requirements for the accuracy, details, and real - time performance of real - scene three - dimensional maps. Therefore, a comprehensive modeling method for real - scene three - dimensional maps that can intelligently select appropriate acquisition methods according to the characteristics of different regions is needed to improve the modeling efficiency and quality and meet diverse application requirements. Summary of the Invention

[0005] The technical problem to be solved by the present invention is: how to solve the problem that existing modeling methods cannot obtain data comprehensively and accurately, and provides a comprehensive modeling method for real - scene three - dimensional maps with intelligent acquisition.

[0006] The present invention solves the above - mentioned technical problem through the following technical solutions. The present invention includes the following steps:

[0007] Step 1: Collect relevant information of the target area and process the relevant information of the target area to obtain the acquisition area evaluation information;

[0008] Step 2: The acquisition area evaluation information includes Class - I acquisition areas, Class - II acquisition areas, and Class - III acquisition areas;

[0009] Step 3: When the acquisition area is a Class - I acquisition area, use the first acquisition method to collect the real - scene three - dimensional construction information of the acquisition area;

[0010] Step 4: When the acquisition area is a Class - II acquisition area, use the second acquisition method to collect the actual three - dimensional construction information of the acquisition area;

[0011] Step 5: When the collection site is a type-3 collection site, use the third collection method to collect the actual 3D construction information of the collection site;

[0012] Step 6: After obtaining the real-scene 3D construction information, conduct verification of the real-scene 3D construction information. If the verification passes, export it for real-scene 3D map construction.

[0013] Furthermore, the specific process of processing the relevant information of the target site to obtain the collection site evaluation information is as follows:

[0014] Extract the relevant information of the target site obtained. The relevant information of the target site includes the target site area information, the target site building quantity information, and the target site vegetation coverage area information;

[0015] Process the target site area information, the target site building quantity information, and the target site vegetation coverage area information to obtain the allocation parameter;

[0016] When the allocation parameter is greater than the preset value a1, it is evaluated as a type-1 collection site;

[0017] When the allocation parameter is between the preset values a1 and a2, it is evaluated as a type-2 collection site;

[0018] When the allocation parameter is less than the preset value a2, it is evaluated as a type-3 collection site;

[0019] a2 < a1.

[0020] Furthermore, the process of obtaining the allocation parameter is as follows:

[0021] Let the target site area information be S (unit: square meters), the target site building quantity information be B (unit: building), and the target site vegetation coverage area information be V (unit: square meters).

[0022] Assign the weight w to the target site area information S 1 , assign the weight w to the target site building quantity information B 2 , assign the weight w to the target site vegetation coverage area information V 3 , and w 1 + w 2 + w 3 = 1;

[0023] Obtain the allocation parameter P through the formula ;

[0024] Among them, S max , B max , V max are respectively the maximum values that the target site area, the building quantity, and the vegetation coverage area may appear in the target site.

[0025] Furthermore, the specific process of collecting the real - scene three - dimensional construction information of the collection site using the first collection method is as follows:

[0026] Step 1: Equipment deployment: Select a lidar scanner, install the lidar on a mobile collection platform, and synchronously deploy multiple high - resolution panoramic cameras. The cameras are distributed around the collection platform, and the camera parameters are calibrated in advance and set to a preset mode;

[0027] Step 2: Data collection: The collection vehicle travels at a preset speed along a predetermined route. The lidar emits laser beams in real - time, measures the time and intensity of the reflected light, and accurately calculates the distance, azimuth, and surface feature signals of the target object to form point - cloud data;

[0028] The panoramic cameras work synchronously, continuously shooting images of the surrounding environment. The captured images not only contain the visual appearance information of the collection site, but also include the colors and textures of buildings, the types and forms of vegetation, and at the same time record environmental details such as the lighting conditions and weather conditions at that time;

[0029] Step 3: Data pre - processing: After the collection is completed, the point - cloud data obtained by the lidar is transmitted to professional processing software. First, noise reduction processing is performed. Algorithms are used to identify and eliminate abnormal points caused by environmental interference. Then, point - cloud registration is carried out. The point - cloud data collected from different positions and perspectives is unified into the same coordinate system. Through feature - matching algorithms, the overlapping parts of adjacent point - cloud data are found, and their relative position relationships are calculated, so that the scattered point - clouds are stitched into a complete three - dimensional scene;

[0030] For the images taken by the panoramic cameras, they are also imported into image - processing software for preliminary processing. Distortion correction is performed on the images to eliminate image distortion caused by the optical characteristics of the camera lens;

[0031] The images are automatically stitched according to the shooting time sequence and the preset overlapping areas;

[0032] Step 4: Information fusion: The processed point - cloud data is fused with the panoramic image sequence. The texture information of the images is used to supplement the appearance details of the point - cloud data. Through preset algorithms, the colors and textures in the images are mapped to the surface of the point - cloud according to the three - dimensional spatial positions reflected by the point - cloud, and the real - scene three - dimensional construction information is obtained.

[0033] Furthermore, the specific process of collecting the actual three - dimensional construction information of the collection site using the second collection method is as follows:

[0034] Step (1): Equipment selection: Use a handheld three - dimensional laser scanner, paired with a portable panoramic cloud - platform camera;

[0035] Step (2): Data acquisition: The collector holds a 3D laser scanner and moves at a preset speed along the pre-planned acquisition path to scan the acquisition site in detail. Meanwhile, operate the panoramic pan-tilt camera synchronously and stay at each acquisition point.

[0036] Step (3): Data processing: Import the point cloud data collected by the handheld laser scanner into computer software. First, conduct manual screening to remove invalid data caused by operator occlusion and misoperation. Then, use a feature-based registration algorithm to splice the point cloud data collected at different stations to form a locally complete 3D model fragment.

[0037] For the panoramic image, perform rapid splicing and correction in the software, use the EXIF information of the image to assist in splicing. After that, preliminarily associate the processed panoramic image with the corresponding point cloud model fragment to obtain the real-scene 3D construction information.

[0038] Furthermore, there are two groups of people in the step (1). Extract the starting point and the ending point of the pre-planned acquisition path.

[0039] Set a group of people at the starting point to conduct data acquisition in the direction from the starting point to the ending point, and mark the data collected by this group of people at each acquisition point as Hi, where i is the number of acquisition points.

[0040] Set a group of people at the ending point to conduct data acquisition in the direction from the ending point to the starting point, and mark the data collected by this group of people at each acquisition point as Ui, where i is the number of acquisition points.

[0041] Before data processing in step (3), first compare and analyze the data Hi and Ui.

[0042] Randomly select at least m data of acquisition points from Hi and mark them as Ym.

[0043] Then select the data of the acquisition points at the corresponding positions of the m acquisition points randomly selected from Hi from Ui and mark them as Em.

[0044] Compare the similarity between Em and Ym. When the number of similarities greater than the preset value in Em and Ym is m, directly proceed with the subsequent data processing.

[0045] When the number of similarities greater than the preset value in Em and Ym is less than m, re-acquire the data of the acquisition points with similarities less than the preset value.

[0046] Furthermore, the specific process of using the third acquisition method to collect the actual 3D construction information of the acquisition site is as follows:

[0047] Step 1): Tool Preparation: Select a drone equipped with an optical camera and a laser ranging module, and prepare ground control point markers at the same time.

[0048] Step 2): Data Acquisition: Control the drone to fly automatically along a preset route. During the flight, the optical camera automatically takes orthophotos at preset time intervals.

[0049] At the same time, the laser ranging module measures the vertical distance to the ground every preset distance and records the relative height information.

[0050] At the collection points set at the collection site, manually control the drone to hover, take close-up images to supplement detailed information, and place control point markers at the corresponding positions on the ground and record their coordinate information.

[0051] Step 3): Data Processing: Import the images and ranging data collected by the drone into professional software, use the ground control points for geometric correction of the images, correct the images to the true geographic coordinate system, and then extract feature points from the orthophotos through photogrammetry algorithms to generate sparse point cloud data. Combine the laser ranging data to optimize the sparse point cloud, increase the point cloud density, screen and process the multi-angle close-up images, and extract the useful texture information therein, that is, obtain the actual three-dimensional construction information.

[0052] Furthermore, during the process of collecting the actual three-dimensional construction information of the collection site using the third collection method, the drone used for data collection was intelligently monitored, and the content of its intelligent monitoring is as follows:

[0053] Through the sensors of the flight control system built into the drone, the flight altitude, flight speed, heading angle, pitch angle, and roll angle of the drone are obtained in real time.

[0054] When any one of the flight altitude, flight speed, heading angle, pitch angle, and roll angle of the drone exceeds its corresponding standard value, the drone automatically activates the return flight program.

[0055] At the same time, the battery power and motor speed of the drone are monitored.

[0056] When the battery power is lower than the preset safe return flight power threshold, the drone automatically activates the return flight program.

[0057] Moreover, if it is detected that the speed of a certain motor fluctuates abnormally or the difference in speed from other motors exceeds the preset range, the system will issue an alarm and attempt to automatically adjust the flight attitude to maintain stable flight. If the speed of a certain motor fluctuates abnormally or the difference in speed from other motors exceeds the preset range and exceeds the preset duration, the drone automatically activates the return flight program.

[0058] During the flight of the unmanned aerial vehicle (UAV), it regularly conducts self-checks on the carried optical camera and simple laser ranging module. When the self-check is abnormal, it restarts the device or adjusts the flight mission, and checks whether the focus of the camera is normal, whether there is a malfunction in the image sensor, and whether the write speed of the memory card meets the requirements, etc. For the laser ranging module, it monitors its laser emission power, receiving sensitivity, and data transmission stability. If it is found that the test images taken by the camera are blurred, the noise increases abnormally, or there are continuous packet losses in the laser ranging data, it promptly prompts the operator at the ground control terminal so that corresponding measures can be taken;

[0059] The UAV is equipped with a small meteorological sensor to real-time sense and collect the meteorological information of the collection site, including wind speed, wind direction, air temperature, and air pressure;

[0060] When the wind speed exceeds the upper limit of the safe flight wind speed of the UAV (generally 10 - 15 m / s for consumer-grade UAVs, depending on the specific model), the system will issue a strong wind warning, reminding the operator to operate carefully or suspend the flight mission to avoid the UAV getting out of control due to strong winds;

[0061] Based on the air temperature and air pressure data, the air density is estimated, and then the flight parameters of the UAV, such as the lift coefficient, are automatically adjusted to ensure flight stability.

[0062] The present invention has the following advantages compared with the prior art: This intelligent acquisition method for integrated modeling of real-scene three-dimensional maps calculates and allocates parameters according to information such as the area of the target site, the number of buildings, and the vegetation coverage area, and divides the collection site into category one, category two, and category three. Different collection methods are adopted for different categories of collection sites. For example, for category one collection sites, a mobile collection platform is used to carry a lidar and a panoramic camera, which is suitable for rapid collection in large areas; for category two collection sites, handheld devices are used, which are convenient for fine collection in small areas; for category three collection sites, UAVs are used, which can obtain complex terrain and high-altitude information. The combination of multiple collection methods meets the requirements of different scenarios and improves the collection efficiency and accuracy.

[0063] After data collection, each type of collection method has a complete data processing process. The data collected by the lidar is denoised and registered, and the images of the panoramic camera are corrected for distortion and stitched. Then, the point cloud data and the images are fused to ensure that the obtained real-scene three-dimensional construction information is accurate and complete, providing high-quality data for the construction of real-scene three-dimensional maps. In the handheld collection method, for example, two groups of personnel collect data from different directions and compare them, which can promptly detect and correct collection errors, further ensuring data quality.

[0064] When using a drone for data collection, the drone is intelligently monitored. By monitoring flight parameters, battery power, motor speed, etc., when the parameters are abnormal, the return flight program is automatically started or the flight attitude is adjusted to ensure the safety of the drone; the camera and laser rangefinder module are self-checked, and problems are handled in a timely manner to ensure the reliability of the collected data; a meteorological sensor is carried, and the flight parameters are adjusted according to the meteorological information to avoid being affected by bad weather during data collection and ensure the smooth progress of the entire data collection process. BRIEF DESCRIPTION OF THE DRAWINGS

[0065] Figure 1 is the overall flowchart of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0066] The embodiments of the present invention will be described in detail below. These embodiments are implemented on the premise of the technical solution of the present invention, and detailed implementation manners and specific operation processes are given. However, the protection scope of the present invention is not limited to the following embodiments.

[0067] As Figure 1 shown, this embodiment provides a technical solution: an integrated modeling method for a real-scene three-dimensional map of intelligent data collection, including the following steps:

[0068] Step 1: Collect relevant information of the target location, and process the relevant information of the target location to obtain the evaluation information of the collection location;

[0069] Step 2: The evaluation information of the collection location includes a first-class collection location, a second-class collection location, and a third-class collection location;

[0070] Step 3: When the collection location is a first-class collection location, use the first collection method to collect the real-scene three-dimensional construction information of the collection location;

[0071] Step 4: When the collection location is a second-class collection location, use the second collection method to collect the actual three-dimensional construction information of the collection location;

[0072] Step 5: When the collection location is a third-class collection location, use the third collection method to collect the actual three-dimensional construction information of the collection location;

[0073] Step 6: When the real-scene three-dimensional construction information is obtained, verify the real-scene three-dimensional construction information. If the verification is passed, export it for the construction of the real-scene three-dimensional map.

[0074] The specific process of processing the relevant information of the target location to obtain the evaluation information of the collection location is as follows:

[0075] Extract the relevant information of the target location obtained. The relevant information of the target location includes the target location area information, the target location building quantity information, and the target location vegetation coverage area information;

[0076] Process the target area information, the number of buildings in the target area, and the vegetation coverage area information of the target area to obtain the allocation parameter;

[0077] When the allocation parameter is greater than the preset value a1, it is evaluated as a first-class collection site;

[0078] When the allocation parameter is between the preset values a1 and a2, it is evaluated as a second-class collection site;

[0079] When the allocation parameter is less than the preset value a2, it is evaluated as a third-class collection site;

[0080] a2 < a1;

[0081] Determining the collection site category through comprehensive consideration of information such as the target area, the number of buildings, and the vegetation coverage area helps to reasonably allocate collection resources. For first-class collection sites, which may have a large area, complex buildings and vegetation, a more efficient mobile collection platform equipped with lidar and panoramic cameras is adopted; second-class collection sites are relatively smaller in scale or slightly simpler in situation, and handheld devices are selected for fine collection; for third-class collection sites, drones are used to obtain unique perspective information. Selecting the appropriate collection method according to different categories can avoid resource waste and improve collection efficiency.

[0082] The clear classification criteria make the collection work more targeted and can obtain more accurate data based on the characteristics of different regions. For different categories of collection sites, the differences in collection equipment and methods can better adapt to their geographical features and environmental conditions. For example, in areas with a large vegetation coverage area, the appropriate collection method can more accurately obtain information such as the shape and type of vegetation, ensuring the accuracy and integrity of the data when constructing the real scene 3D map and improving the quality and accuracy of the map.

[0083] Enhance data usability: The information involved in the collection site classification and evaluation process is closely related to the actual geographical environment. The real scene 3D map obtained after processing can more intuitively and accurately reflect the real situation of the target area. For urban planners, they can clearly see the building distribution and area information from the map; for ecological researchers, it is convenient to obtain vegetation coverage data. This usability enables the map to play an important role in multiple fields and provides strong support for decision-making, research and analysis, etc.

[0084] The process of obtaining the allocation parameter is as follows:

[0085] Let the target area information be S (unit: square meters), the number of buildings in the target area be B (unit: building), and the vegetation coverage area information of the target area be V (unit: square meters).

[0086] Assign the weight w to the target area information S 1 and assign the weight w to the number of buildings in the target area B2 The vegetation coverage area information V of the destination is assigned a weight w 3 , and w 1 +w 2 +w 3 = 1;

[0087] The allocation parameter P is obtained through the formula ;

[0088] where S max , B max , V max are respectively the maximum values that the destination area, the number of buildings, and the vegetation coverage area may appear within the destination;

[0089] The destination area information S, the number of buildings information B, and the vegetation coverage area information V are incorporated into the calculation and are respectively assigned weights, comprehensively considering the influence of different geographical features on the acquisition work. In different regions, the importance of these three factors may be different, and the weight settings can be flexibly adjusted to reflect the actual situation. In urban areas, the number and area of buildings may have a greater impact on the selection of acquisition methods; in nature reserves, the weight of the vegetation coverage area may be higher. In this way, the dominance of a single factor in the evaluation result is avoided, making the evaluation more comprehensive and objective.

[0090] The maximum values that the destination area, the number of buildings, and the vegetation coverage area may appear are introduced into the formula for normalization. This operation unifies data with different dimensions to a relatively comparable scale, eliminating the interference of dimension differences on the calculation result. It enables different types of data to be comprehensively calculated under the same standard, enhancing the scientificity and comparability of the allocation parameter calculation, and making the evaluation results between different destinations more valuable for reference.

[0091] Provide a quantitative basis for the classification of the acquisition area: The calculated allocation parameter P is a quantitative indicator. After comparison with the preset values a1 and a2, the acquisition area can be clearly and accurately divided into Class 1, Class 2, and Class 3. This quantitative classification method avoids the ambiguity of subjective judgment, making the acquisition area evaluation process more standardized and normalized. Selecting the corresponding acquisition method based on the quantitative result can ensure that the acquisition work better meets the actual needs, improve the quality and efficiency of the acquired data, and thus enhance the accuracy of the construction of the real scene three-dimensional map.

[0092] The specific process of using the first acquisition method to collect the real scene three-dimensional construction information of the acquisition area is as follows:

[0093] Step 1: Equipment deployment: Select a lidar scanner, install the lidar on a mobile acquisition platform, and synchronously deploy multiple high-resolution panoramic cameras. The positions are distributed around the acquisition platform, and the camera parameters are calibrated in advance and set to the preset mode;

[0094] Step 2: Data Acquisition: The collection vehicle travels at a preset speed along a predetermined route. The lidar emits laser beams in real time, measures the time and intensity of the reflected light, and accurately calculates the distance, azimuth, and surface feature signals of the target object to form point cloud data.

[0095] The panoramic camera works synchronously, continuously shooting images of the surrounding environment. The captured images not only contain the visual appearance information of the collection site but also include the color and texture of buildings, the types and forms of vegetation, and at the same time record environmental details such as the lighting conditions and weather conditions at that time.

[0096] Step 3: Data Preprocessing: After the acquisition is completed, the point cloud data obtained by the lidar is transmitted to professional processing software. First, noise reduction processing is performed, and algorithms are used to identify and eliminate abnormal points caused by environmental interference. Then, point cloud registration is carried out to unify the point cloud data collected from different positions and perspectives into the same coordinate system. Through feature matching algorithms, the overlapping parts of adjacent point cloud data are found, and their relative position relationships are calculated to stitch the scattered point clouds into a complete three-dimensional scene.

[0097] For the images taken by the panoramic camera, they are also imported into image processing software for preliminary processing. Distortion correction is performed on the images to eliminate image distortion caused by the optical characteristics of the camera lens.

[0098] Automatically stitch the images according to the shooting time sequence and the preset overlapping areas.

[0099] Step 4: Information Fusion: The processed point cloud data is fused with the panoramic image sequence. The texture information of the image is used to supplement the appearance details of the point cloud data. Through preset algorithms, the colors and textures in the image are mapped to the point cloud surface according to the three-dimensional spatial positions reflected by the point cloud to obtain real-scene three-dimensional construction information.

[0100] Device Collaborative Acquisition, Abundant and Comprehensive Information: The method of collaborative operation of a lidar scanner and multiple high-resolution panoramic cameras is adopted. The lidar can accurately measure the distance, azimuth, and surface feature signals of the target object, generate point cloud data, and provide the spatial position and shape information of the object. The panoramic camera can capture rich visual appearance information of the collection site, including the color and texture of buildings, the types and forms of vegetation, and environmental details such as lighting conditions and weather conditions. The combination of the two collects comprehensive and complementary data, providing sufficient information for constructing accurate and realistic real-scene three-dimensional models.

[0101] Data preprocessing ensures data quality: The preprocessing steps after data collection are crucial. Denoising of point cloud data can effectively remove abnormal points generated by environmental interference and improve data accuracy; point cloud registration unifies point cloud data from different positions and perspectives into the same coordinate system and stitches them into a complete 3D scene, ensuring the accuracy of spatial positions. For panoramic camera images, distortion correction eliminates image distortion caused by the optical characteristics of the lens, and automatically stitches the images according to the shooting time sequence and preset overlapping areas, ensuring the geometric accuracy and integrity of the images. These preprocessing operations improve data quality and lay a good foundation for subsequent information fusion and model construction.

[0102] Information fusion enhances model fidelity: Fusing the processed point cloud data with the panoramic image sequence, using the texture information of the images to supplement the appearance details of the point cloud data, and mapping the image colors and textures to the point cloud surface through a preset algorithm. This fusion method enables the constructed real-scene 3D model to not only have an accurate spatial structure but also rich texture and color information, greatly enhancing the model's fidelity and visualization effect, more realistically reflecting the actual situation of the collection site, and meeting the requirements of various application scenarios, such as urban planning, virtual tourism, etc.

[0103] The acquisition process is efficient and orderly: From the parameter calibration during equipment deployment, the collection vehicle driving according to the predetermined route and speed, to the clear processes of data preprocessing and information fusion, the entire acquisition process has strong standardization and orderliness. This helps improve the acquisition efficiency, reduce errors and uncertainties during data acquisition and processing, ensure the smooth progress of the acquisition work, and ensure the rapid and stable acquisition of high-quality real-scene 3D construction information.

[0104] The specific process of collecting the actual 3D construction information of the collection site using the second collection method is as follows:

[0105] Step (1): Equipment selection: Use a handheld 3D laser scanner, equipped with a portable panoramic pan-tilt camera;

[0106] Step (2): Data collection: The collector holds the 3D laser scanner and moves at a preset speed along the pre-planned collection path, carefully scanning the collection site, and synchronously operating the panoramic pan-tilt camera and staying at each collection point;

[0107] Step (3): Data processing: Import the point cloud data collected by the handheld laser scanner into computer software. First, perform manual screening to remove invalid data generated by operator occlusion and misoperation, and then use a feature-based registration algorithm to stitch the point cloud data collected at different positions to form a locally complete 3D model segment;

[0108] For panoramic images, rapid stitching and calibration are performed in software, and the EXIF information of the images is used to assist in stitching. After that, the processed panoramic images are preliminarily associated with the corresponding point cloud model segments to obtain the real-scene three-dimensional construction information;

[0109] A handheld 3D laser scanner and a portable panoramic cloud platform camera are selected. The equipment is lightweight and easy to carry. The acquisition personnel can hold the equipment and move along the pre-planned path, stopping at each acquisition point for scanning. It can penetrate into complex areas such as narrow streets and indoor spaces that are difficult to reach by large acquisition equipment for detailed scanning. Compared with the mobile acquisition platform of the first acquisition method, this method is not restricted by the site space, greatly expanding the acquisition range and meeting diverse acquisition needs.

[0110] In the data processing link, the collected point cloud data is first manually screened to remove invalid data generated by operator occlusion and misoperation, effectively avoiding the interference of incorrect data on the final result. The point cloud data collected at different stations is stitched using a feature-based registration algorithm, improving the accuracy of the stitching of the 3D model segments. The panoramic images are assisted in stitching and calibration using EXIF information, further ensuring the accuracy and integrity of the images. Through these operations, it is ensured that the obtained real-scene three-dimensional construction information can truly reflect the actual situation of the acquisition site.

[0111] Data processing is mainly carried out in computer software, and the process is relatively simple and direct. The stitching of the point cloud data and the processing of the panoramic images are both completed within the software, and the EXIF information is used to assist in stitching the panoramic images, reducing manual intervention and complex operations. By preliminarily associating the processed panoramic images with the point cloud model segments, the real-scene three-dimensional construction information can be quickly obtained, improving the data processing efficiency and saving time and labor costs.

[0112] The verification mechanism improves data reliability: In some cases, two groups of personnel are arranged to collect data bidirectionally from the starting point and the ending point, and a comparison and analysis are carried out before data processing. By randomly selecting the data at the acquisition points for similarity comparison, if the similarity does not meet the standard, re-acquisition is carried out. This method can timely detect problems in the acquisition process, further guarantee the data quality, and improve the reliability of the finally obtained real-scene three-dimensional construction information.

[0113] The personnel in step (1) are two groups, and the starting point and the ending point of the pre-planned acquisition path are extracted;

[0114] One group of personnel is set at the starting point to collect data in the direction from the starting point to the ending point, and the data collected by this group of personnel at each acquisition point is marked as Hi, where i is the number of acquisition points;

[0115] One group of personnel is set at the ending point to collect data in the direction from the ending point to the starting point, and the data collected by this group of personnel at each acquisition point is marked as Ui, where i is the number of acquisition points;

[0116] Before the data processing in step (3), the data Hi and Ui are first compared and analyzed;

[0117] Randomly select at least m data of the acquisition points from Hi and label them as Ym;

[0118] Then select the data of the acquisition points at the corresponding positions of the m acquisition points randomly selected from Hi from Ui and label them as Em;

[0119] Compare the similarity between Em and Ym. When the number of similarities greater than the preset value in Em and Ym is m, directly perform the subsequent data processing;

[0120] When the number of similarities greater than the preset value in Em and Ym is less than m, re-acquire the data of the acquisition points with similarities less than the preset value;

[0121] Arrange two groups of personnel to collect data from the starting point and the ending point along the acquisition path in opposite directions, forming a cross-validation mechanism. Since the operation processes of the two groups of personnel are independent of each other, when collecting data in different directions, the errors caused by factors such as operating habits and environmental impacts are random. By comparing the two groups of data, the acquisition errors caused by accidental factors can be effectively detected. For example, if a group of personnel misjudges the position of the target object due to an occlusion at a certain acquisition point, and the other group of personnel is not interfered by this when collecting data from the opposite direction, this problem can be discovered through data comparison.

[0122] Randomly select the data of the acquisition points for similarity comparison. When the number of similarities greater than the preset value reaches the set standard (m, m is a positive integer, which is proportional to the area size of the acquisition area), it indicates that the collected data has a certain degree of consistency and accuracy and can be processed subsequently. If the similarity is insufficient, re-collect the data of the acquisition points with low similarity, ensuring that the finally used data can accurately reflect the actual situation of the acquisition area. This avoids the entry of incorrect or large-deviation data into the subsequent processing links, affecting the construction accuracy of the real-scene three-dimensional model, ensuring the integrity and reliability of the data, and laying a foundation for constructing a high-quality real-scene three-dimensional map.

[0123] Although an additional step of collecting data and comparing and analyzing by a group of personnel is added, from the overall process, incorrect data can be discovered and corrected in a timely manner, avoiding repeated rework caused by incorrect data in the subsequent data processing. If a large amount of data problems are found only in the later stage of data processing, re-collecting and processing will consume more time and resources. And this pre-comparison mechanism can screen out the problem data in advance and re-collect them in a targeted manner, improving the efficiency of the entire data collection and processing process.

[0124] For the construction of a real - scene three - dimensional map, the credibility of data is of crucial importance. The process of two groups of personnel collecting data and strictly comparing it provides strong support for the accuracy of the data. The data verified in this way can enhance the credibility of the model when used to construct a real - scene three - dimensional model, making applications based on this model, such as urban planning and virtual display, more reliable and practical.

[0125] The specific process of collecting the actual three - dimensional construction information of the collection site using the third collection method is as follows:

[0126] Step 1): Tool preparation: Select a drone equipped with an optical camera and a laser ranging module, and at the same time prepare ground control point markers.

[0127] Step 2): Data collection: Control the drone to fly automatically along a preset route. During the flight, the optical camera automatically takes orthophotos at preset time intervals.

[0128] At the same time, the laser ranging module measures the vertical distance to the ground every preset distance and records the relative height information.

[0129] At the collection points set in the collection site, manually control the drone to hover and take close - up images to supplement detailed information, and place control point markers at the corresponding positions on the ground and record their coordinate information.

[0130] Step 3): Data processing: Import the images and ranging data collected by the drone into professional software, use the ground control points for geometric correction of the images, correct the images to the true geographic coordinate system, then extract feature points from the orthophotos through photogrammetry algorithms to generate sparse point cloud data, optimize the sparse point cloud in combination with the laser ranging data to increase the point cloud density, screen and process the multi - angle close - up images, and extract the useful texture information from them, that is, obtain the actual three - dimensional construction information.

[0131] The drone is equipped with an optical camera and a laser ranging module for data collection. The optical camera takes orthophotos at preset time intervals, recording the planar visual information of the collection site, including the shape, color, and texture of ground objects, etc.; the laser ranging module measures the vertical distance to the ground to obtain relative height information. In addition, at the collection points, manually control the drone to hover and take close - up images to supplement detailed information. These multi - dimensional data complement each other, providing a rich information basis for constructing an accurate real - scene three - dimensional model and can more comprehensively and realistically reflect the actual situation of the collection site.

[0132] The drone has flexible maneuverability, is not restricted by ground terrain and obstacles, and can easily reach areas that are difficult for people to access or where traditional collection equipment cannot reach, such as mountainous areas, rivers, large areas of forests, etc. For the collection of large areas, it can fly automatically according to the preset route, quickly cover the target area, and improve the collection efficiency. Compared with other collection methods, it has significant advantages in dealing with complex terrains and large-scale collection tasks, greatly expanding the collection scope.

[0133] Using ground control point markers for geometric correction of images, the images are corrected to the true geographic coordinate system, ensuring the geographical location accuracy of the collected data. This enables the constructed real-scene three-dimensional model to be accurately matched with the actual geographical environment, which is of great significance in application scenarios that require precise geographical positioning, such as geographical information system (GIS) analysis, urban planning, etc., enhancing the practicality and application value of the data.

[0134] The data processing process uses photogrammetry algorithms to extract feature points from orthophotos to generate sparse point cloud data, combines lidar data to optimize the point cloud density, and simultaneously screens and processes multi-angle close-up images to extract useful texture information. This scientific data processing flow makes full use of the advantages of various types of collected data, can effectively improve the data quality, and construct a real-scene three-dimensional model with rich details and high precision to meet the needs of different industries for high-precision real-scene three-dimensional data.

[0135] During the process of using the third collection method to collect the actual three-dimensional construction information of the collection site, the drone used for data collection was intelligently monitored, and its intelligent monitoring content is as follows:

[0136] Through the sensors of the flight control system built into the drone, the flight altitude, flight speed, heading angle, pitch angle, and roll angle of the drone are obtained in real time;

[0137] When any one of the parameters of the flight altitude, flight speed, heading angle, pitch angle, and roll angle of the drone exceeds its corresponding standard value, the drone automatically starts the return flight program;

[0138] At the same time, the battery power and motor speed of the drone are monitored;

[0139] When the battery power is lower than the preset safe return flight power threshold, the drone automatically starts the return flight program;

[0140] Moreover, if it is detected that the speed of a certain motor fluctuates abnormally or the difference in speed from other motors exceeds the preset range, the system will issue an alarm and attempt to automatically adjust the flight attitude to maintain stable flight. If the speed of a certain motor fluctuates abnormally or the difference in speed from other motors exceeds the preset range and exceeds the preset duration, the drone automatically starts the return flight program.

[0141] During the flight of the drone, it regularly conducts self-checks on the carried optical camera and simple laser ranging module. When the self-check is abnormal, it restarts the device or adjusts the flight mission, and checks whether the camera focus is normal, whether the image sensor has faults, and whether the writing speed of the memory card meets the requirements, etc. For the laser ranging module, monitor its laser emission power, receiving sensitivity, and data transmission stability. If it is found that the test images taken by the camera are blurred, the noise increases abnormally, or there is a continuous packet loss phenomenon in the laser ranging data, prompt the operator in the ground control terminal in time so that corresponding measures can be taken;

[0142] The drone is equipped with a small meteorological sensor to real-time sense and collect meteorological information of the collection site, including wind speed, wind direction, air temperature, and air pressure;

[0143] When the wind speed exceeds the upper limit of the safe flight wind speed of the drone (generally 10 - 15 m / s for consumer drones, depending on the specific model), the system will issue a strong wind warning, reminding the operator to operate carefully or suspend the flight mission to avoid the drone getting out of control due to strong winds;

[0144] According to the air temperature and air pressure data, estimate the air density, and then automatically adjust the flight parameters of the drone, such as the lift coefficient, etc., to ensure flight stability;

[0145] Ensure the safe flight of the drone: Real-time monitor the flight altitude, speed, heading angle, pitch angle, and roll angle through the sensors of the flight control system. When exceeding the standard value, it will automatically return to the home position, which can prevent the drone from crashing due to attitude out of control. Monitor the battery power. When it is lower than the safety threshold, it will automatically return to the home position to avoid losing the drone due to power exhaustion. The monitoring of the motor speed and the corresponding processing mechanism can adjust the flight attitude or return to the home position in time when the motor fails, ensuring the safety of the drone to the greatest extent and reducing the risk of equipment loss.

[0146] Ensure the quality of the collected data: Regularly conduct self-checks on the optical camera and laser ranging module, which can timely detect and solve problems such as abnormal camera focus, image sensor faults, memory card writing problems, and faults in aspects such as the emission power, receiving sensitivity, and data transmission of the laser ranging module. Once it is found that the test images are blurred, the noise increases, or there is packet loss in the laser ranging data, prompt the operator to handle it in time to ensure that the collected data is clear and accurate, providing a reliable data basis for the construction of real scene three-dimensional models.

[0147] Improve the stability and efficiency of the collection work: Equipped with a meteorological sensor to obtain wind speed, wind direction, air temperature, and air pressure information in real time. When the wind speed exceeds the safety upper limit, a warning is issued to avoid strong wind interference with the collection or even causing the drone to get out of control, ensuring the smooth progress of the collection work. Automatically adjust the flight parameters according to the air temperature and air pressure to maintain flight stability, reduce the flight attitude adjustment time caused by environmental factors, improve the collection efficiency, and ensure the completion of the collection task on time.

[0148] Enhance the automation and intelligence level of the system: The entire intelligent monitoring process has a high degree of automation, and there is no need for operators to manually monitor various parameters at all times. The system can automatically judge the state of the drone, the operation of the equipment, and the environmental conditions, and make corresponding responses, such as automatically returning, adjusting the flight attitude, restarting the equipment, prompting the operator, etc., reducing the work intensity and operation difficulty of the operator, and improving the intelligence level of the acquisition system.

[0149] After obtaining the real-scene three-dimensional construction information, conduct a verification of the real-scene three-dimensional construction information. If the verification is passed, export it for real-scene three-dimensional map construction. The specific verification process is as follows:

[0150] Data integrity verification Point cloud data integrity check: Check the coverage of the point cloud data in the acquisition area to ensure that there is no large-area data loss. For example, check whether there is blank point cloud data in some areas due to equipment failure, occlusion, etc. Use the visualization function of point cloud processing software to visually observe the point cloud distribution. If obvious holes or sparse areas are found, mark and record them, and further investigate the reasons.

[0151] Image data integrity check: Verify whether the images taken by the panoramic camera and the optical camera cover all parts of the acquisition area. Check whether the image sequence is continuous and whether there is any omission in the shooting of key areas. Check the timestamp information of the images to judge the continuity of shooting. If there is a large time interval or missing images, it is regarded as incomplete data.

[0152] Geometric accuracy verification Point cloud data accuracy verification: Compare the point cloud data with known high-precision measurement data (such as the accurate coordinates of ground control points). Calculate the deviation between the coordinates of the feature points in the point cloud data and the actual coordinates. If the deviation is within the preset accuracy range, the geometric accuracy of the point cloud data meets the standard; if it exceeds the range, analyze the reasons for the error, such as calibration error of the lidar, motion error during the acquisition process, etc., and make corresponding adjustments or re-acquire.

[0153] Geometric correction check of image data: Check whether the images have been correctly corrected for distortion and geometry. By comparing the corrected images with a standard map or known georeference data, check whether the shapes and positions of the ground objects are accurate. For images with geometric deformation, re-correct them to ensure that the ground objects in the images are consistent with the actual geographical space position.

[0154] Data consistency verification: Check the matching degree between the point cloud and the image. Based on information fusion, check whether the texture information of the point cloud data matches the image. Randomly select multiple point cloud positions and check whether the texture details at the corresponding positions in the image match the surface features reflected by the point cloud. If the point cloud shows a building wall, but the image texture shows vegetation, it indicates that there is a problem with data consistency, and it is necessary to recheck the fusion algorithm or the accuracy of the collected data.

[0155] Data consistency check for different acquisition devices: For data obtained from multiple acquisition devices (such as lidar and cameras), compare their measurement results for the same ground object. For example, check whether the building height measured by lidar is consistent with the height measured or estimated through images. If the difference is large, analyze the reasons and make corrections.

[0156] Logical rationality verification: Check whether the spatial relationships between ground objects in the real scene 3D construction information are logical. For example, buildings should be located above the ground, and vegetation should grow in suitable soil areas. Use spatial analysis algorithms to check the spatial topological relationships of ground objects. If unreasonable spatial layouts are found (such as buildings floating in the air, roads passing through buildings, etc.), perform manual intervention and correction or re-collect relevant data.

[0157] Logical check of data change trends: Observe whether the change trends of data in space and time are reasonable. For continuously collected data, check whether the changes in terrain and ground objects conform to natural laws or actual situations. If unreasonable drastic changes in the terrain of a certain area are found in a short period of time, and factors such as actual engineering construction are excluded, it indicates that the data may be incorrect and further verification is required.

[0158] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the quantity of the indicated technical features. Thus, features defined with "first" and "second" may explicitly or implicitly include at least one such feature. In the description of the present invention, "a plurality" means at least two, such as two, three, etc., unless otherwise specifically defined.

[0159] In the description of this specification, the description referring to terms such as "one embodiment", "some embodiments", "examples", "specific examples", or "some examples", etc. means that the specific features, structures, materials, or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, without contradiction, those skilled in the art can combine and combine the different embodiments or examples described in this specification and the features of different embodiments or examples.

[0160] Although the embodiments of the present invention have been shown and described above, it can be understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those of ordinary skill in the art can make changes, modifications, substitutions, and variations to the above embodiments within the scope of the present invention.

Claims

1. A comprehensive modeling method for real-scene three-dimensional maps with intelligent collection, characterized in that: The following steps are involved: Step 1: Collect relevant information of the target location, and process the relevant information of the target location to obtain the evaluation information of the collection location; Step 2: The collection site assessment information includes Class I collection sites, Class II collection sites, and Class III collection sites; Step 3: When the collection site is a Class I collection site, use the first collection method to collect real-scene three-dimensional construction information of the collection site; Step 4: When the collection site is a Class II collection site, the second collection method is used to collect actual three-dimensional construction information of the collection site; Step 5: When the collection site is a Class III collection site, use the third collection method to collect actual three-dimensional construction information of the collection site; Step 6: After obtaining the real-scene 3D construction information, verify the real-scene 3D construction information, and export it to construct a real-scene 3D map.

2. The method for integrated modeling of real-scene three-dimensional maps collected by intelligent means according to claim 1, characterized in that: The specific process of processing the relevant information of the target location to obtain the collection location assessment information is as follows: Extracting the relevant information of the target location, the relevant information of the target location includes the area information of the target location, the number of buildings in the target location and the vegetation coverage area information of the target location; Processing the target land area information, the target land building quantity information and the target land vegetation coverage area information to obtain allocation parameters; When the allocation parameter is greater than the preset value a1, it is evaluated as a Class I collection site; When the allocation parameter is between the preset values ​​a1 and a2, it is assessed as a Class II collection site; When the allocation parameter is less than the preset value to a2, it is evaluated as a Class III collection site; a2<a1.

3. The method for integrated modeling of real-scene three-dimensional maps with intelligent collection according to claim 2 is characterized in that: The process of obtaining the allocation parameters is as follows: Assume that the area of ​​the target site is S, the number of buildings in the target site is B, and the vegetation coverage area of ​​the target site is V; The target area information S is given a weight of ω1, the target building quantity information B is given a weight of ω2, and the target vegetation coverage area information V is given a weight of w3, and w1+w2+w3=1; By formula Get the allocation parameter P; Among them, S max , B max 、V max They are respectively the maximum values ​​that may appear in the target area, number of buildings and vegetation coverage area within the target area.

4. The method for integrated modeling of real-scene three-dimensional maps collected intelligently according to claim 1 is characterized in that: The specific process of using the first acquisition method to collect the real scene three-dimensional construction information of the collection site is as follows: Step 1: Equipment deployment: Select a LiDAR scanner, install the LiDAR on a mobile acquisition platform, and simultaneously deploy multiple high-resolution panoramic cameras distributed around the acquisition platform. Calibrate the camera parameters in advance and set them to the preset mode. Step 2: Data collection: The collection vehicle travels along the predetermined route at a preset speed. The laser radar emits a laser beam in real time, measures the time and intensity of the reflected light, and accurately calculates the distance, direction and surface feature signals of the target object to form point cloud data. The panoramic cameras work synchronously and continuously capture images of the surrounding environment. The captured images not only contain the visual appearance information of the collection site, but also include the color and texture of the buildings, the types and forms of vegetation, and also record environmental details such as the lighting conditions and weather conditions at the time; Step 3: Data preprocessing: After the acquisition is completed, the point cloud data obtained by the lidar is transferred to professional processing software. First, denoising is performed, and the algorithm is used to identify and remove abnormal points caused by environmental interference. Then, point cloud registration is performed to unify the point cloud data collected at different positions and different perspectives into the same coordinate system. The overlapping parts of adjacent point cloud data are found through feature matching algorithms, and the relative position relationship between them is calculated, so that the scattered point clouds are spliced ​​into a complete three-dimensional scene. The images taken by the panoramic camera are also imported into the image processing software for preliminary processing to correct the image distortion and eliminate the image deformation caused by the optical characteristics of the camera lens; Automatically stitch images according to the shooting time sequence and pre-set overlapping areas; Step 4: Information fusion: Fuse the processed point cloud data with the panoramic image sequence, use the image texture information to supplement the appearance details of the point cloud data, and use the preset algorithm to map the color and texture in the image to the point cloud surface according to the three-dimensional spatial position reflected by the point cloud to obtain the real-scene three-dimensional construction information.

5. The method for integrated modeling of real-scene three-dimensional maps collected intelligently according to claim 1 is characterized in that: The specific process of using the second acquisition method to acquire the actual three-dimensional construction information of the acquisition site is as follows: Step (1): Equipment selection: Use a handheld 3D laser scanner with a portable panoramic gimbal camera; Step (2): Data collection: The data collector holds a 3D laser scanner and moves at a preset speed along a pre-planned collection path to scan the collection site in detail, while operating the panoramic gimbal camera simultaneously and stopping at each collection point; Step (3): Data processing: The point cloud data collected by the handheld laser scanner is imported into the computer software, and firstly manually screened to remove invalid data caused by operator occlusion and misoperation. Then, the point cloud data collected at different positions are spliced ​​using a feature-based registration algorithm to form a partially complete 3D model fragment. For panoramic images, fast stitching and correction are performed in the software, and the image EXIF ​​information is used to assist in stitching. After that, the processed panoramic image is preliminarily associated with the corresponding point cloud model fragments to obtain the real-scene three-dimensional construction information.

6. The method for integrated modeling of real-scene three-dimensional maps with intelligent collection according to claim 5 is characterized in that: The personnel in step (1) are divided into two groups, and the starting point and the end point of the pre-planned collection path are extracted; Set a group of people at the starting point to collect data from the starting point to the end point, and mark the data collected by the group of people at each collection point as Hi, where i is the number of collection points; Set up a group of personnel at the end point to collect data in the direction from the end point to the starting point, and mark the data collected by the group of personnel at each collection point as Ui, where i is the number of collection points; Before the data processing in step (3), the data Hi and Ui are compared and analyzed; Randomly select data from at least m collection points from Hi and mark them as Ym; Then select the data of the collection point at the position corresponding to the m randomly selected collection points in Hi from Ui and mark it as Em; Compare Em and Ym for similarity. When the number of similarities between Em and Ym is greater than the preset value, the subsequent data processing is directly performed. When the number of collection points whose similarities in Em and Ym are greater than a preset value is less than m, data collection of collection points whose similarities are less than the preset value is performed again.

7. The method for integrated modeling of real-scene three-dimensional maps with intelligent collection according to claim 1 is characterized in that: The specific process of using the third acquisition method to acquire the actual three-dimensional construction information of the acquisition site is as follows: Step 1): Tool preparation: Select a drone that carries an optical camera and a laser ranging module, and prepare ground control point markers; Step 2): Data collection: Control the drone to automatically fly along the preset route. During the flight, the optical camera automatically captures orthophotos at preset time intervals; At the same time, the laser ranging module measures the vertical distance to the ground every preset distance and records the relative height information; At the collection point set up at the collection site, manually control the drone to hover, take close-up images, supplement detailed information, and place control point markers at corresponding locations on the ground to record their coordinate information; Step 3): Data processing: Import the images and ranging data collected by the drone into professional software, use ground control points to perform geometric correction of the images, correct the images to the real geographic coordinate system, and then use photogrammetry algorithms to extract feature points from the orthophotos to generate sparse point cloud data. Combined with laser ranging data, the sparse point cloud is optimized to increase the point cloud density, and multi-angle close-up images are screened and processed to extract useful texture information, that is, to obtain actual 3D construction information.

8. The method for integrated modeling of real-scene three-dimensional maps with intelligent collection according to claim 7 is characterized in that: In the process of collecting the actual 3D construction information of the collection site using the third collection method, the drone used for data collection was intelligently monitored. The contents of the intelligent monitoring are as follows: The built-in flight control system sensor of the drone can obtain the drone's flight altitude, flight speed, heading angle, pitch angle and roll angle in real time; When any of the parameters of the drone's flight altitude, flight speed, heading angle, pitch angle, and roll angle exceeds its corresponding standard value, the drone automatically starts the return program; At the same time, the battery power and motor speed of the drone are monitored; When the battery power is lower than the preset safe return power threshold, the drone automatically starts the return procedure; In addition, if the motor speed is detected to fluctuate abnormally or the speed difference with other motors exceeds the preset range, the system will sound an alarm and try to automatically adjust the flight attitude to maintain stable flight. If the motor speed fluctuates abnormally or the speed difference with other motors exceeds the preset range and exceeds the preset time, the drone will automatically start the return procedure; During the flight, the drone will periodically perform self-checks on the optical camera and simple laser ranging module it carries. If the self-check is abnormal, the device will be restarted or the flight mission will be adjusted. The drone is equipped with a small meteorological sensor to sense the meteorological information of the collection site in real time, including wind speed, wind direction, temperature and air pressure; When the wind speed exceeds the upper limit of the safe flight speed of the drone, the system will issue a strong wind warning to remind the operator to operate with caution or suspend the flight mission; Based on the temperature and air pressure data, the air density is estimated and the flight parameters of the drone are automatically adjusted.

Images