High-precision dynamic 3D surveying and mapping system and method based on mobile ultrasonic composite sensor and fixed receiving array

Abstract

The invention discloses a high-precision dynamic 3D surveying and mapping system and method based on a mobile ultrasonic composite sensor and a fixed receiving array, and the system comprises a mobile composite sensing unit which comprises a 3D positioning module and is used for achieving the rapid preliminary surveying and mapping of the surface of an object; the surface high-precision surveying and mapping module is used for realizing high-precision 3D surface surveying and mapping; and an embedded controller. The fixed receiving array is arranged at the boundary of a surveying and mapping area, is in polyhedral geometric distribution and is used for capturing positioning signals; the data processing unit comprises a positioning thread processing unit and is used for real-time pose calculation; the imaging thread processing unit is used for surface topography reconstruction; and the data fusion module is used for mapping the imaging details to the positioning coordinate system. According to the invention, the surface microstructure of the object is captured through the mobile ultrasonic composite sensor and the fixed receiver, coordinate positioning and surface contour reconstruction are realized in combination with a cross-modal data fusion technology, the surveying and mapping precision and efficiency can be improved, and the method has good environmental robustness.

Description

Technical Field

[0001] The present invention belongs to the technical field of three-dimensional mapping, and particularly relates to a high-precision dynamic 3D mapping system and method based on a mobile ultrasonic composite sensor and a fixed receiving array. Background Art

[0002] In the technical field of three-dimensional mapping, with the continuous growth of the demand for high-precision, dynamic and complex environment-adaptive 3D mapping in many industries such as industrial manufacturing, cultural heritage protection, and scientific research exploration, the innovation of related technologies is imminent. At present, both traditional single ultrasonic positioning technology and emerging multi-sensor fusion technology have significant limitations.

[0003] Traditional ultrasonic 3D positioning systems, typically represented by the time-of-flight (ToF) scheme, are limited by the wavelength characteristics of ultrasonic waves themselves and have insurmountable obstacles in surface detail resolution. Generally speaking, the resolution it can achieve is generally greater than 1 mm. In the field of precision manufacturing, for example, in the quality inspection of micro-electromechanical system (MEMS) components, sub-millimeter or even finer surface defect detection is crucial for ensuring product performance. However, the resolution level of traditional ultrasonic 3D positioning systems cannot accurately identify these subtle defects, seriously affecting product quality control. Another example is in the work of cultural heritage protection. The delicate textures and subtle features formed on the surface of cultural relics over time are important carriers of their historical and artistic values. The low resolution of traditional ultrasonic 3D positioning systems causes a large amount of these key information to be lost during the mapping process, making it difficult to achieve accurate digital restoration and protection of cultural relics.

[0004] High-frequency ultrasonic imaging technology, when the frequency is increased to greater than 1 MHz, theoretically can obtain sub-millimeter-level surface textures, bringing new possibilities for high-precision surface detail mapping. However, this technology has two fatal defects in practical applications. First, it highly depends on the close contact between the probe and the surface of the object to be measured. In the face of objects with complex shapes, such as sculptures with irregular curved surfaces and mechanical parts with complex internal structures, it is difficult for the probe to fully and closely fit, resulting in data loss in some areas. Second, during the measurement process, this technology cannot synchronously obtain the pose information of the probe in space. In a dynamic mapping scenario, the position or pose of the object may change at any time. If the spatial pose of the probe cannot be known in real time, the surface texture data collected at different times cannot be accurately positioned and stitched in three-dimensional space, making it almost impossible to construct an accurate and continuous three-dimensional model.

[0005] In order to overcome the limitations of a single sensor in terms of measurement capabilities, multi-sensor fusion solutions have gradually become a research hotspot. However, in practical applications, heterogeneous data fusion faces the extremely difficult problem of data spatiotemporal alignment. Taking the fusion of lidar and ultrasonic sensors as an example, lidar has been widely used in terrain mapping, autonomous driving and other fields due to its high-precision distance measurement capabilities and fast data acquisition rate. However, its high equipment cost limits its large-scale application. At the same time, in severe weather (such as heavy rain, dust) or complex scenes with multiple obstructions, laser signals are easily interfered with, resulting in a significant decrease in measurement accuracy. Although ultrasonic sensors have the significant advantages of low cost and insensitivity to changes in ambient light, they have obvious shortcomings compared to lidar in terms of measurement accuracy and range.

[0006] Judging from the existing relevant literature, there are also many problems that need to be solved. The ultrasonic array imaging system proposed by patent US2022003XXXA1 has made innovative attempts in the design of imaging system architecture. However, in the actual dynamic mapping application scenario, the system has failed to effectively solve the key problem of real-time tracking of the probe's motion trajectory. In the process of dynamic mapping, the position and posture of the probe are constantly changing. If its motion trajectory cannot be grasped in real time and accurately, the collected ultrasonic signal will not be able to accurately correspond to the actual spatial position, which will lead to serious deviations in the imaging results, making it difficult to meet the needs of high-precision imaging in practical applications such as industrial inspection and architectural mapping. As recorded in the paper "Multi-modal 3D Reconstruction", the camera has unique advantages in obtaining rich texture information on the surface of objects by adopting a camera and ultrasound fusion solution. However, as an optical component, the camera has extremely high requirements for ambient light conditions. In a dark environment, due to insufficient light, clear and effective image data cannot be obtained. In occlusion scenes, the information of the occluded object cannot be collected by the camera at all, resulting in a large amount of data missing in the final 3D reconstruction result, which seriously undermines the integrity and accuracy of the reconstructed model and cannot provide reliable data support for subsequent analysis and decision-making.

[0007] In summary, the current shortcomings of ultrasonic 3D positioning and related multi-sensor fusion technologies in core performance indicators such as accuracy, flexibility and environmental adaptability have become key factors restricting their in-depth application in many fields. Therefore, there is an urgent need for a new technical solution that can effectively solve the above problems to meet the urgent needs of modern industry, scientific research, cultural protection and other fields for high-precision dynamic 3D mapping, and promote the further development and application expansion of three-dimensional mapping technology.

[0008] The information disclosed in this background section is only intended to enhance the overall understanding of the background of the present invention and should not be regarded as an admission or any form of suggestion that this information constitutes prior art known to those of ordinary skill in the art. Summary of the Invention

[0009] An object of the present invention is to provide a high-precision dynamic 3D mapping system and method based on a mobile ultrasonic composite sensor and a fixed receiving array, thereby overcoming the defects in the above-mentioned prior art.

[0010] To achieve the above object, the present invention provides a high-precision dynamic 3D mapping system based on a mobile ultrasonic composite sensor and a fixed receiving array, including: A mobile composite sensing unit, including: a 3D positioning module, which is an ultrasonic transmitter for spatial coordinate calculation and realizes rapid preliminary mapping by moving on the surface of an object; a surface high-precision mapping module, optionally equipped with an ultrasonic array module, a capacitive sensor module, an eddy current module, a laser module, and a contact three-coordinate measuring head, for realizing high-precision mapping of the object surface, such as a detachable high-frequency ultrasonic array module operating in the pulse echo mode; and an embedded controller; A fixed receiving array, arranged at the boundary of the mapping area and distributed in a polyhedral geometry for positioning signal capture; A data processing unit, including a positioning thread processing unit for real-time pose calculation; an imaging thread processing unit for surface topography reconstruction; and a data fusion module for mapping imaging details to the positioning coordinate system.

[0011] Further, preferably, the fixed receiving array is used for capturing positioning signals in the low-frequency band of 20 - 100 kHz.

[0012] Further, preferably, the fixed receiving array includes at least 3 high-sensitivity ultrasonic receivers.

[0013] Further, preferably, it is applicable to industrial inspection, medical organ modeling, and digital mapping of cultural relics.

[0014] The present invention also provides a high-precision dynamic 3D mapping method based on a mobile ultrasonic composite sensor and a fixed receiving array, including the following steps: Signal transmission and synchronization, using coding modulation technology to generate recognizable ultrasonic signals, and ensuring the synchronization of the mobile ultrasonic composite sensor and the fixed receiving array in time and space through a wireless clock synchronization mechanism, so that the transmitted signals can be accurately received and processed; Signal reception and preprocessing, the fixed receiving array receives ultrasonic signals, uses adaptive filtering technology to extract effective signals, and applies dynamic gain control (DGC) to suppress noise to improve signal quality; Time Difference of Arrival (TDOA) calculation uses Cross-Correlation Function (CCF) and Generalized Cross-Correlation with Phase Transform (GCC-PHAT) algorithm to accurately measure the time difference of signals arriving at each receiver. Three-dimensional coordinate calculation, based on the measured time delay difference, uses the TDOA positioning model to deduce the position of the emission source, and optimizes the position calculation result through a non-linear optimization algorithm to improve the positioning accuracy. Point cloud generation and registration, based on multi-view scanning data, uses the Iterative Closest Point (ICP) registration algorithm for point cloud registration, and combines Kalman filtering to optimize the registration process to achieve the fusion of multi-view scanning data and generate point cloud data. Surface reconstruction, for the generated discrete point cloud, uses Poisson reconstruction or Marching Cubes algorithm to generate a continuous surface model. Error compensation and optimization, uses the Random Sample Consensus (RANSAC) outlier rejection algorithm to remove outliers in the point cloud data, and uses adaptive grid subdivision technology to optimize the surface model to improve the model accuracy and integrity.

[0015] Furthermore, as an option, the Time Difference of Arrival (TDOA) calculation method is as follows: Use the Generalized Cross-Correlation with Phase Transform (GCC-PHAT) algorithm to process the signal, suppress multi-path interference through phase transform weighting, and calculate the time difference of signals arriving at different receivers to achieve nanosecond-level accuracy and millimeter-level positioning error. Anti-interference optimization, introduce polarization filtering technology to polarize the received signal and remove interference signals with polarization characteristics different from the target signal; at the same time, use spatial beamforming technology to adjust the weighting coefficients of each receiver according to the spatial position of the receiver array and the signal propagation direction to form a beam pointing to the target signal source, suppress Non-Line-of-Sight (NLOS) signals, and improve the accuracy of time difference calculation.

[0016] Furthermore, as an option, the three-dimensional coordinate calculation method is as follows: Hyperboloid equation construction, for each pair of receivers, based on the time difference of arrival of the signal Δt ij , construct the corresponding hyperboloid equation; use the spatial positions of each receiver as the foci, and use the relationship between the time difference and the signal propagation speed to determine the parameters in the hyperboloid equation, establishing the geometric relationship between the time difference of signal arrival and the spatial position. Objective function construction and nonlinear least squares optimization: Based on multiple hyperboloid equations constructed, an objective function for nonlinear least squares optimization is constructed. The coordinates x, y, z of the moving emission source are taken as unknown variables and incorporated into the objective function. By minimizing the value of the objective function, the three-dimensional coordinates of the moving emission source are solved. An iterative algorithm is used to continuously adjust the coordinate values to make the objective function converge to the minimum value, thereby obtaining the accurate three-dimensional coordinates x, y, z of the moving emission source and realizing the accurate calculation of the spatial position of the moving emission source from the TDOA data.

[0017] Further, preferably, the method for point cloud generation and registration is as follows: Coarse registration: In the point cloud data obtained by multiple scans, feature points are extracted. Based on the extracted feature points, the initial transformation matrix is calculated using SVD decomposition to preliminarily align the point cloud data to a unified coordinate system, providing a basis for subsequent fine registration. Fine registration: The KD tree acceleration algorithm is adopted to efficiently find the nearest neighbor point pairs in the point cloud data. For the found nearest neighbor point pairs, the optimal rigid body transformation is calculated to obtain the rotation matrix R and the translation vector t. The above steps are continuously iterated until the mean square error is less than 0.1 mm to achieve the accurate registration of the point cloud data. Global optimization: The graph optimization framework g2o is introduced to incorporate the poses of multiple frames of point clouds into the optimization scope. By jointly optimizing the poses of multiple frames of point clouds, the cumulative error in the point cloud registration process is eliminated to ensure that all point cloud data are accurately aligned to a unified coordinate system.

[0018] Further, preferably, the method for surface reconstruction is as follows: Model generation based on Poisson reconstruction: For discrete point cloud data with complex topological structures, the Poisson reconstruction algorithm is adopted. Through implicit function fitting, a watertight mesh is generated to construct a continuous surface model to meet the surface modeling requirements of objects with complex shapes. Model generation based on Marching Cubes: For discrete point cloud data in scenarios such as high-precision medical modeling, the Marching Cubes algorithm is used. By extracting the isosurface, a triangular mesh is generated to construct a continuous surface model to meet the application scenarios with extremely high requirements for model accuracy. Parameter optimization: Adaptive octree depth adjustment. According to the density of the discrete point cloud, the octree depth is dynamically adjusted. The default depth is set to 10. When the point cloud density is high, the octree depth is appropriately increased to improve the model detail accuracy. When the point cloud density is low, the octree depth is reduced to reduce the calculation amount and improve the modeling efficiency. Normal vector consistency correction, based on the principal component analysis (PCA) method, corrects the direction of the normal vectors of the discrete point cloud. By analyzing the principal components of the local neighborhood of the point cloud, the correct direction of the normal vector is determined to ensure that the generated continuous surface model meets the requirements in terms of normal vector consistency and improves the model quality.

[0019] Furthermore, as an optimization, the method for error compensation and optimization is as follows: Ambient sound speed correction, real-time collection of temperature and humidity data in the environment, based on the pre-established relationship model between temperature and humidity and sound speed, dynamically calculates the sound speed in the current environment, and uses the calculated sound speed to correct the relevant calculations of the point cloud data related to the propagation of ultrasonic signals to compensate for the measurement errors caused by changes in the ambient sound speed. Multipath suppression, spatial domain filtering operation, constructs the direction-of-arrival (DOA) constraint based on the geometric distribution characteristics of the receiving array. By analyzing and screening the DOA of the received signals, it suppresses the interference of multipath signals to ensure that the received signals mainly come from the target direction and improves the accuracy of point cloud data collection. Polarization filtering operation, emits orthogonal polarization signals, and performs polarization analysis on the received signals at the receiving end. For the interference signals generated by non-matching reflections, it uses the polarization characteristic differences to suppress them and reduces the influence of multipath signals on the original point cloud data. Outlier removal, uses the RANSAC algorithm to process the collected point cloud data. In the way of random sample consensus, it samples and fits the data multiple times to identify and remove the outliers. The threshold is set to 3 times the median absolute deviation (MAD), and the points that deviate from the normal data range by more than this threshold are determined as outliers and removed.

[0020] Compared with the prior art, the present invention has the following beneficial effects: The present invention captures the surface microstructure of an object through a mobile ultrasonic composite sensor and a fixed receiver, and combines cross-modal data fusion technology to achieve coordinate positioning and surface contour reconstruction, which can improve the mapping accuracy and efficiency and has good environmental robustness. The present invention uses a time-division and frequency-division composite multiplexing mechanism for the positioning signal and the imaging signal. After basic positioning and fine mapping processing, it can obtain fine surface microstructures and improve the mapping accuracy. The present invention performs optimization after surface contour reconstruction to eliminate errors and has higher mapping accuracy. Description of the Drawings

[0021] Figure 1 It is a schematic flow chart of the high-precision dynamic 3D mapping method based on the mobile ultrasonic composite sensor and the fixed receiving array of the present invention. Figure 2Schematic diagram of the application scenario of the high-precision dynamic 3D mapping system based on a mobile ultrasonic composite sensor and a fixed receiving array of the present invention (digital imaging of the full-size profile of the wing). Detailed implementation manners

[0022] The following describes the detailed implementation manners of the present invention in detail, but it should be understood that the protection scope of the present invention is not limited by the detailed implementation manners.

[0023] The following gives a brief overview of one or more aspects to provide a basic understanding of these aspects. This overview is not an exhaustive survey of all contemplated aspects, and is neither intended to identify key or decisive elements of all aspects nor to attempt to define the scope of any or all aspects. Its sole purpose is to present some concepts of one or more aspects in a simplified form as a prelude to the more detailed description that follows.

[0024] A high-precision dynamic 3D mapping system based on a mobile ultrasonic composite sensor and a fixed receiving array, comprising: A mobile composite sensing unit, including an ultrasonic transmitter with an operating frequency of 40 kHz and a half-power angle of 60°, for spatial coordinate calculation; a surface mapping module (optional): an ultrasonic array module, a capacitive sensor module, an eddy current module, a laser module, and a contact three-coordinate probe can be selected. For example: a detachable high-frequency ultrasonic array module with a specification of 0.5 - 20 MHz, the number of array elements ≥ 64, an aperture ≤ 10 mm, operating in the pulse echo mode; and an embedded controller, supporting dynamic switching of the transmission mode (positioning signal / imaging signal); A fixed receiving array, arranged at the boundary of the mapping area, with at least 3 high-sensitivity ultrasonic receivers arranged in a polyhedral geometry for capturing positioning signals in the low-frequency band of 20 - 100 kHz; A data processing unit, including a multi-threaded processing architecture, which includes a positioning thread processing unit for real-time pose calculation based on TDOA (update rate ≥ 100 Hz); an imaging thread processing unit for realizing μm-level surface topography reconstruction based on the synthetic aperture focusing (SAFT) algorithm; and a data fusion module for mapping imaging details to the positioning coordinate system through a spatio-temporal calibration matrix, and this module is based on an adaptive weighted fusion algorithm for feature points.

[0025] The mechanical design in this system meets the following conditions: A composite ultrasonic sensor for manual, automatic, and semi-automatic scanning; An automatic and semi-automatic scanning frame, a magnetic / vacuum adsorption base and a universal adjustment bracket to ensure that the imaging probe fits the curved surface An automatic supply device for acoustic coupling agent (switchable to air coupling or water immersion coupling mode for non-contact scenarios).

[0026] The present invention also provides a high-precision dynamic 3D mapping method based on a mobile ultrasonic composite sensor and a fixed receiving array, comprising the following steps: Signal transmission and synchronization: Using coding modulation technology (Chirp-BPSK) to generate recognizable ultrasonic signals, and through a wireless clock synchronization mechanism, ensuring the synchronization of the mobile ultrasonic composite sensor and the fixed receiving array in time and space, so that the transmitted signals can be accurately received and processed; Signal reception and preprocessing: The fixed receiving array receives ultrasonic signals, adopts adaptive filtering technology to extract effective signals, and uses dynamic gain control (DGC) to suppress noise to improve signal quality; Time difference of arrival (TDOA) calculation: Using cross-correlation analysis (CCF) and the generalized cross-correlation (GCC-PHAT) algorithm to accurately measure the time difference of signals arriving at each receiver; Three-dimensional coordinate calculation: Based on the measured time delay difference, deriving the position of the emission source with the help of the TDOA positioning model, and optimizing the position calculation result through the non-linear optimization (Levenberg-Marquardt) algorithm to improve the positioning accuracy; Point cloud generation and registration: Based on multi-view scanning data, using the iterative closest point (ICP) registration algorithm for point cloud registration, and combining Kalman filtering to optimize the registration process, realizing the fusion of multi-view scanning data and generating point cloud data; Surface reconstruction: For the generated discrete point cloud, adopting Poisson reconstruction or Marching Cubes algorithm to generate a continuous surface model; Error compensation and optimization: Using the RANSAC outlier rejection algorithm to remove outliers in the point cloud data, and adopting adaptive grid subdivision technology to optimize the surface model to improve the accuracy and integrity of the model.

[0027] a. In the above method, the time difference of arrival (TDOA) calculation method is as follows: Using the generalized cross-correlation (GCC-PHAT) algorithm to process the signals, suppressing multipath interference through phase transformation weighting, thereby calculating the time difference of signals arriving at different receivers to achieve nanosecond-level accuracy, corresponding to millimeter-level positioning error; Anti-interference optimization: Introducing polarization filtering technology to perform polarization processing on the received signals to remove interference signals with polarization characteristics different from the target signals; at the same time, adopting spatial domain beamforming technology, according to the spatial positions of the receiver array and the signal propagation direction, adjusting the weighting coefficients of each receiver to form a beam pointing to the target signal source, suppressing non-line-of-sight (NLOS) signals, and improving the accuracy of time difference calculation; The goal is to accurately calculate the time difference of signals arriving at different receivers, and the accuracy needs to reach the nanosecond level (corresponding to millimeter-level positioning error).

[0028] b. The three-dimensional coordinate calculation method is as follows: Hyperboloid equation construction: For each pair of receivers, based on the time difference of arrival Δt of the signals ij , construct the corresponding hyperboloid equation; taking the spatial positions of each receiver as foci, using the relationship between the time difference and the signal propagation speed, determine the parameters in the hyperboloid equation, and establish the geometric connection between the time difference of arrival and the spatial position; Objective function construction and nonlinear least squares optimization: Based on the constructed multiple hyperboloid equations, construct the objective function for nonlinear least squares optimization, incorporate the coordinates x, y, z of the mobile emitter as unknown variables into the objective function, and solve for the three-dimensional coordinates of the mobile emitter by minimizing the value of the objective function; use the iterative algorithm to continuously adjust the coordinate values to make the objective function converge to the minimum value, thereby obtaining the accurate three-dimensional coordinates x, y, z of the mobile emitter, and realizing the accurate calculation of the spatial position of the mobile emitter from the TDOA data; The goal is to calculate the three-dimensional coordinates x, y, z of the mobile emitter from the TDOA data.

[0029] c. The method for point cloud generation and registration is as follows: Coarse registration: In the point cloud data obtained by multiple scans, extract feature points; based on the extracted feature points, use SVD decomposition to calculate the initial transformation matrix, and preliminarily align the point cloud data to a unified coordinate system to provide a basis for subsequent fine registration; Fine registration: Adopt the KD tree acceleration algorithm to efficiently find the nearest neighbor point pairs in the point cloud data; for the found nearest neighbor point pairs, calculate the optimal rigid body transformation to obtain the rotation matrix R and the translation vector t; continuously iterate the above steps until the mean square error is less than 0.1 mm to achieve the accurate registration of the point cloud data; Global optimization: Introduce the graph optimization framework g2o, incorporate the poses of multiple frames of point clouds into the optimization scope, and eliminate the cumulative error in the point cloud registration process by jointly optimizing the poses of multiple frames of point clouds to ensure that all point cloud data are accurately aligned to a unified coordinate system; The goal is to calculate the three-dimensional coordinates x, y, z of the mobile emitter from the TDOA data.

[0030] d. The method for surface reconstruction is as follows: Model generation based on Poisson reconstruction: For the discrete point cloud data with complex topological structures, adopt the Poisson reconstruction algorithm, and generate a watertight mesh and construct a continuous surface model by means of implicit function fitting to meet the surface modeling requirements of objects with complex shapes; Model generation based on Marching Cubes: For discrete point cloud data in scenarios such as high-precision medical modeling, the Marching Cubes algorithm is used. By extracting the isosurface, a triangular mesh is generated to construct a continuous surface model, meeting application scenarios with extremely high requirements for model accuracy. Parameter optimization and adaptive adjustment of octree depth: According to the density of the discrete point cloud, the depth of the octree is dynamically adjusted. The default depth is set to 10. When the point cloud density is high, the octree depth is appropriately increased to improve the detail accuracy of the model; when the point cloud density is low, the octree depth is reduced to reduce the computational amount and improve the modeling efficiency. Normal vector consistency correction: Based on the principal component analysis (PCA) method, the direction of the normal vector of the discrete point cloud is corrected. By analyzing the principal components of the local neighborhood of the point cloud, the correct direction of the normal vector is determined to ensure that the generated continuous surface model meets the requirements in terms of normal vector consistency and improves the model quality. The goal is to generate a continuous surface model from discrete point cloud.

[0031] e. The method of error compensation and optimization is as follows: Environmental sound speed correction: The temperature and humidity data in the environment are collected in real time. Based on the pre-established relationship model between temperature, humidity and sound speed, the sound speed in the current environment is dynamically calculated. The calculated sound speed is used to correct the relevant calculations of the point cloud data related to the propagation of ultrasonic signals, compensating for the measurement errors caused by the change of environmental sound speed. Multipath suppression - spatial domain filtering operation: According to the geometric distribution characteristics of the receiving array, a direction of arrival (DOA) constraint is constructed. By analyzing and screening the DOA of the received signals, the interference of multipath signals is suppressed to ensure that the received signals mainly come from the target direction and improve the accuracy of point cloud data collection. Polarization filtering operation: Orthogonal polarization signals are transmitted. At the receiving end, the polarization analysis of the received signals is carried out. For the interference signals generated by non-matching reflections, the polarization characteristics difference is used to suppress them, reducing the influence of multipath signals on the original point cloud data. Outlier removal: The RANSAC algorithm is used to process the collected point cloud data. In the way of random sample consensus, the data is sampled and model-fitted multiple times to identify and remove outliers. The threshold is set to 3 times the median absolute deviation (MAD). The points that deviate from the normal data range by more than this threshold are determined as outliers and removed.

[0032] The core points of the present invention are as follows: (1) The modal cooperation mechanism is adopted. Specifically, time-division and frequency-division composite multiplexing is used for the positioning signal and the imaging signal, which is divided into a basic positioning stage: transmitting a 40 kHz coded Chirp signal; and a fine mapping stage: switching to short pulses above 0.5 MHz, and obtaining the surface microstructure through SAFT processing of the receiving array.

[0033] (2) Cross-modal calibration technology. Specifically, a reference reflector (metal ball, diameter 0.5 mm) is used to synchronously capture the positioned signal and the imaging signal, realizing: coordinate system alignment error compensation and system delay calibration.

[0034] (3) Intelligent workflow control. Specifically, an adaptive scanning strategy based on point cloud density feedback: when the curvature change of the positioning data > threshold, automatically trigger local high-frequency imaging; the imaging area result feeds back to the positioning algorithm to optimize the subsequent path planning.

[0035] The present invention is used in the fields of industrial inspection, medical organ modeling, cultural relic digital mapping, etc., which will be described below in conjunction with specific embodiments.

[0036] Embodiment 1: Application in the quality inspection of industrial turbine blades: Turbine blade contour scanning to detect the deformation amount (accuracy ±0.05 mm). Specifically: Equipment setup: Place the fixed receiving array around the turbine blade on a stable measurement platform to ensure its uniform distribution around the blade and enable omnidirectional reception of ultrasonic signals; the mobile ultrasonic composite sensor is installed on a robotic arm that can precisely control the movement trajectory. The robotic arm has multi-axis linkage function and can scan the turbine blade according to a preset path; Initialize the system settings, including calibrating the transmission and reception parameters of the ultrasonic sensor, adjusting the sensitivity of the fixed receiving array, and confirming the movement accuracy of the robotic arm, etc., to ensure that the system is in the best working state; Scanning process: According to the complex shape of the turbine blade, use the motion control and positioning module to plan the movement trajectory of the robotic arm. This trajectory design fully considers the curvature change of the blade to ensure that the mobile ultrasonic composite sensor can comprehensively cover the blade surface while maintaining a certain distance. During the scanning process, the mobile ultrasonic composite sensor emits ultrasonic signals at a high frequency. After the signals are reflected by the turbine blade surface, they are captured by each receiving unit in the fixed receiving array; The signal processing and analysis module processes the received ultrasonic signals in real time, generates recognizable signals through the coding modulation (Chirp-BPSK) technology, combines with the wireless clock synchronization mechanism to ensure the accuracy of the signals in space and time, uses the generalized cross-correlation (GCC-PHAT) algorithm to accurately calculate the time difference of the signals arriving at different receivers, and simultaneously introduces anti-interference optimization measures such as polarization filtering and spatial beamforming to effectively suppress multipath interference and non-line-of-sight (NLOS) signals, thus ensuring the accuracy of the time difference calculation; According to the TDOA positioning model, the distance information between the mobile ultrasonic composite sensor and each point on the surface of the turbine blade is deduced from the calculated time delay difference. The nonlinear least squares optimization method is used to iteratively optimize the distance calculation results to further improve the measurement accuracy, and finally achieve high-precision scanning of the turbine blade profile with an accuracy of up to ±0.05 mm; Deformation detection: Compare and analyze the currently scanned contour data of the turbine blade with the pre-stored standard contour data. Use professional data analysis software to calculate the contour deviation value point by point. For each detection point, if the deviation between its actual contour and the standard contour exceeds ±0.05 mm, it is determined that the area where the point is located has deformation. By calculating the deviations of all detection points on the entire blade surface, the deformation amount of the turbine blade is comprehensively detected; Generate a deviation chromatogram in real time and mark the out-of-tolerance areas. Specifically: Data processing and conversion: The signal processing and analysis module sorts and converts the detected contour deviation data of the turbine blade, maps the deviation value of each detection point to a preset color space range. For example, it is set that the deviation within ±0.01 mm is green, ±0.01 - ±0.03 mm is yellow, ±0.03 - ±0.05 mm is orange, and exceeding ±0.05 mm is red. In this way, the abstract deviation numerical values are converted into intuitive color information; Deviation chromatogram generation: The data storage and display module generates the deviation chromatogram of the turbine blade in real time according to the above color mapping rules. Based on the three-dimensional model of the turbine blade, the color corresponding to each detection point is filled into the corresponding position to form a color image that intuitively shows the deviation distribution on the blade surface. During the generation process, the complex shape and surface characteristics of the blade are fully considered to ensure that the color filling is accurately matched with the actual position, enabling the operator to clearly see the deviation status of each part of the blade; Out-of-tolerance area marking: In the deviation chromatogram, for the areas where the color is displayed as red (i.e., the deviation exceeds ±0.05 mm), the data storage and display module automatically marks them with eye-catching graphic identifiers, such as drawing a red border around the out-of-tolerance area or marking it with a flashing warning symbol, so that the operator can quickly identify the areas that need to be focused on. At the same time, beside or below the chromatogram, the specific position coordinates and deviation values of the out-of-tolerance area are listed in text form, providing detailed data support for subsequent repair and adjustment work. By generating the deviation chromatogram in real time and marking the out-of-tolerance areas, it can help industrial quality inspection personnel quickly and accurately evaluate the quality status of turbine blades, timely discover and handle existing problems, and effectively improve the efficiency of industrial production and product quality.

[0037] Example 2: Applications in the field of digital twins: Full-scale modeling of factory equipment, combined with CAD data comparison and analysis. Specifically: Preliminary preparation: Collect detailed information about the factory equipment, including equipment type, structure drawings, working principles, and previous maintenance records, etc. For the target equipment, determine its key measurement parts and areas of focus, so as to collect data more targeted during subsequent measurements; Check the transmission and reception performance of the mobile ultrasonic composite sensor to ensure that it can stably and accurately transmit and capture ultrasonic signals. At the same time, calibrate the spatial position and reception sensitivity of the fixed receiving array to ensure that each receiving unit can receive signals evenly and efficiently. In addition, test the motion control and positioning module to ensure that it can accurately control the mobile ultrasonic composite sensor to move along the predetermined trajectory; Equipment scanning and data acquisition: According to the shape and structure characteristics of the equipment, use the motion control and positioning module to plan the movement path of the mobile ultrasonic composite sensor. This path is designed to comprehensively cover the surface of the equipment. For parts with complex shapes, such as equipment with irregular curved surfaces or internal structures, adopt a layered and segmented scanning method to ensure that every detail can be measured; During the scanning process, the mobile ultrasonic composite sensor emits ultrasonic signals modulated by chirp-BPSK. After the signals are reflected by the surface of the device, they are received by the fixed receiving array. The signal processing and analysis module processes the received signals in real time, uses the generalized cross-correlation (GCC-PHAT) algorithm to accurately calculate the time difference of the signals arriving at different receivers, and solves the distance information between the sensor and each point on the device surface through the TDOA positioning model. At the same time, anti-interference technologies such as polarization filtering and spatial beamforming are introduced to effectively suppress multipath interference and non-line-of-sight (NLOS) signals, ensuring high-precision distance measurement. After full-scale scanning, a large amount of discrete point cloud data of the device surface is obtained. These data reflect the actual size and shape information of the device in the current state; Point cloud data processing and model construction: The collected point cloud data is transmitted to the data processing software. First, noise reduction processing is carried out to remove abnormal points caused by environmental interference or measurement errors. Then, using the point cloud registration algorithm, the point cloud data collected from different perspectives is stitched and aligned to be in the same coordinate system. On this basis, surface reconstruction algorithms such as Poisson reconstruction or Marching Cubes algorithm are used to convert the discrete point cloud data into a continuous surface model, thus realizing the full-scale modeling of factory equipment; Combined with CAD data comparison and analysis: Obtain the CAD design drawing data of the device and import it into the same software platform as the point cloud data processing. Through specific algorithms and tools, the constructed actual model of the device is accurately aligned and compared with the CAD model. During the comparison process, the deviation values between the models are calculated, including dimensional deviation, shape deviation, etc. For each detection point, analyze the difference between its actual position and the CAD design position, and present the deviation values in an intuitive way, such as through color coding or numerical annotation; Through comprehensive comparison and analysis, it is possible to clearly understand the compliance of the device with the design standards during manufacturing and use. For areas with large deviations found, further analyze the reasons for their occurrence, which may include manufacturing process errors, wear and deformation caused by long-term use, etc. These analysis results provide an important basis for subsequent equipment maintenance and optimization.

[0038] Dynamically update the worn area and predict the maintenance cycle. Specifically: Regular monitoring and data collection: At regular time intervals, regularly use the high-precision dynamic 3D mapping system based on the mobile ultrasonic composite sensor and the fixed receiving array to repeatedly scan the factory equipment to obtain the state data of the equipment at different time points. The process and data collection method of each scan are the same as those during the initial modeling to ensure the consistency and comparability of the data; Wear area analysis and dynamic update. Compare the newly collected device point cloud data with the initial modeling data and the previous monitoring data each time. By calculating the differences between the models, identify the areas on the device surface that may be worn. For the worn areas, use parameter optimization methods such as adaptive octree depth adjustment and normal vector consistency correction to perform more accurate analysis and description of the details of the worn parts; According to the changes in the worn areas, update the digital twin model of the device in real time. In the model, mark the position, shape, and degree changes of the worn areas in an obvious way, such as using different colors or textures to represent different degrees of wear. By dynamically updating the worn areas, the wear development trend of the device can be intuitively tracked, and the health status of the device can be grasped in a timely manner; Maintenance cycle prediction. Based on the dynamic monitoring of the device worn areas and the accumulation of historical data, use data analysis and machine learning algorithms to establish a device wear prediction model. This model comprehensively considers various factors such as the operating time, workload, environmental factors, and wear change trend of the device to predict the wear situation of the device in the future for a period of time; According to the results of the wear prediction model, combined with the design life and safe operation standards of the device, formulate a reasonable maintenance cycle. When it is predicted that the wear degree of the device is about to reach the threshold that affects its normal operation or safety performance, the system automatically issues a warning message to remind the maintenance personnel to perform device maintenance and repair in a timely manner, avoiding production interruptions and safety accidents caused by device failures. By dynamically updating the worn areas and predicting the maintenance cycle, it provides strong support for the intelligent management and preventive maintenance of factory equipment, effectively improving the reliability and production efficiency of the equipment, and reducing the maintenance cost and production risk.

[0039] Example 3: Application in medical plastic surgery: Facial 3D contour reconstruction to assist in customizing maxillofacial prostheses. Specifically: Patient facial data acquisition. Guide the patient to be comfortably within the measurement area, reasonably arrange the fixed receiving array around the patient's face to ensure that ultrasonic signals can be received omnidirectionally. At the same time, the operator controls the device equipped with a mobile ultrasonic composite sensor to scan the patient's face along a pre-planned path. During the scanning process, fully consider the complex curved surface characteristics of the face and focus on scanning key parts such as the eye sockets, cheekbones, and mandible to ensure comprehensive and accurate facial surface information is obtained; The mobile ultrasonic composite sensor emits ultrasonic signals encoded and modulated (Chirp-BPSK). After being reflected by the patient's facial skin and bone surface, these signals are quickly captured by the fixed receiving array. The signal processing and analysis module immediately processes the received signals. Using the generalized cross-correlation (GCC-PHAT) algorithm, it accurately calculates the time difference of the signals arriving at different receivers. At the same time, by introducing polarization filtering and spatial domain beamforming technologies, it effectively resists environmental noise and multipath interference, ensuring the high accuracy of the time difference calculation. Then, based on the TDOA positioning model, it accurately calculates the distance information between the sensor and each point on the face, thereby obtaining a large amount of discrete point cloud data reflecting the current state of the patient's face; Point cloud data processing and facial model construction. The collected facial point cloud data is transmitted to professional medical imaging processing software. First, using advanced denoising algorithms, abnormal points generated due to environmental interference, minor patient movements, etc. are removed to ensure the purity of the data. Then, point cloud registration algorithms are used to seamlessly splice and accurately align the point cloud data obtained from different scanning perspectives, making them unified in the same coordinate system. On this basis, surface reconstruction algorithms suitable for the complex facial structure such as Poisson reconstruction or Marching Cubes are selected to convert the discrete point cloud data into a continuous, smooth, and high-precision facial three-dimensional surface model; Customized assistance for the maxillofacial prosthesis. The constructed three-dimensional facial model of the patient is interacted with the maxillofacial prosthesis design software. The maxillofacial prosthesis design engineer designs the maxillofacial prosthesis personalizedly based on the patient's specific condition, oral anatomical structure, and overall facial aesthetic requirements in the software. By accurately measuring parameters such as the dimensions and curvatures of relevant facial parts and combining CAD technology, the shape of the maxillofacial prosthesis that fits the patient's oral and facial morphology is designed. For example, for patients with jawbone defects caused by tumor resection, the design of the maxillofacial prosthesis needs to accurately match the edge shape of the remaining jawbone, and at the same time consider the overall facial symmetry and chewing function requirements. After the design is completed, using 3D printing technology, the maxillofacial prosthesis is accurately manufactured according to the design data, greatly improving the fit between the maxillofacial prosthesis and the actual situation of the patient's face, and providing a more comfortable and better-functioning maxillofacial prosthesis repair plan for the patient.

[0040] Intraoperative real-time navigation to ensure the implant matching degree > 99%. Specifically: Preoperative preparation and model fusion. Before the operation, the high-precision dynamic 3D mapping system based on the mobile ultrasonic composite sensor and the fixed receiving array is used again to scan the patient's face to obtain the latest facial state data, and it is fused and analyzed with the preoperative constructed facial three-dimensional model and the maxillofacial prosthesis model. Through accurate matching and comparison, the consistency between the maxillofacial prosthesis and the actual situation of the patient's face is further confirmed, and possible minor differences are adjusted and optimized in advance. At the same time, the fused model data is transmitted to the surgical navigation system to provide an accurate reference basis for intraoperative real-time navigation; Intraoperative real-time measurement and navigation. During the operation, the mobile ultrasonic composite sensor and the fixed receiving array are flexibly arranged around the operation area to ensure the real-time measurement of the spatial position changes of the operation site. As the operation progresses, when an implant such as a jaw implant needs to be implanted, the operator moves the ultrasonic composite sensor to obtain the relative position relationship data between the implant and the surrounding tissues in real time. The signal processing and analysis module quickly processes and analyzes these data, calculates the deviation between the current actual position of the implant and the ideal position planned before the operation. According to the calculated deviation value, the surgical navigation system provides real-time navigation information to the surgeon in an intuitive manner, such as displaying the direction and distance that the implant should be adjusted in the form of virtual lines or marks in the surgical field of view. The doctor accurately adjusts the position of the implant based on the navigation information to ensure that the implant fits closely with the patient's facial bones, tissues, etc. Through continuous real-time measurement and navigation, the position of the implant is continuously optimized until the matching degree of the implant reaches more than 99%, effectively improving the success rate and treatment effect of the operation and reducing the occurrence of postoperative complications.

[0041] The foregoing description of the specific exemplary embodiments of the present invention is for the purposes of illustration and exemplification. These descriptions are not intended to limit the invention to the precise forms disclosed, and it is apparent that many changes and variations are possible in light of the above teaching. The purpose of selecting and describing the exemplary embodiments is to explain the specific principles of the invention and its practical applications, so that those skilled in the art can implement and utilize the various different exemplary embodiments of the invention, as well as various different selections and changes. The scope of the invention is intended to be defined by the claims and their equivalents.

Claims

1. A high-precision dynamic 3D mapping system based on a mobile ultrasonic composite sensor and a fixed receiving array, characterized in that, Comprising: A mobile composite sensing unit, comprising: a 3D positioning module, the 3D positioning module being an ultrasonic transmitter for spatial coordinate calculation, moving on the surface of an object to achieve rapid preliminary mapping; a surface high-precision mapping module, optionally including an ultrasonic array module, a capacitive sensor module, an eddy current module, a laser module, and a contact-type coordinate measuring probe, for achieving high-precision mapping of the object surface; and an embedded controller; A fixed receiving array, arranged at the boundary of the mapping area, in a polyhedral geometric distribution, for positioning signal capture; A data processing unit, including a positioning thread processing unit for real-time pose calculation; an imaging thread processing unit for surface topography reconstruction; and a data fusion module for mapping imaging details to the positioning coordinate system.

2. The high-precision dynamic 3D mapping system based on a mobile ultrasonic composite sensor and a fixed receiving array according to claim 1, wherein: The fixed receiving array is used for capturing positioning signals in the 20 - 200 kHz low-frequency band.

3. The high-precision dynamic 3D mapping system based on a mobile ultrasonic composite sensor and a fixed receiving array according to claim 1, wherein: The fixed receiving array includes at least 3 high-sensitivity ultrasonic receivers.

4. The high-precision dynamic 3D mapping system based on a mobile ultrasonic composite sensor and a fixed receiving array according to claim 1, characterized in that: Applicable to industrial inspection, medical organ modeling, and digital mapping of cultural relics.

5. A high-precision dynamic 3D mapping method based on a mobile ultrasonic composite sensor and a fixed receiving array, characterized in that, Including the following steps: Signal transmission and synchronization, using coding modulation technology to generate recognizable ultrasonic signals, and ensuring the synchronization of the mobile ultrasonic composite sensor and the fixed receiving array in time and space through a wireless clock synchronization mechanism, so that the transmitted signals can be accurately received and processed; Signal reception and preprocessing, the fixed receiving array receives ultrasonic signals, uses adaptive filtering technology to extract effective signals, and uses dynamic gain control DGC to suppress noise to improve signal quality; Time Difference of Arrival (TDOA) calculation, using cross-correlation analysis (CCF) and Generalized Cross-Correlation (GCC-PHAT) algorithms to accurately measure the time difference of the signal arriving at each receiver; Three-dimensional coordinate calculation, based on the measured time delay difference, deriving the position of the emission source with the help of the TDOA positioning model, and optimizing the position calculation result through a non-linear optimization algorithm to improve the positioning accuracy; Point cloud generation and registration, based on multi-view scanning data, using the Iterative Closest Point (ICP) registration algorithm for point cloud registration, and optimizing the registration process in combination with Kalman filtering to achieve the fusion of multi-view scanning data and generate point cloud data; Surface reconstruction, for the generated discrete point cloud, using Poisson reconstruction or Marching Cubes algorithm to generate a continuous surface model; Error compensation and optimization, using the RANSAC outlier rejection algorithm to remove outliers in the point cloud data, and using adaptive mesh subdivision technology to optimize the surface model to improve the accuracy and integrity of the model.

6. The high-precision dynamic 3D mapping method based on a mobile ultrasonic composite sensor and a fixed receiving array according to claim 5, wherein The method for calculating the Time Difference of Arrival (TDOA) is as follows: Using the Generalized Cross-Correlation (GCC-PHAT) algorithm to process the signal, suppressing multipath interference through phase transformation weighting, thereby calculating the time difference of the signal arriving at different receivers to achieve nanosecond-level accuracy, corresponding to millimeter-level positioning error; Anti-interference optimization: Introduce the polarization filtering technology to perform polarization processing on the received signals, removing interference signals with polarization characteristics different from those of the target signal. At the same time, adopt the spatial domain beamforming technology to adjust the weighting coefficients of each receiver according to the spatial positions of the receiver array and the signal propagation direction, forming a beam pointing to the target signal source, suppressing the non-line-of-sight (NLOS) signals, and improving the accuracy of time difference calculation.

7. The high-precision dynamic 3D mapping method based on a mobile ultrasonic composite sensor and a fixed receiving array according to claim 5, characterized in that, The three-dimensional coordinate calculation method is as follows: Hyperboloid equation construction. For each pair of receivers, based on the time difference of arrival Δt of the signals ij , construct the corresponding hyperboloid equation; taking the spatial positions of each receiver as the foci, using the relationship between the time difference and the signal propagation speed, determine the parameters in the hyperboloid equation, and establish the geometric connection between the time difference of arrival and the spatial position; Objective function construction and non-linear least squares optimization: Based on the constructed multiple hyperbolic equations, construct the objective function of non-linear least squares optimization, incorporate the coordinates x, y, z of the mobile transmitter as unknown variables into the objective function, and solve the three-dimensional coordinates of the mobile transmitter by minimizing the value of the objective function. Use the iterative algorithm to continuously adjust the coordinate values to make the objective function converge to the minimum value, thereby obtaining the accurate three-dimensional coordinates x, y, z of the mobile transmitter and realizing the accurate calculation of the spatial position of the mobile transmitter from the TDOA data.

8. The high-precision dynamic 3D mapping method based on a mobile ultrasonic composite sensor and a fixed receiving array according to claim 5, wherein The method for point cloud generation and registration is as follows: Coarse registration: Extract feature points from the point cloud data obtained by multiple scans. Based on the extracted feature points, calculate the initial transformation matrix using the SVD decomposition to preliminarily align the point cloud data to a unified coordinate system, providing a basis for subsequent fine registration. Fine registration: Adopt the KD tree acceleration algorithm to efficiently find the nearest neighbor point pairs in the point cloud data. For the found nearest neighbor point pairs, calculate the optimal rigid body transformation to obtain the rotation matrix R and the translation vector t. Continuously iterate the above steps until the mean square error is less than 0.1 mm to achieve the accurate registration of the point cloud data. Global optimization: Introduce the graph optimization framework g2o, incorporate the poses of multiple frames of point clouds into the optimization scope, and eliminate the cumulative error in the point cloud registration process by jointly optimizing the poses of multiple frames of point clouds to ensure that all point cloud data are accurately aligned to a unified coordinate system.

9. The high-precision dynamic 3D mapping method based on a mobile ultrasonic composite sensor and a fixed receiving array according to claim 5, wherein The method for surface reconstruction is as follows: Model generation based on Poisson reconstruction: For the discrete point cloud data with complex topological structures, adopt the Poisson reconstruction algorithm to generate a watertight mesh and construct a continuous surface model by means of implicit function fitting to meet the surface modeling requirements of objects with complex shapes. Model generation based on Marching Cubes: Apply the Marching Cubes algorithm to the discrete point cloud data in scenarios such as high-precision medical modeling, and generate a triangular mesh and construct a continuous surface model by extracting the isosurface to meet the application scenarios with extremely high requirements for model accuracy. Parameter optimization: Adaptive octree depth adjustment, dynamically adjust the octree depth according to the density of the discrete point cloud. The default depth is set to 10. When the point cloud density is high, appropriately increase the octree depth to improve the model detail accuracy. When the point cloud density is low, reduce the octree depth to reduce the computational amount and improve the modeling efficiency. Normal vector consistency correction: Based on the principal component analysis (PCA) method, correct the direction of the normal vectors of the discrete point cloud. Determine the correct direction of the normal vectors by analyzing the principal components of the local neighborhood of the point cloud to ensure that the generated continuous surface model meets the requirements in terms of normal vector consistency and improves the model quality.

10. The high-precision dynamic 3D mapping method based on a mobile ultrasonic composite sensor and a fixed receiving array according to claim 5, wherein The error compensation and optimization method is as follows: Ambient sound velocity correction: collect temperature and humidity data in the environment in real time, dynamically calculate the sound velocity in the current environment based on the pre-established temperature, humidity and sound velocity relationship model, and use the calculated sound velocity to correct the point cloud data related calculations involving ultrasonic signal propagation to compensate for the measurement error caused by changes in ambient sound velocity; Multipath suppression, spatial filtering operation, based on the geometric distribution characteristics of the receiving array, constructs the direction of arrival (DOA) constraint, and suppresses multipath signal interference by analyzing and screening the direction of arrival of the received signal, ensuring that the received signal mainly comes from the target direction, thereby improving the accuracy of point cloud data acquisition; Polarization filtering operation, transmits orthogonal polarization signals, performs polarization analysis on the received signals at the receiving end, and suppresses the interference signals generated by non-matching reflections by using the difference in polarization characteristics, thereby reducing the impact of multipath signals on the original point cloud data; For outlier removal, the RANSAC algorithm is used to process the collected point cloud data. The data is sampled and model fitted multiple times in a random sampling consistency manner to identify and remove outliers. The threshold is set to 3 times the median absolute deviation (MAD). Points that deviate from the normal data range by more than the threshold are identified as outliers and removed.

Images