An automated point cloud precision correction survey vehicle

Abstract

This invention specifically relates to an automated point cloud accuracy correction mapping vehicle. The invention includes an autonomously cruising unmanned mapping vehicle, a self-stabilizing platform, a prism-tracking automatic measurement robot, a high-precision positioning module, a dual-redundant wireless communication module, and an onboard controller. It is equipped with a mapping drone carrying a fixed collaborative prism and an airborne lidar, constructing an integrated air-ground fully automated mapping correction system. The unmanned mapping vehicle can achieve autonomous cruising across the entire domain, intelligent obstacle avoidance, and path planning. The self-stabilizing platform mounted on top of the vehicle dynamically compensates for attitude deviations caused by vehicle bumps, tilts, and swaying, ensuring the stability of the measurement benchmark throughout the process. The fully trackable automatic measurement robot, fixed above the self-stabilizing platform, uses machine vision recognition and servo tracking algorithms to track and aim at the collaborative prism mounted on the mapping drone in real time and at high frequency, dynamically acquiring high-precision spatial true coordinate data.

Description

Technical Field

[0001] This invention belongs to the field of air-ground collaborative measurement and three-dimensional laser point cloud accuracy correction technology in building engineering, and specifically relates to an automated point cloud accuracy correction surveying vehicle. Background Technology

[0002] UAV-based lidar 3D mapping is currently the mainstream technology in engineering site and spatial mapping. With its advantages of fast operation speed, wide coverage, high spatial resolution, and strong terrain adaptability, it can quickly complete the acquisition of 3D laser point cloud data over large areas, and is widely used in important scenarios such as full-area topographic mapping, building construction completion acceptance, ecological space surveys, and national land space censuses. However, existing traditional UAV-based lidar mapping technology has significant inherent technical shortcomings in practical engineering applications, severely restricting the accuracy of mapping results and their ability to be digitized.

[0003] Specifically, during drone flight operations, multiple interference factors, including satellite GPS positioning drift, high-frequency aerial attitude jitter, external airflow disturbances, installation deviations of airborne equipment, and multipath effects, result in raw laser point cloud data collected by airborne lidar generally exhibiting quality defects such as overall coordinate offset, systematic elevation deviation, local point cloud distortion, and numerous discrete noise points. The raw point cloud accuracy is low, only sufficient for low-precision terrain visualization, and fails to meet the stringent accuracy standards of engineering precision measurement and high-precision reverse modeling. Furthermore, it cannot support the construction, verification, and standardized engineering delivery of compliant and usable high-precision BIM models. BIM models generated from low-precision distorted point clouds are highly susceptible to problems such as component size distortion, spatial misalignment, structural surface deformation, and inaccurate elevation data. This directly renders the model unusable for construction guidance, quantity calculation, final acceptance, and digital operation and maintenance, severely hindering the implementation of smart construction and digital delivery.

[0004] Currently, the mainstream method for correcting point cloud errors in RTK drones relies on traditional manual operations. This involves manually deploying a large number of ground control points and targets, collecting the true coordinates of these control points using handheld RTK devices, and then using third-party software for offline registration and correction of the raw point cloud data. This manual correction method suffers from several unavoidable industry pain points and technical shortcomings: First, the manual deployment of control points is cumbersome, time-consuming, labor-intensive, and extremely inefficient. In scenarios involving long coastlines, large-scale fieldwork, and extensive land surveying, the workload is extremely heavy, and the project costs are high. Second, due to the limited workload, the density and sparse distribution of control points are limited, allowing only coarse registration and correction in localized areas. It cannot eliminate minute distortions and systemic global biases across the entire area, and discrete target points are insufficient to address the local distortion noise in massive point clouds. The first problem is that while fine-grained correction is achieved, the overall point cloud correction effect is poor and the accuracy uniformity is insufficient. The second problem is that the manual selection, layout, and data collection process is prone to human error, which further amplifies the original point cloud error. In addition, manual layout on steep slopes and dangerous areas in the field is difficult and poses a high risk of safety, resulting in a large number of blind spots in the surveying. The third problem is that this mode only supports offline correction after the fact, and cannot realize the simultaneous operation of surveying and accuracy correction. The data processing is lagging behind and cannot meet the needs of modern automated, real-time, and unmanned surveying operations.

[0005] Meanwhile, the existing air-ground collaborative mapping equipment system suffers from insufficient supporting systems, and generally suffers from problems such as asynchronous air-ground data timing, inconsistent benchmarks, and poor dynamic tracking stability. High data transmission latency and inaccurate timestamp matching between UAVs and ground mapping equipment easily lead to timing misalignment errors. Traditional ground mapping equipment lacks adaptive leveling capabilities for vehicle attitude, resulting in unstable measurement benchmarks under conditions of road bumps and vehicle tilting, making it impossible to achieve high-precision, continuous, and stable tracking and aiming of moving UAV prisms. This makes it difficult to obtain reliable, high-frequency ground truth benchmark data, further leading to inaccurate point cloud error calculation and limited correction accuracy.

[0006] Furthermore, existing conventional vehicle-mounted mobile surveying equipment has limited functionality, capable only of acquiring point cloud data for single ground scenarios. It generally lacks core functions such as automatic tracking of aerial UAV collaborative prisms, real-time acquisition of dynamic spatial ground truth, and air-ground collaborative layered point cloud accuracy correction, making it impossible to construct an integrated air-ground surveying and correction system. Simultaneously, traditional total stations with simple gimbal structures lack dynamic attitude compensation capabilities for vehicle bumps and swaying, only meeting static fixed-point measurement needs. Under dynamic operating conditions with moving vehicles, they cannot continuously and stably aim at high-speed moving UAV collaborative prisms, making it difficult to achieve full-domain, continuous, and high-frequency dynamic ground truth data acquisition. In summary, the industry has consistently failed to achieve fully automated, high-precision surveying throughout the entire process of "automatic arrival - automatic centering - automatic calibration - automatic tracking - real-time correction - automatic re-measurement," and lacks truly mature dedicated equipment for air-ground collaborative point cloud accuracy correction. This has become a key technological bottleneck restricting the large-scale application of UAV high-precision surveying technology in engineering BIM delivery, precision terrain monitoring, and large-scale site digital surveying. Summary of the Invention

[0007] This invention aims to address the accuracy deficiencies of existing UAV laser point cloud mapping, such as coordinate deviation, geometric distortion, and noise interference, as well as the technical problems of low efficiency, high difficulty in high-risk scenarios, and inability to achieve fully automated real-time collaborative correction using traditional manual control points. It provides an automated point cloud accuracy correction mapping vehicle and its supporting accuracy correction method. This invention abandons the manual deployment mode, relying on autonomous vehicle navigation, dynamic attitude stabilization, automatic UAV prism tracking, air-to-ground data time-series synchronization, and layered progressive point cloud correction technology to achieve fully automated air-to-ground collaborative mapping, high-frequency truth acquisition, and high-precision correction of the entire point cloud domain. It effectively solves industry pain points such as UAV point cloud positioning drift, elevation distortion, and local deformation, providing efficient, unmanned, and high-precision technical support for precision engineering surveying, terrain monitoring, and standardized BIM model delivery.

[0008] The technical solution adopted in the method of this invention is as follows:

[0009] An automated point cloud accuracy correction mapping vehicle includes an unmanned mapping vehicle body, a self-stabilizing platform fixedly installed on the top of the unmanned mapping vehicle body, an automatic measurement robot and a laser ranging system installed on the bearing end of the self-stabilizing platform, and an integrated vehicle controller, a high-precision positioning module, a laser radar module and a wireless communication module inside the unmanned mapping vehicle body; an automatic ground measurement point calibration system is installed inside the unmanned mapping vehicle body.

[0010] The surveying vehicle is equipped with a surveying drone. The surveying drone is equipped with a collaborative prism and an airborne lidar to collect raw laser point cloud data, as well as a wireless communication module to maintain communication with the surveying vehicle.

[0011] The surveying vehicle is equipped with a ground control station, which integrates a 5G communication module, an unmanned equipment remote control module, and a real-time data transmission module. The ground control station establishes a two-way wireless communication link with the unmanned surveying vehicle and the surveying drone, respectively, for remote manual / automatic collaborative control of the unmanned surveying vehicle and the surveying drone's operating trajectory, measurement attitude, and surveying parameters. At the same time, it receives, summarizes, and stores high-precision true coordinate data collected by the unmanned surveying vehicle and raw point cloud data collected by the drone's airborne lidar in real time, realizing real-time transmission, synchronous interaction, and remote monitoring of air-to-ground surveying data. It also works with the vehicle-mounted controller to complete remote accuracy correction and data optimization processing of the point cloud data.

[0012] The vehicle controller is configured to perform the following steps:

[0013] The satellite positioning system drives the unmanned vehicle to the benchmark measurement point. Image recognition and laser calibration drive the unmanned vehicle to move slightly, so that the laser point of the measurement robot can be aligned with the benchmark measurement point. Then, the image recognition algorithm captures the aerial mapping drone and the collaborative prism target in real time, and drives the self-stabilizing platform to dynamically level and compensate for angles, so as to maintain the stability of the automatic measurement robot's posture.

[0014] The automatic measurement robot is controlled to automatically track and aim at the collaborative prism, and high-precision three-dimensional true coordinate data of the prism is obtained in real time.

[0015] The system receives raw laser point cloud data transmitted back by a surveying drone, constructs an error correction model by combining the measured true coordinates of the prism, performs global deviation correction and local distortion correction on the raw laser point cloud data, and outputs high-precision corrected 3D point cloud data.

[0016] As a further improvement of the present invention, the self-stabilizing platform is a three-axis anti-shake self-stabilizing structure, with built-in attitude sensors, tilt angle detection modules, vehicle speed detection modules, and servo adjustment motors. It collects real-time data on the unmanned surveying vehicle's driving speed, measurement robot height, tilt angle, and overall vehicle status, including posture sway. Relying on the built-in dynamic attitude adaptive adjustment algorithm, it automatically adjusts and compensates for pitch, roll, and yaw angles in real time according to different driving conditions and changes in vehicle posture. It dynamically calibrates the platform's level state, ensuring that the automatic measurement robot mounted on the platform maintains an absolutely level measurement posture throughout all operating conditions, including unmanned vehicle start-stop, constant speed driving, and bumpy driving. At the same time, it collects and outputs accurate height and spatial posture data of the measurement robot in real time, ensuring the stability and accuracy of the measurement benchmark.

[0017] As a further improvement of the present invention, the automatic calibration system for ground measuring points adopts a ground fixed measuring point tracking mechanism, which has a built-in laser grating sensor and camera module. During operation, it is located at the bottom of the unmanned mapping vehicle. The camera module identifies the position of the ground measuring points, the laser grating grid covers the measuring point area, the horizontal deviation between the reference point of the measuring robot and the measuring points is calculated, and the deviation data is uploaded to the vehicle controller. The vehicle controller drives the unmanned mapping vehicle and the self-stabilizing platform to automatically complete the alignment calibration with the reference point.

[0018] As a further improvement of the present invention, the automatic measurement robot integrates a high-precision total station, a high-definition visual recognition camera, a laser rangefinder, and a servo tracking mechanism; the high-definition visual recognition camera is used to acquire panoramic images of the airspace in real time, accurately identify, lock, and dynamically position the spatial position of the surveying drone and the collaborative prism; the servo tracking mechanism works in conjunction with the self-stabilizing platform to achieve 360° horizontal and ±90° pitch rotation without blind spots, completing fully automatic tracking and precise aiming of the collaborative prism in both static and moving states, ensuring the continuity and high precision of the static and dynamic measurement processes.

[0019] As a further improvement of the present invention, the collaborative prism adopts a high-reflectivity 540° omnidirectional prism, which has an ultra-large angle reflection recognition range and high reflectivity, effectively improving the recognition efficiency and recognition accuracy of the high-definition visual recognition camera, and is suitable for high-precision dynamic tracking operations with lasers; the collaborative prism and the UAV-borne lidar adopt a fixed integrated layout, and their relative positions remain constant, constructing a unified and stable airborne mapping benchmark, avoiding measurement errors caused by relative displacement, and providing a reliable benchmark for subsequent point cloud data accuracy correction.

[0020] As a further improvement of the present invention, the high-precision positioning module includes a geodesic GPS / BeiDou dual-mode receiver and an inertial navigation unit, equipped with an RTK differential positioning fusion algorithm, which collects the unmanned surveying vehicle's position coordinates, driving attitude and motion parameters in real time and completes the real-time positioning of the vehicle; it dynamically analyzes and autonomously corrects positioning deviations by combining site environment and historical positioning data, reducing errors caused by terrain and satellite signal drift; relying on satellite differential and inertial navigation fusion compensation, it outputs centimeter-level positioning coordinates when the vehicle is moving and outputs millimeter-level high-precision vehicle positioning coordinates when stationary, providing high-precision reference coordinates for the calculation of measured data and the construction of a global point cloud error correction model;

[0021] The vehicle-mounted controller incorporates a point cloud layering correction algorithm, which specifically includes: collecting multiple sets of measured true coordinates of the prism at different airspace heights and horizontal positions and corresponding sampled coordinates of the UAV point cloud; calculating global coordinate offset, elevation deviation, and attitude distortion parameters; constructing a global linear correction model to complete the overall coordinate translation, rotation, and scaling correction of the point cloud; and filtering out abnormal point clouds by using a true residual threshold for local distortion points and noise points to complete local fine-grained correction.

[0022] As a further improvement of the present invention, the unmanned mapping vehicle is equipped with an autonomous navigation module and an obstacle avoidance sensor array, and also features a 360° panoramic LiDAR. The 360° panoramic LiDAR can scan the surrounding environment in real time, construct a high-precision real-time environmental model, and accurately identify obstacles, terrain undulations, and work boundaries. Combined with high-precision RTK millimeter-level positioning data and dynamic planning of the environmental model, the unmanned vehicle's mapping route is optimized. Simultaneously, the mapping path can be pre-set via an electronic map, providing dual assurance of the mapping path's rationality and accuracy. The onboard controller has built-in puncture point location storage and automatic location tracking. The system records the spatial coordinates, location numbers, and attitude information of all surveying points in real time during the operation, forming a local point database. During subsequent re-surveys and supplementary surveys, the system can automatically retrieve historical point data, combine real-time positioning and environmental models to autonomously plan the driving path, and combine the vehicle-mounted ground fixed measuring point tracking mechanism to accurately search for and automatically arrive at the preset location to complete the fixed-point re-survey and supplementary sampling operations. It can achieve fully automatic cruising, intelligent obstacle avoidance, accurate point finding and trajectory driving in complex sites, and, in conjunction with the UAV airspace surveying operation mode, efficiently complete the integrated operation of air-ground collaborative fully automatic point cloud acquisition and accuracy correction.

[0023] As a further improvement of the present invention, the automatic measurement robot is equipped with an automatic calibration module. Before each surveying operation, the device completes fully automatic self-check and accuracy calibration based on the preset ground reference points. By comparing the standard coordinates of the reference points with the robot's measured coordinates, it automatically calculates and compensates for multiple error sources such as the installation deviation of the self-stabilizing platform, the mechanical assembly error of the equipment, and the inherent measurement error of the system. This enables adaptive correction of the device's measurement posture and measurement parameters, effectively eliminating the cumulative error of the system. No manual calibration and debugging are required throughout the entire process, ensuring the robot's prism tracking response accuracy, dynamic aiming accuracy, and spatial true coordinate acquisition accuracy throughout the operation, providing a reliable measurement foundation for subsequent high-precision correction of point cloud data.

[0024] As a further improvement of the present invention, the wireless communication module adopts a 5G+wireless bridge dual-mode redundant communication architecture, which combines low-latency transmission and long-distance stable communication capabilities. During the operation of the unmanned surveying vehicle and the surveying drone, a high-precision timing module is independently equipped and records the data acquisition timestamp synchronously, so as to achieve a unified time reference for air and ground equipment. Relying on the complementary advantages of the dual communication modes, low-latency and high-reliability two-way data interaction between the unmanned vehicle and the drone is completed, and the true coordinate data of the prism collected by the automatic measurement robot and the raw point cloud data collected by the drone's airborne lidar are synchronized in real time. The timing error caused by the asynchronous data acquisition of air and ground equipment is offset by the timestamp matching correction algorithm, which reduces the time dimension deviation from the source, ensures the timing consistency and matching accuracy of the point cloud correction data, and provides accurate and synchronous raw data support for the global error correction model.

[0025] This invention also provides an automated point cloud accuracy correction method, applied to the aforementioned automated point cloud accuracy correction mapping vehicle, comprising the following steps: S1, job deployment and equipment self-check: The unmanned mapping vehicle autonomously travels to the target mapping area according to the preset mapping path and real-time environment model, and sequentially completes RTK millimeter-level positioning calibration, self-stabilizing platform horizontal attitude self-check, automatic measurement robot parameter initialization, and whole-machine communication link self-check to ensure that the reference parameters of each device are normal and the communication is stable; the ground control station simultaneously binds and connects the unmanned mapping vehicle and the mapping drone through 5G+wireless bridge dual communication mode to complete equipment status inspection and initial parameter synchronization; the mapping drone takes off to the preset mapping airspace, and relies on the airborne lidar and cooperative prism to perform full-area scanning and acquisition of the site to obtain the original three-dimensional laser point cloud data of the site;

[0026] S2. Measurement Point Acquisition and Automatic Calibration: The automatic calibration system for ground measurement points at the bottom of the vehicle body acquires real-time images of the ground below through a built-in camera module, identifies and captures fixed ground measurement points set up on the site; the system controls the laser grating sensor to project a grating grid to fully cover the measurement point area, calculates the horizontal deviation between the measurement robot's reference point and the ground measurement point, and uploads the deviation data to the vehicle controller in real time; the controller issues adjustment commands according to the feedback error, synchronously drives the entire unmanned surveying vehicle to move, and fine-tunes the attitude of the self-stabilizing gimbal platform, continuously correcting the position deviation in a closed loop until the measurement robot's reference point and the ground measurement point are precisely aligned, completing the automatic alignment calibration of the entire machine's measurement reference;

[0027] S3. Air-Ground Collaborative Dynamic Tracking and Measurement: The surveying drone carries a collaborative prism and takes off to conduct a full-domain scanning operation. The drone uses its own satellite positioning module to obtain the airspace position coordinates in real time and actively transmits the satellite position coordinate data to the unmanned surveying vehicle through a wireless communication link as a guidance signal. After receiving the drone's satellite position information, the unmanned surveying vehicle quickly locks the target area using the depth camera of the automatic measurement robot and the prism tracking function. It actively tracks and locks the collaborative prism carried by the drone and obtains the drone's high-precision spatial true position in real time through dynamic tracking and measurement of the prism. Relying on the real-time attitude compensation capability of the self-stabilizing platform, the pitch, roll, and yaw angles are dynamically adjusted. With the help of the servo tracking mechanism, the automatic measurement robot continuously tracks and aims at the center of the collaborative prism with high precision. The three-dimensional spatial true coordinates of the prism and the drone are collected in real time in a high-frequency sampling mode to form a high-precision ground true value dataset.

[0028] S4. Intelligent calculation of multi-dimensional error parameters: The vehicle controller completes air-to-ground data matching based on the synchronization timestamp, compares and analyzes the actual measured coordinates of the prism collected by the automatic measurement robot with the coordinate information of the corresponding position in the original point cloud data of the UAV, and intelligently calculates multi-dimensional error parameters such as global offset error, elevation system error, spatial angle distortion error and local point residual in the point cloud data, and constructs a dynamic error correction model adapted to the current work site.

[0029] S5. Hierarchical Point Cloud Accuracy Correction: Based on the established global error correction model, the original laser point cloud data is subjected to overall coordinate correction, which uniformly corrects the overall positioning drift and overall offset of the point cloud; then, the single point data is screened and identified by the preset local residual threshold, and the distorted noise, abnormal floating points and deviation exceeding the limit points are automatically removed, so as to realize the point-by-point correction, noise reduction optimization and local distortion repair of the original laser point cloud data, and complete the all-round and multi-level accuracy optimization of the point cloud data.

[0030] S6. Data storage and remote output: The high-precision 3D point cloud data after accuracy correction is stored locally with encryption, and simultaneously uploaded to the ground control station and surveying terminal in real time through dual redundant communication links, completing an integrated surveying and mapping operation process of fully automatic point cloud acquisition, error correction and data output in air-ground collaboration.

[0031] Compared with the prior art, the present invention has the following beneficial effects:

[0032] (1) Completely overturns the traditional manual surveying and correction mode, realizing unmanned operation throughout the entire process. This invention eliminates the need for manual entry to set up ground control points, manual target calibration and subsequent offline correction, completely getting rid of the dependence on manual point setting, station setting and calibration. The efficiency of field operations is increased by more than 90% compared with traditional manual operations. The entire work relies on unmanned vehicles for autonomous operation, automatic robot tracking and intelligent system correction, which greatly reduces the intensity of manual labor, completely avoids the safety risks of manual operations in water-adjacent, steep slope and high-risk areas in the field, and multiplies the efficiency of large-scale surveying and mapping operations.

[0033] (2) A unique three-axis, six-degree-of-freedom dynamic attitude compensation technology solves the problem of unstable dynamic measurement benchmarks. For the bumpy and tilting conditions of unmanned vehicles, a three-axis, six-degree-of-freedom self-stabilizing platform provides real-time attitude compensation and leveling, ensuring stable operation of the prism-tracking high-precision total station throughout the entire process. This completely solves the defects of traditional fixed measuring equipment, such as large dynamic aiming deviation, unstable tracking, and large fluctuations in true data, significantly improving the stability and accuracy of dynamic true value acquisition. The self-stabilizing platform, with vehicle speed feedforward, maintains a 2″ level horizontal accuracy and ±1mm elevation stability even at driving speed, doubling the leveling response speed. The 540° omnidirectional prism achieves blind-spot-free coverage of the upper hemisphere, ensuring no loss of lock when the UAV passes overhead. Tracking stability under dual dynamic conditions is 90% higher than traditional equipment, enabling continuous acquisition of reliable millimeter-level true data.

[0034] (3) A layered and progressive dual correction mechanism is adopted, and the accuracy of point cloud correction is comprehensively upgraded. Unlike traditional ICP registration without true value, this invention uses the millimeter-level true value of the total station as the core to build a global correction model. It completes the global system error correction by sampling true values ​​at multiple positions and heights, eliminating overall drift and elevation deviation. Then, it completes local noise removal and distortion repair by filtering the residual threshold. It comprehensively solves the technical pain points of UAV point cloud positioning drift, inaccurate elevation, local deformation, and redundant noise points. The accuracy, integrity, and uniformity of point cloud data are greatly improved, which can directly meet the requirements of high-precision BIM model modeling, verification and standardized engineering delivery.

[0035] (4) Real-time collaborative air-ground operation enables simultaneous completion of surveying and correction. Relying on the air-ground linkage mechanism of UAV satellite coarse positioning guidance + ground robot precise tracking, combined with dual-link low-latency data transmission and high-precision timestamp synchronization technology, real-time data acquisition, real-time matching and real-time correction are achieved without the need for offline processing, truly realizing integrated, real-time and intelligent surveying and mapping operations.

[0036] (5) It has wide scene adaptability and strong anti-interference ability, and is applicable to all-area surveying and mapping scenarios. It can be widely adapted to complex scenarios such as coastal tidal flat monitoring, strip highway surveying, mountain topographic mapping, large-scale construction site surveying, mine earthwork measurement, and land space survey. It is not limited by site topography, high-risk areas, or weak signal areas, and has extremely strong environmental adaptability and engineering practicality.

[0037] (6) Equipment self-calibration + data self-correction, system accuracy can be continuously optimized in a closed loop. The equipment is automatically calibrated at startup and the vehicle positioning is dynamically self-corrected. It can compensate for multiple error sources such as mechanical assembly error, installation deviation, and satellite signal drift in real time, forming a complete accuracy closed loop control system. The long-term operation has stable accuracy and high reliability.

[0038] (7) Intelligent retesting and supplementary testing capabilities to meet long-term monitoring needs. All measuring point information is automatically stored in the puncture point database. During subsequent retesting and supplementary testing, the system can navigate autonomously, find points automatically, and center the measurement automatically without manual operation. It is particularly suitable for scenarios such as foundation pits and slopes that require long-term millimeter-level deformation monitoring, which greatly improves the efficiency of periodic monitoring operations. Attached Figure Description

[0039] Figure 1 This is a schematic diagram of the overall hardware assembly structure of the automated point cloud accuracy correction mapping vehicle of the present invention;

[0040] Figure 2 This is a schematic diagram of the gimbal on the top of the unmanned surveying vehicle of the present invention;

[0041] Figure 3 This is a schematic diagram of the installation structure of the measuring robot of the present invention;

[0042] Figure 4 A schematic diagram of a structure for mounting a prism and lidar on a drone;

[0043] Figure 5 This is a schematic diagram of the air-ground collaborative mapping operation of the present invention;

[0044] Figure 6 This is a flowchart of the automated point cloud accuracy correction method of the present invention.

[0045] In the attached image:

[0046] 1. Unmanned surveying vehicle body; 12. Obstacle avoidance radar; 2. Prism tracking high-precision total station; 21. Top lidar; 22. Laser rangefinder; 23. Three-axis attitude sensor; 24. Bottom laser rangefinder; 25. Total station eyepiece; 3. Self-stabilizing platform; 4. High-definition recognition camera; 5. Vehicle-mounted lidar; 6. Vehicle-mounted RTK; 7. Control and communication box; 8. 360° rotating platform; 81. Under-vehicle high-definition camera; 9. Surveying UAV carrier; 91. Airborne lidar; 92. Airborne 540° prism; 10. Measured building; 101. Ground reference point. Detailed Implementation

[0047] The present invention will be further explained in detail below with reference to the accompanying drawings, so that those skilled in the art can better understand and implement the present invention. However, the following examples are only used to explain the present invention and are not intended to limit the present invention.

[0048] like Figure 1-5 As shown, an automated point cloud accuracy correction mapping vehicle includes an unmanned mapping vehicle body 1. A self-stabilizing platform 3 is fixedly installed on the top of the unmanned mapping vehicle body 1. A prism tracking high-precision total station 2, i.e., an automatic measurement robot, and a laser ranging system (including a bottom laser rangefinder 24) are installed on the bearing end of the self-stabilizing platform 3. The unmanned mapping vehicle body 1 integrates an on-board controller (located in the control and communication box 7), a high-precision positioning module (i.e., an on-board RTK 6), a laser radar module (including an on-board laser radar 5 and an obstacle avoidance radar 12), and a wireless communication module (located in the control and communication box 7). An automatic ground measurement point calibration system is installed on the bottom of the unmanned mapping vehicle body 1, including a 360° rotating platform 8 and a high-definition camera 81 under the vehicle.

[0049] This invention includes a mapping drone 9 and a ground control station, forming an integrated air-ground fully unmanned mapping system. The mapping drone 9 is equipped with a 540° collaborative prism 92 and an airborne lidar 91 to collect raw laser point cloud data, and is equipped with a wireless communication module to maintain communication with the mapping vehicle. The ground control station integrates a 5G communication module, an unmanned equipment remote control module, and a real-time data transmission module. It establishes bidirectional wireless communication links with the unmanned mapping vehicle 1 and the mapping drone 9, respectively, for remote manual / automatic collaborative control of the unmanned mapping vehicle and the mapping drone's operating trajectory, measurement attitude, and mapping parameters. Simultaneously, it receives, summarizes, and stores high-precision true coordinate data collected by the unmanned mapping vehicle and raw point cloud data collected by the drone's airborne lidar in real time, realizing real-time transmission, synchronous interaction, and remote monitoring of air-ground mapping data. It also works with the vehicle-mounted controller to perform remote accuracy correction and data optimization processing of the point cloud data.

[0050] The vehicle-mounted controller (located in the control and communication box 7) is configured to perform the following steps: the satellite positioning system drives the unmanned vehicle to the reference measurement point position; image recognition (high-definition camera 81 under the vehicle) and laser calibration drive the unmanned vehicle to move slightly, so that the total station laser point is aimed at the reference measurement point; then, the high-definition recognition camera 4 captures the aerial mapping UAV and the cooperative prism target in real time, and drives the self-stabilizing platform 3 to dynamically level and compensate for angles, so as to keep the attitude of the prism tracking high-precision total station 2 stable; control the prism tracking high-precision total station 2 to automatically track and aim at the 540° cooperative prism 92, and obtain the high-precision three-dimensional true coordinate data of the prism in real time; receive the original laser point cloud data transmitted back by the UAV, combine it with the measured true coordinates of the prism to construct an error correction model, perform global deviation correction and local distortion correction on the original laser point cloud data, and output the high-precision corrected three-dimensional point cloud data.

[0051] The self-stabilizing platform 3 features a three-axis anti-shake self-stabilizing structure, incorporating a three-axis attitude sensor 23, a tilt detection module, a vehicle speed detection module, and a servo adjustment motor. It can collect real-time data on the unmanned surveying vehicle's speed, total station height, tilt angle, and attitude sway, among other vehicle status data. Utilizing a built-in dynamic attitude adaptive adjustment algorithm, it can automatically and in real-time adjust and compensate for pitch, roll, and yaw angles based on different driving conditions and vehicle attitude changes, dynamically calibrating the platform's level. This ensures that the prism-tracking high-precision total station 2 mounted on the platform maintains an absolutely level measurement posture under all operating conditions, including unmanned vehicle start-stop, constant speed driving, and bumpy driving. Simultaneously, it can collect and output accurate height and spatial attitude data of the total station in real-time, guaranteeing the stability and accuracy of the measurement benchmark.

[0052] Specifically, the self-stabilizing platform 3 adopts a three-axis, six-DOF self-stabilizing platform, including a lower mounting flange, an upper bearing plate, three sets of high-precision servo electric cylinders arranged in a 120° equilateral triangle, and a dual-axis high-precision tilt sensor (resolution 0.1″) fixed on the upper bearing plate, a 200Hz sampling frequency MEMS inertial measurement unit, and a ±1mm precision pulsed laser ranging module (i.e., the bottom laser rangefinder 24) with the center vertically downward. The platform has a built-in vehicle speed detection module that can read the chassis wheel speed sensor data in real time and adopts an attitude-height-vehicle speed joint feedforward PID control algorithm: In terms of attitude control, the electric cylinders are driven to extend and retract differentially based on the tilt sensor and IMU data, and the ±15° range is compensated with a 100Hz response frequency. The platform exhibits dynamic leveling accuracy better than 2″ despite vehicle tilt. Regarding elevation stability, three electric cylinders are synchronously extended and retracted based on the vertical height data between the platform and the ground obtained from the laser ranging module. This compensates for vertical vibrations caused by potholes and bumps in the ground, controlling the vertical vibration amplitude within ±1mm. For speed feedforward, the control gain is dynamically adjusted according to real-time driving speed: high gain feedback ensures accuracy at low speeds (<5km / h), while increasing the feedforward weight at medium to high speeds (5-12km / h) to anticipate attitude disturbances caused by road bumps, reducing the leveling response time from the traditional 100ms to less than 50ms. Stable accuracy is maintained even at high speeds, breaking through the limitation of traditional self-stabilizing platforms that can only be used statically or at low speeds. Because the upper support plate remains horizontal, the laser ranging module fixed on it emits a beam that is always vertical, unaffected by vehicle tilt, allowing real-time acquisition of the true vertical distance between the platform and the ground, providing millimeter-level true values ​​for elevation correction and terrain warning.

[0053] The automatic calibration system for ground measuring points adopts a ground-fixed measuring point tracking mechanism, which is installed at the center vertical axis of the unmanned vehicle chassis. It includes a 360° rotating platform 8 and a vehicle-mounted high-definition camera 81, a crosshair laser grating projector, and a supplementary lighting unit. During operation, the vehicle-mounted high-definition camera 81 identifies the position of the ground measuring points, the laser grating grid covers the measuring point area, the horizontal deviation between the total station reference point and the measuring points is calculated, and the deviation data is uploaded to the controller. The controller drives the surveying unmanned vehicle and the self-stabilizing platform to automatically complete the alignment calibration with the reference point.

[0054] Specifically, the automatic calibration system for ground measuring points uses a 20-megapixel global shutter industrial camera (i.e., a high-definition camera 81 under the vehicle) and an 850nm crosshair laser grating projector to achieve fully automatic unmanned centering of the measurement benchmark: During operation, the industrial camera acquires ground images and uses a deep learning image recognition algorithm to locate the pre-embedded high-reflectivity benchmark marks on the ground. The laser grating projector projects a 100mm×100mm grid grating onto the ground, and calculates the horizontal deviation and height difference between the benchmark point and the vertical axis of the total station through grating deformation. After the deviation data is uploaded to the controller, the unmanned vehicle is first driven to make centimeter-level micro-adjustments, and then the self-stabilizing platform is used for millimeter-level fine adjustments. The entire centering process takes ≤30 seconds, and the final centering accuracy is better than 0.5mm, which is better than the accuracy of traditional manual centering (1-2mm). The entire process does not require manual operation, completely solving the pain points of low efficiency, large error, and inability to set up stations in dangerous areas by traditional mobile measuring equipment.

[0055] Meanwhile, the system is equipped with an automatic calibration mechanism upon startup: each time the equipment is powered on, the unmanned vehicle automatically navigates to the nearest known benchmark point, completes centering and leveling through the automatic calibration system of ground measuring points, measures the three-dimensional coordinates of the benchmark point and compares them with the known true values, automatically calculates the installation deviation of the self-stabilizing platform, the line-of-sight error of the total station, the index difference, and the systematic errors caused by mechanical wear, and writes the error parameters into the compensation file for automatic correction in subsequent measurements. The entire calibration process takes ≤2 minutes, requires no manual operation, ensures that the equipment has no cumulative error in long-term use, and does not require periodic return to the factory for calibration.

[0056] The Prism Tracking High-Precision Total Station 2 (Automatic Measurement Robot) integrates a high-precision total station body, a high-definition recognition camera 4, a laser rangefinder 22, and a servo tracking mechanism. The high-definition recognition camera 4 is used to acquire panoramic images of the airspace in real time, accurately identify, lock, and dynamically position the spatial position of the surveying UAV 9 and the 540° cooperative prism 92. The servo tracking mechanism works in conjunction with the three-axis six-degree-of-freedom self-stabilizing platform to achieve 360° horizontal and ±90° pitch rotation without blind spots, completing fully automatic tracking and precise aiming of the cooperative prism in both static and moving states, ensuring the continuity and high precision of static and dynamic measurement processes.

[0057] Specifically, the Prism Tracking High-Precision Total Station 2 employs a blind-spot-free automatic tracking measurement system, integrating 850nm infrared active illumination, a sub-pixel-level CCD vision module, and a dual-axis direct-drive servo turntable. Combined with a POS feedforward + vision-locked tracking algorithm, it can achieve lock-free tracking of UAVs within a 1500m range with a flight speed ≤18m / s, achieving an angle measurement accuracy better than 1″ and a distance measurement accuracy better than 1mm + 1ppm. It outputs the true 3D value of the prism at a 10Hz frequency. For temporary occlusion scenarios, it incorporates an occlusion prediction and rapid re-acquisition mechanism with a re-acquisition time ≤0.5s. During operation, it accurately locates the spatial position of the aerial mapping UAV and the collaborative prism through optimized image recognition and contour matching algorithms, achieving intelligent target locking and continuous tracking. The built-in laser rangefinder 22 and top-mounted lidar 21 can measure the spatial distance from the equipment to the collaborative prism in real-time and at high frequency, and calculate the high-precision 3D coordinates of the prism by combining angle observation data.

[0058] The Prism Tracking High-Precision Total Station 2 is equipped with an automatic calibration module. Before each surveying operation, the device can perform fully automatic self-checks and accuracy calibration based on preset ground benchmarks. By comparing the standard coordinates of the benchmarks with the measured coordinates of the total station, it automatically calculates and compensates for multiple error sources, such as installation deviations of the three-axis six-DOF self-stabilizing platform, mechanical assembly errors of the equipment, and inherent measurement errors of the system. This enables adaptive correction of the device's measurement posture and parameters, effectively eliminating accumulated system errors. No manual calibration or debugging is required throughout the entire process, ensuring the robot's prism tracking response accuracy, dynamic aiming accuracy, and spatial true coordinate acquisition accuracy throughout the operation. This provides a reliable measurement foundation for subsequent high-precision correction of point cloud data.

[0059] The 540° collaborative prism 92 adopts a high-reflectivity 540° omnidirectional prism, which has an ultra-large angle reflection recognition range and high reflectivity characteristics. It can effectively improve the recognition efficiency and recognition accuracy of the high-definition recognition camera 4 and is suitable for high-precision dynamic laser tracking operations. The 540° collaborative prism 92 and the UAV airborne lidar 91 adopt a fixed integrated layout, and their relative positions remain constant, thus constructing a unified and stable airborne mapping benchmark, avoiding measurement errors caused by relative displacement, and providing a reliable benchmark for subsequent point cloud data accuracy correction.

[0060] Specifically, the 540° collaborative prism 92 uses high-reflectivity glass with a reflectivity of 92%, significantly improving visual recognition efficiency and laser tracking accuracy in complex airspace environments. The 540° omnidirectional prism achieves blind-spot-free coverage of the upper hemisphere, ensuring lock-on even when the drone is overhead. Tracking stability under dual dynamic conditions is 90% higher than traditional equipment, and reliable millimeter-level true-value data can be continuously acquired. The collaborative prism and the airborne LiDAR 91 are rigidly connected, with a fixed relative position and no offset. Their relative position parameters are pre-stored in the vehicle-mounted controller (located in the control and communication box 7), establishing a precise conversion benchmark between the drone's laser point cloud data and the prism's measured true-value coordinates, thus avoiding system errors caused by relative equipment displacement at the source.

[0061] The high-precision positioning module, namely the vehicle-mounted RTK 6, includes a geodesic GPS / BeiDou dual-mode receiver and an inertial navigation unit. Equipped with an RTK differential positioning fusion algorithm, it can collect the unmanned surveying vehicle's position coordinates, driving attitude, and motion parameters in real time and complete the vehicle's real-time positioning. It dynamically analyzes and autonomously corrects positioning deviations by combining site environment and historical positioning data, reducing errors caused by terrain and satellite signal drift. Relying on satellite differential and inertial navigation fusion compensation, it outputs centimeter-level positioning coordinates when the vehicle is moving and outputs millimeter-level high-precision vehicle positioning coordinates when stationary, providing high-precision reference coordinates for the calculation of measured data and the construction of global point cloud error correction models.

[0062] The wireless communication module (located in the control and communication box 7) adopts a 5G+wireless bridge dual-mode redundant communication architecture, which combines low-latency transmission and long-distance stable communication capabilities. During operation, the unmanned surveying vehicle 1 and the surveying drone 9 are independently equipped with high-precision timing modules and synchronously record data acquisition timestamps to achieve a unified time reference for air and ground equipment. Relying on the complementary advantages of the dual communication modes, low-latency and highly reliable two-way data interaction between the unmanned vehicle and the drone is completed, and the prism true coordinate data collected by the prism tracking high-precision total station 2 and the raw point cloud data collected by the drone's airborne lidar 91 are synchronized in real time. The timing error caused by the asynchronous data acquisition of air and ground equipment is offset by the timestamp matching correction algorithm, which reduces the time dimension deviation from the source, ensures the timing consistency and matching accuracy of the point cloud correction data, and provides accurate and synchronous raw data support for the global error correction model.

[0063] Specifically, the dual-mode wireless communication module adopts a dual-link redundant transmission architecture of 5G public network communication + 5.8G industrial wireless bridge, with a communication distance of ≥5km and an automatic switching time of ≤100ms when a single link is interrupted, ensuring uninterrupted data transmission of control commands and sensors. Time synchronization adopts a dual mechanism of "hardware PTP triggering + software delay compensation": at the hardware level, multi-sensor hardware triggering synchronization is achieved through the PTP precise time protocol, and at the software level, the fixed sampling delay of each sensor is calibrated, and residual time errors are corrected through timestamp matching algorithms. Finally, the time matching accuracy of multi-source data is better than 0.5μs, eliminating the timing misalignment error of air-to-ground data from the root.

[0064] The vehicle-mounted controller incorporates a point cloud layering correction algorithm, which includes: collecting multiple sets of measured true coordinates of the prism at different airspace altitudes and horizontal positions and corresponding sampled coordinates of the UAV point cloud; calculating global coordinate offset, elevation deviation, and attitude distortion parameters; constructing a global linear correction model to complete the overall coordinate translation, rotation, and scaling correction of the point cloud; and filtering out abnormal point clouds by using a true residual threshold for local distortion points and noise points to complete local fine correction.

[0065] Specifically, the point cloud layered correction algorithm adopts a layered progressive dual correction mechanism: a global correction model is constructed with millimeter-level true values ​​from a total station as the core, and global system error correction is completed through multi-position and multi-height true value sampling to eliminate overall drift and elevation deviation; then, local noise removal and distortion repair are completed through residual threshold screening, comprehensively solving the technical pain points of UAV point cloud positioning drift, inaccurate elevation, local deformation, and redundant noise points. Global correction uses a seven-parameter model (3 translations, 3 rotations, and 1 scale) to uniformly correct the overall system drift, coordinate offset, and elevation distortion of the point cloud, completing the overall correction of the entire domain; local correction uses the continuous ground point cloud scanned by the vehicle-mounted LiDAR 5 as the area reference and the vertical elevation obtained by the laser rangefinder 24 at the bottom of the self-stabilizing platform as the single-point constraint, calculates the residual between the point cloud and the true value model point by point, automatically removes noise points, floating points, and distorted points with residuals greater than 3 times the standard error, and performs weighted smoothing correction on areas with large local residuals to complete the local fine optimization of the point cloud.

[0066] The unmanned mapping vehicle 1 has a built-in autonomous navigation module and obstacle avoidance sensor array (including obstacle avoidance radar 12), and is equipped with an onboard LiDAR 5 (360° panoramic LiDAR). The onboard LiDAR 5 can scan the surrounding environment in real time, build a high-precision real-time environment model, and accurately identify obstacles, terrain undulations, and work boundaries. Combined with onboard RTK... 6. High-precision millimeter-level positioning data and environmental model dynamic planning optimize the unmanned vehicle's mapping route. Simultaneously, the mapping path can be preset in advance through electronic maps, ensuring the rationality and accuracy of the mapping path. The vehicle controller has a built-in puncture point location storage and automatic point search program, which can record the spatial coordinates, location numbers, and attitude information of all mapping puncture points in real time during the operation, forming a local puncture point database. During subsequent re-measurement and supplementary measurement operations, the system can automatically retrieve historical puncture point data, combine real-time positioning and environmental model to autonomously plan the driving path, and combine the vehicle-mounted ground fixed measurement point tracking mechanism (360° rotating platform 8 and vehicle under-body high-definition camera 81) to accurately search for and automatically arrive at the preset position to complete the fixed-point re-measurement and supplementary data collection operations. It realizes fully automatic cruising, intelligent obstacle avoidance, accurate point search and trajectory driving in complex sites. Combined with the UAV airspace mapping operation mode, it efficiently completes the integrated operation of air-ground collaborative fully automatic point cloud acquisition and accuracy correction.

[0067] Specifically, the unmanned mapping vehicle 1 adopts an off-road four-wheel chassis adapted to complex unpaved roads, with a maximum driving speed of 15km / h, a climbing ability of ≥35°, and an obstacle crossing height of ≥15cm. It has a built-in multi-line laser obstacle avoidance sensor array (including obstacle avoidance radar 12) and a wheeled servo drive unit. It supports the import of electronic maps of the survey area and automatically presets the full-domain cruise path, and can realize intelligent path planning, fully automatic cruise, real-time obstacle recognition and autonomous obstacle avoidance. It can adapt to the rhythm of UAV airspace scanning operations and complete integrated fully automatic mapping and point cloud accuracy correction operations through air-ground cooperation. The vehicle-mounted controller has a built-in puncture point database module. During operation, it automatically records information such as the global coordinates, point number, measurement time, point cloud accuracy, and on-site photos of all measurement points to form a local measurement point archive. When performing periodic re-measurement operations, the system automatically loads historical puncture point data, combines RTK positioning and laser SLAM navigation, autonomously plans the optimal path to drive above the target measurement point, and automatically starts the ground measurement point automatic calibration system to complete the centering measurement. There is no need for manual input of coordinates or finding points. The centering accuracy of the re-measurement point is better than 1mm, with no cumulative error and no point misalignment or omission. It is particularly suitable for scenarios such as foundation pits and slopes that require long-term millimeter-level deformation monitoring.

[0068] like Figure 6 As shown, the present invention also provides an automated point cloud accuracy correction method, applied to the aforementioned automated point cloud accuracy correction mapping vehicle, comprising the following steps:

[0069] S1. Work Deployment and Equipment Self-Check:

[0070] The unmanned surveying vehicle 1 autonomously travels to the target surveying area according to the preset surveying path and real-time environment model. It sequentially completes the onboard RTK 6mm-level positioning calibration, the horizontal attitude self-check of the three-axis six-DOF self-stabilizing platform, the parameter initialization of the prism-tracking high-precision total station 2, and the self-check of the whole machine communication link to ensure that the reference parameters of each device are normal and the communication is stable. The ground control station binds and connects the unmanned surveying vehicle 1 and the surveying drone 9 simultaneously through the dual communication mode of 5G + wireless bridge to complete the equipment status inspection and initial parameter synchronization. The surveying drone 9 takes off to the preset surveying airspace and uses the airborne lidar 91 and the 540° cooperative prism 92 to perform a full-area scan of the work site to obtain the original three-dimensional laser point cloud data of the site. After the equipment is powered on, it automatically performs a self-test process: the unmanned surveying vehicle automatically drives to the known ground benchmark point in the survey area, the on-board controller starts the automatic calibration module, controls the prism tracking high-precision total station 2 to accurately aim at the benchmark target, collect the benchmark true value data, automatically calculate the installation deviation of the self-stabilizing platform and the inherent error of the total station system and complete the parameter compensation calibration; the three-axis six-degree-of-freedom self-stabilizing platform dynamically runs to complete the self-calibration of the three-axis dynamic compensation parameters; the air-to-ground 5G + wireless bridge dual-channel automatic handshake matching unifies the high-precision timestamp of the equipment to ensure that the air-to-ground data timing is completely synchronized.

[0071] S2. Measurement point acquisition and automatic calibration:

[0072] The automatic calibration system for ground measurement points at the bottom of the vehicle uses a high-definition camera 81 under the vehicle to collect real-time images of the ground below and identify and capture the fixed ground measurement points set up on the site. The system controls the laser grating sensor to project a grating grid to fully cover the measurement point area, calculates the horizontal deviation between the prism-tracking high-precision total station 2 reference point and the ground measurement point, and uploads the deviation data to the on-board controller in real time. The controller issues adjustment commands according to the feedback error, synchronously drives the unmanned surveying vehicle to move and the three-axis six-degree-of-freedom self-stabilizing platform to fine-tune the attitude, continuously correcting the position deviation in a closed loop until the total station reference point and the ground measurement point are accurately aligned, completing the automatic alignment calibration of the entire machine's measurement reference.

[0073] S3, Air-Ground Cooperative Dynamic Tracking Measurement:

[0074] The surveying drone 9, carrying a 540° collaborative prism 92, takes off and conducts a full-domain scanning operation. The drone uses its own satellite positioning module to obtain the airspace position coordinates in real time and actively transmits the satellite position coordinate data to the unmanned surveying vehicle 1 via a wireless communication link as a guidance signal. After receiving the drone's satellite position information, the unmanned surveying vehicle 1 quickly locks onto the target area using a high-definition recognition camera 4 and a prism tracking function. It actively tracks and locks onto the 540° collaborative prism 92 carried by the drone and obtains the drone's high-precision spatial true position in real time through dynamic tracking measurement of the prism. Relying on the real-time attitude compensation capability of the three-axis six-degree-of-freedom self-stabilizing platform, it dynamically adjusts the pitch, roll, and yaw angles. In conjunction with the servo tracking mechanism, it drives the prism tracking high-precision total station 2 to continuously and accurately track and aim at the center of the collaborative prism. It collects the three-dimensional spatial true coordinates of the prism and the drone in real time in a high-frequency sampling mode to form a high-precision ground true data set.

[0075] S4. Intelligent calculation of multi-dimensional error parameters:

[0076] The vehicle-mounted controller completes air-to-ground data matching based on the synchronization timestamp. It compares and analyzes the measured true coordinates of the prism collected by the prism-tracking high-precision total station 2 with the coordinate information of the corresponding position in the UAV's original point cloud data. It intelligently calculates multi-dimensional error parameters such as global offset error, elevation system error, spatial angle distortion error, and local point residuals in the point cloud data, and constructs a dynamic error correction model adapted to the current working site. The system selects multiple evenly distributed paired data sets in the low-altitude, mid-altitude, and high-altitude areas of the survey area, accurately calculates the global offset of the X / Y / Z axes, elevation system deviation, coordinate rotation matrix, and scaling distortion coefficient (i.e., 7 parameters in total: 3 translations, 3 rotations, and 1 scale), and constructs a linear global error correction model.

[0077] S5, Hierarchical Point Cloud Accuracy Correction:

[0078] Based on the established global error correction model, the original laser point cloud data is subjected to overall coordinate correction, uniformly correcting the overall positioning drift and overall offset of the point cloud. Then, by using a preset local residual threshold, individual point data is screened and identified, automatically eliminating distorted noise points, abnormal floating points, and points with excessive deviations. This achieves point-by-point correction, noise reduction optimization, and local distortion repair of the original laser point cloud data, completing comprehensive and multi-level accuracy optimization of the point cloud data. Specifically, firstly, the overall system drift, coordinate offset, and elevation distortion of the point cloud are uniformly corrected through a global linear correction model, completing the overall correction across the entire domain. Then, using the continuous ground point cloud scanned by the vehicle-mounted LiDAR 5 as the area reference and the vertical elevation obtained by the bottom laser rangefinder 24 as the single-point constraint, the residual between the point cloud and the true model is calculated point by point. Noise points, floating points, and distorted points with residuals greater than 3 times the standard error are automatically eliminated. Weighted smoothing correction is performed on areas with large local residuals, completing the local fine optimization of the point cloud.

[0079] S6. Data storage and remote output:

[0080] The high-precision 3D point cloud data, after accuracy correction, is locally encrypted and stored, while simultaneously being uploaded in real-time to the ground control station and surveying terminal via dual redundant communication links. This completes an integrated surveying and mapping workflow of fully automated air-ground collaborative point cloud acquisition, error correction, and data output. The corrected high-precision 3D point cloud data is encrypted and stored on an onboard solid-state drive (located in the control and communication box 7) to ensure data security. Under normal public network conditions, it is automatically transmitted to the remote surveying server via the 5G public network. In remote areas without public network access, it automatically switches to an industrial wireless bridge to transmit to the ground surveying terminal, adapting to data output needs across all scenarios.

[0081] This embodiment uses the completion acceptance and long-term deformation monitoring scenario of a subway deep foundation pit (200m long, 100m wide, and 18m deep) as an example to illustrate the technical solution of the present invention in detail:

[0082] In this embodiment, the unmanned mapping vehicle 1 adopts a four-wheel off-road chassis with a maximum driving speed of 15km / h, a climbing ability of 35°, and an obstacle crossing height of 15cm, adapting to the unpaved road on the top of the foundation pit crown beam; the servo electric cylinder of the three-axis six-degree-of-freedom self-stabilizing platform adopts a 100mm stroke high-precision ball screw electric cylinder with a positioning accuracy of 0.01mm, and is equipped with a 0.1″ resolution tilt sensor and a 200Hz IMU (i.e., three-axis attitude sensor 23) with a dynamic leveling accuracy of 2″; the bottom laser rangefinder 24 adopts a vertical laser ranging module with a ranging range of 0.5-30m, an accuracy of ±1mm, and a sampling frequency of 50Hz.

[0083] The prism-tracking high-precision total station 2 adopts a 1″-class servo total station with an angle measurement accuracy of 1″ and a distance measurement accuracy of 1mm+1ppm. It is equipped with a 20-megapixel global camera (i.e., a high-definition recognition camera 4) and a synchronous tracking distance of 1500m, which can stably track drones flying at a speed of 18m / s. The matching 540° cooperative prism 92 uses K9 high-reflectivity glass with a reflectivity of 92%, and is at a fixed distance from the airborne lidar 91 on the drone, with a relative position calibration error of 0.08mm.

[0084] The ground measurement point automatic calibration system uses a 20-megapixel industrial camera (i.e., a high-definition camera 81 under the vehicle) and an 850nm laser grating projector (integrated into a 360° rotating platform 8), with a centering accuracy of 0.4mm and an average centering time of 25 seconds; the vehicle-mounted RTK 6 adopts a geodetic RTK+IMU combination, with dynamic positioning accuracy at the centimeter level and static accuracy at the millimeter level; the communication module (located in the control and communication box 7) adopts a 5G+5.8G wireless bridge with dual backup, a communication distance of 5km, and a link switching time of 80ms.

[0085] The operation process in this embodiment is as follows:

[0086] Deployment and self-calibration: Technicians import the foundation pit boundary at the ground control station, and the system automatically plans the inspection path along the cap beam and the flight path of the UAV at a height of 50m; after the UAV is turned on, it autonomously drives to the benchmark point at the entrance of the foundation pit, and the ground measuring point automatic calibration system (360° rotating platform 8 and high-definition camera 81 under the vehicle) completes the centering in 25 seconds, automatically measures the coordinates of the benchmark point and compensates for system errors. The entire calibration process takes 1 minute and 40 seconds and requires no manual operation.

[0087] Collaborative operation: The unmanned vehicle autonomously travels along the crown beam at a speed of 6 km / h, and the three-axis six-DOF self-stabilizing platform adjusts its attitude and altitude in real time to maintain the stability of the measurement benchmark; after taking off, the surveying UAV 9 flies along the terrain at a speed of 8 m / s, and the airborne lidar 91 collects the original point cloud; the prism tracking high-precision total station 2 automatically locks onto the UAV's 540° collaborative prism 92 and outputs the true coordinates at a frequency of 10 Hz; the vehicle-mounted lidar 5 and the bottom laser rangefinder 24 synchronously collect ground benchmark data, and all data are transmitted in real time to the edge controller in the control and communication box 7 after being marked with a unified hardware timestamp.

[0088] Real-time correction: The controller runs a hierarchical correction algorithm in real time, processing one frame of point cloud every 80ms. First, it completes global correction by solving 7 parameters through multiple sets of true values. Then, it completes local denoising and distortion correction by using vehicle-mounted point cloud and laser elevation constraints. The corrected point cloud is transmitted back to the control terminal in real time, and the point cloud results can be viewed in real time during the operation.

[0089] Output and Archiving: The entire field operation took 42 minutes. Upon completion, the point cloud results for the entire foundation pit were obtained. According to third-party testing, the point cloud plane accuracy was 2.1cm and the elevation accuracy was 1.2cm, which met the requirements of 1:500 large-scale mapping. It can be directly imported into BIM software to complete the comparison and acceptance of the as-built model. All measurement point locations are automatically stored in the puncture point database.

[0090] In subsequent quarterly foundation pit deformation monitoring operations, the unmanned vehicle automatically loads the puncture point database, autonomously navigates to the location of 20 monitoring points, automatically centers and measures, and the entire re-measurement process takes 30 minutes without human intervention. The elevation measurement accuracy of the monitoring points is 0.8mm, which can accurately capture millimeter-level deformation of the foundation pit.

[0091] Example 1: Implementation of Equipment Hardware Assembly and Power-On Self-Test

[0092] The unmanned mapping vehicle 1 of this invention adopts a four-wheel independent drive intelligent obstacle avoidance unmanned chassis. The overall structure adopts a layered modular integrated layout. The bottom layer of the chassis is equipped with a drive power supply and servo drive module, which provides power support for the vehicle's movement and attitude adjustment. The middle layer of the chassis integrates an on-board industrial controller, a Beidou + GPS dual-mode high-precision GNSS positioning module, an inertial navigation unit, a 5G industrial communication module, a wireless bridge transmitter, and a high-speed storage hard disk (the above modules are integrated in the control and communication box 7), realizing positioning calculation, data processing, algorithm operation, and data storage functions. The upper layer of the chassis equipment compartment integrates an autonomous navigation module and a multi-line laser obstacle avoidance sensor array (including obstacle avoidance radar 12), which is responsible for site environment modeling, path planning, and obstacle recognition.

[0093] The top center of the unmanned surveying vehicle 1 is rigidly fixed to a three-axis, six-degree-of-freedom self-stabilizing platform via a high-strength shock-absorbing bracket. The gimbal adopts a three-axis independent servo drive architecture for pitch, roll, and yaw, which has a fast response speed and high compensation accuracy. A prism-tracking high-precision total station 2 is fixed on the top of the gimbal carrying plate, and a high-definition recognition camera 4 is installed on the side of the total station. The platform has a built-in three-axis attitude sensor 23 and tilt detection module, which can transmit vehicle body sway and attitude deviation data to the vehicle controller in real time.

[0094] The prism-tracking high-precision total station 2 adopts a geodetic high-precision total station as its main body. A high-definition recognition camera 4 is integrated above the total station lens to acquire aerial images in real time and transmit them to the onboard image processing unit. The servo rotation mechanism at the bottom of the total station communicates with the three-axis, six-degree-of-freedom self-stabilizing platform servo system, and the controller synchronously outputs gimbal compensation commands and prism tracking angle commands to achieve 360° continuous tracking and aiming without blind spots. The total station body integrates a top-mounted lidar 21, a laser rangefinder 22, a bottom-mounted laser rangefinder 24, and a total station eyepiece 25, forming a complete high-precision measurement unit.

[0095] The accompanying mapping drone 9 is a quadcopter industrial-grade lidar drone. A 540° collaborative prism 92 is rigidly fixed at the lower part of the fuselage. The drone is equipped with an airborne lidar 91 device at the bottom. The relative position parameters of the prism and lidar are pre-entered into the vehicle controller to establish a unified coordinate transformation benchmark. The drone is equipped with a wireless bridge receiver, an airborne POS positioning unit and a lidar acquisition module, and together with the ground unmanned mapping vehicle and ground control station, they form an air-ground collaborative fully automatic mapping and correction system.

[0096] After the equipment is powered on, it automatically performs a self-test process: the unmanned surveying vehicle automatically drives to the known ground benchmark point in the survey area, the on-board controller starts the automatic calibration module, controls the prism tracking high-precision total station 2 to accurately aim at the benchmark target, collect the benchmark true value data, automatically calculate the installation deviation of the self-stabilizing platform and the inherent error of the total station system and complete the parameter compensation calibration; the three-axis six-degree-of-freedom self-stabilizing platform runs dynamically for 30 seconds to complete the self-calibration of the three-axis dynamic compensation parameters; the air-to-ground 5G + wireless bridge dual-channel automatic handshake matching unifies the high-precision timestamp of the equipment to ensure that the air-to-ground data timing is completely synchronized.

[0097] Example 2: Implementation Process of Vehicle-Mounted Algorithm Operation and Hierarchical Point Cloud Correction

[0098] The vehicle-mounted controller (located within the control and communication box 7) is equipped with an industrial processor and a high-speed graphics processing unit to ensure the real-time performance of image recognition, servo control, and error calculation. The specific algorithm execution flow is as follows:

[0099] 1. Visual target recognition and dynamic gimbal stabilization: The high-definition recognition camera outputs 60 frames of high-definition aerial images per second. The controller quickly extracts the 92 feature contours of the highly reflective 540° collaborative prism through threshold segmentation and contour matching algorithms, and calculates the pixel coordinates and spatial position of the prism. Combined with the coarse positioning coordinates of the POS transmitted back by the UAV, the gimbal's three-axis compensation angle is calculated in real time. The servo motor dynamically compensates for the attitude deviation caused by vehicle bumps and tilts, always keeping the total station lens accurately aligned with the center of the prism, and stably outputting more than 5 sets of prism three-dimensional true coordinates per second.

[0100] 2. Time-series synchronization and matching of air and ground data: The raw point cloud data generated by the UAV's airborne LiDAR 91 scan is accompanied by a high-precision acquisition timestamp. The surveying vehicle synchronously records the prism true value acquisition timestamp. The controller accurately matches the point cloud sampling data and prism true value coordinates at the same time based on the millisecond-level timestamp, forming a sufficient and reliable spatiotemporal matching dataset.

[0101] 3. Global Error Model Construction and Overall Correction: The system selects multiple evenly distributed paired data sets from low, medium, and high altitudes in the survey area, accurately calculates the global offset of the X / Y / Z axes, elevation system deviation, coordinate rotation matrix, and scaling distortion coefficient, and constructs a linear global error correction model; it performs unified coordinate translation, rotation, and scaling correction on the original point cloud data of the entire field, completely eliminating the overall system deviation caused by UAV POS cumulative drift and flight attitude disturbance.

[0102] 4. Local fine-grained denoising and distortion repair: The residual is calculated by comparing the globally corrected point cloud coordinates with the corresponding prism true coordinates. The system presets a residual threshold of ±3cm. Points exceeding the threshold are identified as local distortion points or radar clutter noise points and are automatically removed. For local terrain distortion and sparse point areas, a spatial interpolation smoothing algorithm is used to complete local repair, achieving secondary fine-grained correction.

[0103] 5. High-precision data storage and dual-link upload: The corrected high-precision 3D point cloud data is encrypted and stored on the vehicle's solid-state drive to ensure data security; under normal public network conditions, it is automatically transmitted to the remote mapping server via the 5G public network; in remote areas without public network, it automatically switches to the industrial wireless bridge to transmit to the ground mapping terminal, adapting to the data output needs of all scenarios.

[0104] Example 3: Implementation of BIM-based detailed as-built surveying for building construction projects

[0105] I. Core Intelligent Strategy for Building Construction Adaptation: Adaptive Adjustment of Curvature and Structural Complexity

[0106] This embodiment superimposes a building structure feature perception and speed adaptive intelligent decision-making layer on top of the basic air-ground collaborative tracking and point cloud correction algorithm, and adapts it differently for regular building sites and irregular complex structures. The specific implementation process is as follows:

[0107] 1. Building Scene Complexity Perception: The controller collects data from the laser rangefinder 22 and the bottom laser rangefinder 24 in real time, along with visual tracking offset and gimbal attitude dynamic compensation data. By calculating the data change rate Rc (Rc=ΔDc / Δt) per unit time, it quantifies and judges the structural complexity and contour curvature change range of the building survey area. Among them, building beam-column joints, concave and convex structures of the facade, irregular areas of the roof, and local structural shapes will cause large fluctuations in the data change rate, which are judged as high-complexity areas; the data change rate of the site ground, open paved areas, and regular straight walls is close to zero, which are judged as regular and simple areas.

[0108] 2. Graded Speed ​​Intelligent Decision Logic: The controller presets a baseline cruising speed for building construction scenarios, a speed reduction threshold for high-complexity scenarios, and a speed increase threshold for regular sites. When it detects rapid changes in structural curvature and drastic data fluctuations, with |Rc| exceeding the preset speed reduction threshold, it determines that the system is currently in a complex surveying area such as irregular building structures or beam-column joints. The system automatically and smoothly reduces the cruising speed of the unmanned surveying vehicle and the scanning speed of the drone, providing sufficient response time for precise prism tracking, time-series matching of air and ground data, and multi-dimensional error calculation. This prevents tracking inaccuracies and aggravated point cloud distortion in complex structural areas, ensuring the surveying accuracy of core BIM components. When it detects a flat site and regular wall structures, with |Rc| consistently below the speed increase threshold and remaining stable for a preset duration, the system automatically and smoothly increases speed, restoring the baseline cruising speed. This improves the efficiency of surveying large areas and achieves a dynamic balance between accuracy and efficiency.

[0109] 3. BIM Surveying Dedicated Sampling Frame Rate Adaptation: To meet the needs of detailed building modeling, the system adaptively adjusts the 91-scan frame rate of the UAV-borne LiDAR and the frequency of ground truth data acquisition. In key BIM modeling areas such as building facades, beams, columns, roofs, and node splicing, the system automatically densifies sampling points and increases the frequency of truth data acquisition, refining structural outline details. In non-critical areas such as open areas and flat ground, redundant sampling data is appropriately simplified, ensuring BIM modeling accuracy while reducing the pressure on backend data processing and improving delivery efficiency.

[0110] II. Standardized, Fully Automated Operation Process for Building Construction Using BIM

[0111] S1. Initialization of Operation Deployment: The ground control station imports electronic drawings and BIM benchmark coordinates of the building construction site, and automatically plans the full-area cruise path of the unmanned surveying vehicle and the low-altitude circling flight route of the UAV. The unmanned surveying vehicle 1 autonomously drives to the benchmark control point of the building site, automatically completes the onboard RTK 6 mm-level positioning calibration, dynamic calibration of the three-axis six-degree-of-freedom self-stabilizing platform, benchmark self-check calibration of the prism tracking high-precision total station 2, and high-precision timestamp synchronization of the air-to-ground communication link. The surveying UAV 9 takes off and conducts full-area low-altitude laser scanning along the building facade, roof, site ground and ancillary structures, collecting original three-dimensional laser point cloud data of the building construction site, providing the original data foundation for subsequent BIM reverse modeling and as-built verification.

[0112] S2. Real-time tracking of the air-ground collaborative prism: The UAV pushes its own POS coarse positioning coordinates to the vehicle controller in real time, and the vehicle-mounted high-definition recognition camera 4 quickly locks onto the aerial 540° collaborative prism 92 target; In response to uneven road surfaces, vehicle undulations caused by vehicles rolling over the road surface, and vehicle body swaying caused by wind disturbance during operation, the three-axis six-degree-of-freedom self-stabilizing platform provides high-frequency dynamic compensation for pitch and roll attitude deviations to ensure the stability of the measurement equipment reference; The prism tracking high-precision total station 2 servo mechanism continuously and stably tracks and accurately illuminates the prism center, and works with the laser rangefinder 22 to collect prism spatial distance and angle data at high frequency, continuously outputting high-precision prism three-dimensional true coordinates, providing continuous and reliable ground reference data for building point cloud correction.

[0113] S3. Multi-dimensional error parameter calculation: The vehicle controller batch matches the time-synchronized prism true coordinates with the airborne point cloud sampling data of the building area. For different areas such as the vertical structure of the building facade, the horizontal structure of the roof, the ground of the site, and the ancillary structures, it uniformly calculates a complete set of correction parameters such as global coordinate offset, elevation system error, attitude rotation distortion, and scale deviation of the survey area. It constructs a dynamic error correction model adapted to the complex structural scene of the building and eliminates the overall system deviation caused by the flight disturbance of the UAV.

[0114] S4. Layered Two-Level Point Cloud Accuracy Correction: The global linear correction model uniformly corrects the overall coordinate drift, elevation deviation, and attitude distortion of the entire building site point cloud, ensuring that the overall position, floor height, and span dimensions of the building conform to the design benchmark. By automatically filtering out reflective noise on the building facade, wall cutout noise, roof fragmentation points, and site interference noise through preset residual thresholds, the system performs interpolation smoothing repair on local distortion areas of key structures such as walls, beams, columns, roofs, and floors, resulting in a high-precision 3D point cloud of the building with clear structural outlines, accurate dimensions, and flatness that conforms to reality, meeting the accuracy requirements of BIM fine modeling, component size verification, and as-built comparison analysis.

[0115] S5. Output Archiving: The corrected high-precision 3D point cloud of the building is automatically encrypted and stored locally, and uploaded in real time to the remote BIM surveying terminal via dual links of 5G and industrial wireless bridge. It can be directly used for reverse BIM as-built modeling of buildings, design model comparison and verification, accurate calculation of engineering quantities, and construction quality acceptance. The whole process is unmanned and fully automated, completing high-precision surveying of building sites and outputting usable BIM results.

[0116] III. Standardized, Fully Automated Operation Process for Building Construction Using BIM

[0117] S1. Initialization of Operation Deployment: The ground control station imports electronic drawings and BIM benchmark coordinates of the building construction site, and automatically plans the full-area cruise path of the unmanned surveying vehicle and the low-altitude circling flight route of the UAV. The unmanned surveying vehicle autonomously drives to the benchmark control point of the building site, automatically completes GNSS millimeter-level positioning calibration, three-axis platform dynamic calibration, automatic measurement robot benchmark self-check calibration, and high-precision timestamp synchronization of the air-to-ground communication link. The surveying UAV takes off and conducts full-area low-altitude laser scanning along the building facade, roof, site ground and ancillary structures, collecting original three-dimensional laser point cloud data of the building construction site, providing the original data foundation for subsequent BIM reverse modeling and as-built verification.

[0118] S2. Real-time tracking of the air-to-ground collaborative prism: The UAV pushes its own coarse positioning coordinates to the vehicle controller in real time, and the vehicle's high-definition vision system quickly locks onto the aerial collaborative prism target; to address the unevenness of the road surface at the construction site, the undulation of the road surface caused by vehicles on site, and the swaying of the vehicle body caused by wind disturbance during operation, the three-axis gimbal uses high-frequency dynamic compensation for pitch and roll attitude deviations to ensure the stability of the measurement equipment's reference; the automatic measurement robot's servo mechanism continuously and smoothly tracks and accurately aims at the center of the prism, and, in conjunction with laser ranging, high-frequency acquisition of the prism's spatial distance and angle data, continuously outputs high-precision prism three-dimensional true coordinates, providing continuous and reliable ground reference data for the correction of the construction point cloud.

[0119] S3. Multi-dimensional error parameter calculation: The vehicle controller batch matches the time-synchronized prism true coordinates with the airborne point cloud sampling data of the building area. For different areas such as the vertical structure of the building facade, the horizontal structure of the roof, the ground of the site, and the ancillary structures, it uniformly calculates a complete set of correction parameters such as global coordinate offset, elevation system error, attitude rotation distortion, and scale deviation of the survey area. It constructs a dynamic error correction model adapted to the complex structural scene of the building and eliminates the overall system deviation caused by the flight disturbance of the UAV.

[0120] S4. Layered Two-Level Point Cloud Accuracy Correction: The global linear correction model uniformly corrects the overall coordinate drift, elevation deviation, and attitude distortion of the entire building site point cloud, ensuring that the overall position, floor height, and span dimensions of the building conform to the design benchmark. By automatically filtering out reflective noise on the building facade, wall cutout noise, roof fragmentation points, and site interference noise through preset residual thresholds, the system performs interpolation smoothing repair on local distortion areas of key structures such as walls, beams, columns, roofs, and floors, resulting in a high-precision 3D point cloud of the building with clear structural outlines, accurate dimensions, and flatness that conforms to reality, meeting the accuracy requirements of BIM fine modeling, component size verification, and as-built comparison analysis.

[0121] S5. Output Archiving: The corrected high-precision 3D point cloud of the building is automatically encrypted and stored locally, and uploaded in real time to the remote BIM surveying terminal via dual links of 5G and industrial wireless bridge. It can be directly used for reverse BIM as-built modeling of buildings, design model comparison and verification, accurate calculation of engineering quantities, and construction quality acceptance. The whole process is unmanned and fully automated, completing high-precision surveying of building sites and outputting usable BIM results.

[0122] Example 4: Implementation of Layered Progressive High-Precision Point Cloud Correction and BIM Result Output

[0123] Based on the device architecture and intelligent adaptive control of the aforementioned embodiments, Embodiment 4 of the present invention further deepens the layered two-level point cloud accuracy correction algorithm, focusing on the core requirements of high-precision engineering measurement and standardized BIM model delivery, and realizes the whole process of global error correction, local noise reduction and repair, refined result optimization and standardized archiving output, thereby comprehensively improving the accuracy of point cloud data and engineering usability.

[0124] I. Core High-Precision Correction Algorithm System

[0125] This invention relies on an unmanned mapping vehicle 1 equipped with a three-axis, six-DOF self-stabilizing platform and a prism-tracking high-precision total station 2. Using a UAV equipped with a 540° collaborative prism 92 as a high-precision spatial truth medium, it constructs a collaborative, hierarchical, and progressive mathematical model for millimeter-level point cloud error correction and elimination. Through multi-dimensional, end-to-end algorithmic formulas, it systematically eliminates equipment eccentricity errors, vehicle attitude disturbance errors, UAV flight drift errors, spatiotemporal asynchrony errors, global system biases, and local random distortions, ultimately correcting the original UAV laser point cloud from traditional centimeter-level accuracy to millimeter-level accuracy. The entire core error correction algorithm system is as follows:

[0126] 1. Prism observation gross error elimination model

[0127] A high-precision total station with prism tracking, operated by an unmanned vehicle (UAV), continuously tracks and acquires the 3D coordinate sequence of a 540° collaborative prism from a UAV. To eliminate gross errors caused by sea breeze disturbances, momentary occlusion, and tracking jitter, the 2σ standard deviation criterion is used to remove outliers from the original observation sequence. Let the prism observation coordinate time series be: The total number of valid samples is n. The formula for calculating the coordinate mean and standard deviation is the arithmetic mean of each coordinate component and the Bezier standard deviation. The criterion for removing outliers is: if the deviation of any coordinate component observation from the mean exceeds twice the standard deviation, the current frame observation is determined to be an outlier and is directly removed, retaining the high-precision valid observation sequence.

[0128] Formulas for calculating the mean and standard deviation of coordinates:

[0129]

[0130]

[0131]

[0132] Gross error removal criteria: If , , If any condition is met, the current frame observation is determined to be a gross error and is directly discarded, while the high-precision valid observation sequence is retained.

[0133] 2. Truth-value coordinate time-series smoothing filtering model

[0134] The prism truth sequence after gross error removal is filtered using a three-point moving average to suppress high-frequency micro-jitter noise and obtain spatiotemporally continuous and highly stable reference truth coordinates. :

[0135]

[0136]

[0137]

[0138] 3. UAV prism and airborne lidar eccentricity calibration model

[0139] Because the UAV's onboard lidar and the cooperative prism are rigidly mounted with a fixed physical eccentricity, the eccentricity compensation is calculated through repeated multi-attitude calibration to eliminate the system error. The eccentricity compensation is obtained by subtracting the instrument system constant from the average difference between the prism coordinates and radar coordinates from multiple sets (e.g., 15 sets) of synchronous observations. Through pre-calibration of the eccentricity parameters, fixed deviations caused by hardware installation are completely eliminated. This is combined with the instrument system constant. Make compensation adjustments:

[0140]

[0141]

[0142]

[0143] By pre-calibrating the eccentricity parameter, the fixed deviation caused by hardware installation can be completely eliminated.

[0144] 4. Spatial-Temporal Synchronization Interpolation Model for Ground-to-Air Space

[0145] To address the inconsistency between the sampling frequency of the UAV-borne lidar 91 and the observation frequency of the ground total station, high-precision timestamp linear interpolation is employed to achieve accurate temporal alignment between the prism's true value and the airborne point cloud data. Within the interval, interpolation yields the original radar coordinates corresponding to the true time:

[0146]

[0147] The same interpolation formula is used in the Y and Z dimensions to achieve one-to-one correspondence between air and ground data and temporal misalignment matching.

[0148] 5. Truth-value transformation model of UAV attitude coupling

[0149] By combining the real-time roll, pitch, and yaw attitude angles of the UAV with the installation eccentricity parameters, the true values ​​of the prism observations are converted into the true spatial coordinates of the center of the airborne lidar 91, eliminating the coordinate offset caused by the attitude disturbance of the UAV and obtaining the true millimeter-level reference coordinates of the radar. .

[0150] 6. Point cloud global error correction model (rotation + translation global correction)

[0151] Establish an overall distortion correction equation for UAV point clouds to uniformly solve the global system errors caused by flight drift and attitude deflection:

[0152]

[0153] in: The original point cloud coordinates of the UAV. Here is the attitude distortion rotation matrix. This is the global translation deviation vector. The noise is random. The optimal rotation and translation parameters are solved by least squares iteration to complete the overall attitude correction and coordinate zeroing correction of the entire point cloud, solving the problems of overall point cloud drift, overall elevation shift and overall torsional distortion.

[0154] 7. Local residual threshold denoising and refined repair model

[0155] After global correction is completed, local noise and irregular distortion points are eliminated by judging the three-dimensional residuals. The three-dimensional residual calculation formula is the square root of the sum of the squares of the differences between the corrected values ​​and the true values ​​of each coordinate component. A millimeter-level residual threshold is set. The three-dimensional residual calculation formula is as follows:

[0156]

[0157] Set millimeter-level residual threshold ,right The system automatically removes distortion noise, wall reflection noise, airflow interference points, and local structural deformation points. Spatial interpolation is used to complete and repair sparse and missing areas, ultimately obtaining a millimeter-level high-precision 3D point cloud of buildings that is uniform across the entire area, complete in detail, and free of distortion, fully meeting the accuracy requirements for BIM fine modeling and engineering completion verification.

[0158] The overall working principle of this invention is as follows: An autonomous unmanned mapping vehicle 1 serves as a high-precision ground reference carrier, utilizing onboard RTK 6 (BeiDou + GPS dual-mode GNSS and inertial navigation fusion positioning) technology to construct a stable ground mapping reference. A three-axis, six-degree-of-freedom self-stabilizing platform dynamically compensates for attitude errors caused by vehicle movement bumps, tilts, and swaying, ensuring the horizontal stability of the prism-tracking high-precision total station 2 throughout the process. Based on UAV satellite coarse positioning guidance and high-definition recognition camera 4 machine vision recognition technology, the total station automatically tracks the entire domain, and the 540° collaborative prism 92 mounted on the laser high-frequency aiming mapping UAV 9 acquires high-precision UAV spatial true coordinates in real time. Through the 5G+ wireless bridge dual-redundant communication architecture and high-precision timestamp synchronization technology within the control and communication box 7, low-latency, error-free synchronous interaction of air and ground data is achieved. Relying on the layered progressive point cloud correction algorithm built into the onboard controller, the global system is first calculated. The system corrects overall coordinate translation, rotation, scaling, and offset errors. Then, it uses the vehicle-mounted LiDAR 5 and the bottom laser rangefinder 24 to provide area + single-point constraints to complete local residual denoising and distortion repair, ultimately outputting millimeter-level high-precision 3D point cloud results. At the same time, it automatically stores all measurement point information through the puncture point database. Combined with the 360° rotating platform 8 and the vehicle-mounted high-definition camera 81, it supports automatic point finding and automatic centering measurement during subsequent re-measurements and supplementary measurements. This forms a closed loop of unmanned high-precision surveying and mapping with "automatic arrival - automatic centering - automatic calibration - automatic tracking - real-time correction - automatic re-measurement", which completely solves the industry pain points of insufficient accuracy of traditional UAV surveying and mapping and low efficiency of manual correction. It provides efficient, unmanned, and high-precision technical equipment support for engineering BIM delivery, precision terrain monitoring, and large-scale site digital surveying and mapping.

[0159] The specific implementation schemes described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific implementation schemes of the present invention and are not intended to limit the scope of the present invention. Any equivalent changes and modifications made by those skilled in the art without departing from the concept and principles of the present invention should fall within the scope of protection of the present invention.

Claims

1. An automated point cloud accuracy correction mapping vehicle, comprising an unmanned mapping vehicle body, characterized in that, The unmanned surveying vehicle is fixedly equipped with a self-stabilizing platform on its top. The self-stabilizing platform is equipped with an automatic measurement robot and a laser ranging system on its bearing end. The unmanned surveying vehicle integrates an on-board controller, a high-precision positioning module, a laser radar module, and a wireless communication module. The unmanned surveying vehicle is also equipped with an automatic ground measurement point calibration system. The surveying vehicle is equipped with a surveying drone. The surveying drone is equipped with a collaborative prism and an airborne lidar to collect raw laser point cloud data, as well as a wireless communication module to maintain communication with the surveying vehicle. The surveying vehicle is equipped with a ground control station, which integrates a 5G communication module, an unmanned equipment remote control module, and a real-time data transmission module. The ground control station establishes a two-way wireless communication link with the unmanned surveying vehicle and the surveying drone, respectively, for remote manual / automatic collaborative control of the unmanned surveying vehicle and the surveying drone's operating trajectory, measurement attitude, and surveying parameters. At the same time, it receives, summarizes, and stores high-precision true coordinate data collected by the unmanned surveying vehicle and raw point cloud data collected by the drone's airborne lidar in real time, realizing real-time transmission, synchronous interaction, and remote monitoring of air-to-ground surveying data. It also works with the vehicle-mounted controller to complete remote accuracy correction and data optimization processing of the point cloud data. The vehicle controller is configured to perform the following steps: The satellite positioning system drives the unmanned vehicle to the benchmark measurement point. Image recognition and laser calibration drive the unmanned vehicle to move slightly, so that the laser point of the measurement robot can be aligned with the benchmark measurement point. Then, the image recognition algorithm captures the aerial mapping drone and the collaborative prism target in real time, and drives the self-stabilizing platform to dynamically level and compensate for angles, so as to maintain the stability of the automatic measurement robot's posture. The automatic measurement robot is controlled to automatically track and aim at the collaborative prism, and high-precision three-dimensional true coordinate data of the prism is obtained in real time. The system receives raw laser point cloud data transmitted back by a surveying drone, constructs an error correction model by combining the measured true coordinates of the prism, performs global deviation correction and local distortion correction on the raw laser point cloud data, and outputs high-precision corrected 3D point cloud data.

2. The automated point cloud accuracy correction mapping vehicle according to claim 1, characterized in that, The self-stabilizing platform is a three-axis anti-shake self-stabilizing structure with built-in attitude sensors, tilt detection modules, vehicle speed detection modules, and servo adjustment motors. It collects real-time data on the unmanned surveying vehicle's speed, robot height, tilt angle, and overall vehicle status, including posture sway. Relying on a built-in dynamic attitude adaptive adjustment algorithm, it automatically adjusts and compensates for pitch, roll, and yaw angles in real time according to different driving conditions and changes in vehicle posture. It dynamically calibrates the platform's level, ensuring that the automatic measurement robot mounted on the platform maintains an absolutely level measurement posture throughout all operating conditions, including when the unmanned vehicle starts and stops, travels at a constant speed, and travels on bumpy surfaces. Simultaneously, it collects and outputs accurate height and spatial posture data of the measurement robot in real time, ensuring the stability and accuracy of the measurement benchmark.

3. The automated point cloud accuracy correction mapping vehicle according to claim 2, characterized in that, The automatic calibration system for ground measuring points adopts a ground-fixed measuring point tracking mechanism, which has a built-in laser grating sensor and camera module. During operation, it is located at the bottom of the unmanned mapping vehicle. The camera module identifies the position of the ground measuring points, the laser grating grid covers the measuring point area, the horizontal deviation between the robot's reference point and the measuring points is calculated, and the deviation data is uploaded to the vehicle controller. The vehicle controller drives the unmanned mapping vehicle and the self-stabilizing platform to automatically complete the alignment calibration with the reference point.

4. The automated point cloud accuracy correction mapping vehicle according to claim 3, characterized in that, The automated measurement robot integrates a high-precision total station, a high-definition visual recognition camera, a laser rangefinder, and a servo tracking mechanism. The high-definition visual recognition camera is used to acquire panoramic images of the airspace in real time, accurately identify, lock, and dynamically position the spatial position of the surveying drone and the collaborative prism. The servo tracking mechanism works in conjunction with the self-stabilizing platform to achieve 360° horizontal and ±90° pitch rotation without blind spots, completing fully automated tracking and precise aiming of the collaborative prism in both static and moving states, ensuring the continuity and high precision of the static and dynamic measurement processes.

5. The automated point cloud accuracy correction mapping vehicle according to claim 4, characterized in that, The collaborative prism is a high-reflectivity 540° omnidirectional prism, which has an ultra-large angle reflection recognition range and high reflectivity, effectively improving the recognition efficiency and accuracy of high-definition visual recognition cameras and is suitable for high-precision dynamic laser tracking operations. The collaborative prism and the UAV-borne lidar adopt a fixed integrated layout, and their relative positions remain constant, constructing a unified and stable airborne mapping benchmark, avoiding measurement errors caused by relative displacement, and providing a reliable benchmark for subsequent point cloud data accuracy correction.

6. The automated point cloud accuracy correction mapping vehicle according to claim 5, characterized in that, The high-precision positioning module includes a geodesic GPS / BeiDou dual-mode receiver and an inertial navigation unit, equipped with an RTK differential positioning fusion algorithm. It collects the unmanned surveying vehicle's position coordinates, driving attitude, and motion parameters in real time and completes real-time vehicle positioning. It dynamically analyzes and autonomously corrects positioning deviations by combining site environment and historical positioning data, reducing errors caused by terrain and satellite signal drift. Relying on satellite differential and inertial navigation fusion compensation, it outputs centimeter-level positioning coordinates when the vehicle is moving and outputs millimeter-level high-precision vehicle positioning coordinates when stationary, providing high-precision reference coordinates for the calculation of measured data and the construction of a global point cloud error correction model. The vehicle-mounted controller incorporates a point cloud layering correction algorithm, which specifically includes: collecting multiple sets of measured true coordinates of the prism at different airspace heights and horizontal positions and corresponding sampled coordinates of the UAV point cloud; calculating global coordinate offset, elevation deviation, and attitude distortion parameters; constructing a global linear correction model to complete the overall coordinate translation, rotation, and scaling correction of the point cloud; and filtering out abnormal point clouds by using a true residual threshold for local distortion points and noise points to complete local fine-grained correction.

7. The automated point cloud accuracy correction mapping vehicle according to claim 6, characterized in that, The unmanned mapping vehicle is equipped with an autonomous navigation module and an obstacle avoidance sensor array, and also features a 360° panoramic LiDAR. This 360° panoramic LiDAR can scan the surrounding environment in real time, constructing a high-precision real-time environmental model to accurately identify obstacles, terrain undulations, and operational boundaries. Combined with high-precision RTK millimeter-level positioning data and the environmental model, the vehicle dynamically plans and optimizes its mapping route. Simultaneously, it can pre-set mapping paths via electronic maps, ensuring both the rationality and accuracy of the mapping path. The onboard controller has a built-in puncture point location storage and automatic point-finding program, recording the spatial coordinates, location numbers, and attitude information of all mapping puncture points in real time during the operation, forming a local puncture point database. During subsequent re-measurement and supplementary measurement operations, the system can automatically retrieve historical puncture point data, autonomously plan its driving path using real-time positioning and the environmental model, and, combined with the onboard ground fixed measurement point tracking mechanism, accurately search for and automatically reach the preset location to complete the fixed-point re-measurement and supplementary sampling operations. It enables fully automatic cruising, intelligent obstacle avoidance, precise point finding and trajectory driving in complex terrains, and, in conjunction with UAV airspace mapping operation mode, efficiently completes integrated air-ground collaborative fully automatic point cloud acquisition and accuracy correction operations.

8. The automated point cloud accuracy correction mapping vehicle according to claim 7, characterized in that, The automated measurement robot is equipped with an automatic calibration module. Before each surveying operation, the robot performs a fully automatic self-check and accuracy calibration based on a pre-set ground reference point. By comparing the standard coordinates of the reference point with the robot's measured coordinates, it automatically calculates and compensates for multiple error sources, such as the installation deviation of the self-stabilizing platform, the mechanical assembly error of the equipment, and the inherent measurement error of the system. This enables adaptive correction of the equipment's measurement posture and measurement parameters, effectively eliminating accumulated system errors. No manual calibration or debugging is required throughout the entire process, ensuring the robot's prism tracking response accuracy, dynamic aiming accuracy, and spatial true coordinate acquisition accuracy throughout the operation. This provides a reliable measurement foundation for subsequent high-precision correction of point cloud data.

9. The automated point cloud accuracy correction mapping vehicle according to claim 8, characterized in that, The wireless communication module adopts a 5G+wireless bridge dual-mode redundant communication architecture, combining low-latency transmission and long-distance stable communication capabilities. During operation, both the unmanned surveying vehicle and the surveying drone are independently equipped with high-precision timing modules that synchronously record data acquisition timestamps, achieving a unified time reference for both air and ground equipment. Leveraging the complementary advantages of the dual communication modes, low-latency, highly reliable two-way data interaction is achieved between the unmanned vehicle and the drone, synchronizing in real-time the true coordinate data of the prism collected by the automatic measurement robot with the raw point cloud data collected by the drone's onboard lidar. A timestamp matching correction algorithm is used to offset timing errors caused by asynchronous data acquisition between air and ground equipment, reducing time dimension deviations at the source and ensuring the timing consistency and matching accuracy of the point cloud correction data, providing accurate and synchronous raw data support for the global error correction model.

10. An automated point cloud accuracy correction method, applied to the automated point cloud accuracy correction mapping vehicle according to any one of claims 1-9, characterized in that, Includes the following steps: S1. Operation Deployment and Equipment Self-Check: The unmanned mapping vehicle autonomously travels to the target mapping area according to the preset mapping path and real-time environment model, and sequentially completes RTK millimeter-level positioning calibration, self-stabilizing platform horizontal attitude self-check, automatic measurement robot parameter initialization, and whole-machine communication link self-check to ensure that the reference parameters of each device are normal and communication is stable; the ground control station simultaneously binds and connects the unmanned mapping vehicle and the mapping drone through 5G+wireless bridge dual communication mode to complete equipment status inspection and initial parameter synchronization; the mapping drone takes off to the preset mapping airspace and uses airborne lidar and cooperative prism to perform full-area scanning and acquisition of the work site to obtain the original three-dimensional laser point cloud data of the site. S2. Measurement Point Acquisition and Automatic Calibration: The automatic calibration system for ground measurement points at the bottom of the vehicle body acquires real-time images of the ground below through a built-in camera module, identifies and captures fixed ground measurement points set up on the site; the system controls the laser grating sensor to project a grating grid to fully cover the measurement point area, calculates the horizontal deviation between the measurement robot's reference point and the ground measurement point, and uploads the deviation data to the vehicle controller in real time; the controller issues adjustment commands according to the feedback error, synchronously drives the entire unmanned surveying vehicle to move, and fine-tunes the attitude of the self-stabilizing gimbal platform, continuously correcting the position deviation in a closed loop until the measurement robot's reference point and the ground measurement point are precisely aligned, completing the automatic alignment calibration of the entire machine's measurement reference; S3. Air-Ground Collaborative Dynamic Tracking and Measurement: The surveying drone carries a collaborative prism and takes off to conduct a full-domain scanning operation. The drone uses its own satellite positioning module to obtain the airspace position coordinates in real time and actively transmits the satellite position coordinate data to the unmanned surveying vehicle through a wireless communication link as a guidance signal. After receiving the drone's satellite position information, the unmanned surveying vehicle quickly locks the target area using the depth camera of the automatic measurement robot and the prism tracking function. It actively tracks and locks the collaborative prism carried by the drone and obtains the drone's high-precision spatial true position in real time through dynamic tracking and measurement of the prism. Relying on the real-time attitude compensation capability of the self-stabilizing platform, the pitch, roll, and yaw angles are dynamically adjusted. With the help of the servo tracking mechanism, the automatic measurement robot continuously tracks and aims at the center of the collaborative prism with high precision. The three-dimensional spatial true coordinates of the prism and the drone are collected in real time in a high-frequency sampling mode to form a high-precision ground true value dataset. S4. Intelligent calculation of multi-dimensional error parameters: The vehicle controller completes air-to-ground data matching based on the synchronization timestamp, compares and analyzes the actual measured coordinates of the prism collected by the automatic measurement robot with the coordinate information of the corresponding position in the original point cloud data of the UAV, and intelligently calculates multi-dimensional error parameters such as global offset error, elevation system error, spatial angle distortion error and local point residual in the point cloud data, and constructs a dynamic error correction model adapted to the current work site. S5. Hierarchical Point Cloud Accuracy Correction: Based on the established global error correction model, the original laser point cloud data is subjected to overall coordinate correction, which uniformly corrects the overall positioning drift and overall offset of the point cloud; then, the single point data is screened and identified by the preset local residual threshold, and the distorted noise, abnormal floating points and deviation exceeding the limit points are automatically removed, so as to realize the point-by-point correction, noise reduction optimization and local distortion repair of the original laser point cloud data, and complete the all-round and multi-level accuracy optimization of the point cloud data. S6. Data storage and remote output: The high-precision 3D point cloud data after accuracy correction is stored locally with encryption, and simultaneously uploaded to the ground control station and surveying terminal in real time through dual redundant communication links, completing an integrated surveying and mapping operation process of fully automatic point cloud acquisition, error correction and data output in air-ground collaboration.

Images