A method and system for calculating the medium loss of indoor structures based on three-dimensional point cloud profiles and wireless signal samples

Abstract

The application discloses a kind of indoor construction medium loss calculation method and system based on three-dimensional point cloud profile and wireless signal sample, belongs to information communication, embodied intelligent robot field. Embodied intelligent car equipped with depth camera, two-dimensional laser radar is collected three-dimensional point cloud by SLAM, and the wireless signal of three height layers in base_link coordinate system of vehicle body is synchronously acquired;Figure 3 is the medium loss calculation process schematic diagram of the application, and the logic of complete set of data processing and parameter solving is presented. Grid map is constructed and construction profile is extracted, and after wireless signal registration, special sample is filtered according to construction. The application solves the problem that traditional scheme adopts fixed material parameter, and is difficult to adapt to real indoor electromagnetic environment. Single-point penetration loss is calculated based on free-space loss model, and single-frequency equivalent attenuation parameter is obtained by statistical optimization, and loss matrix is generated by traversing multiple frequency bands. Without manual labeling construction, scene loss is calculated by fusing multipath effect, standing wave and signal incidence angle, which can adapt to various indoor constructions and is suitable for radio electromagnetic model calibration, with the advantages of automation, multi-communication system compatibility and reproducible calculation results.

Description

Technical Field

[0001] This invention relates to the fields of information and communication technology and embodied intelligent robot technology, specifically to a method for extracting structural medium loss parameters in indoor wireless signal propagation modeling. Background Technology

[0002] In indoor wireless propagation modeling (Lee empirical model, ray tracing simulation model, etc.), the dielectric penetration loss parameters of structures such as walls, partitions, and furniture are the core factors affecting model accuracy. Existing technologies for obtaining dielectric loss parameters mainly fall into two categories, both of which have significant drawbacks: 1. Standard / Empirical Database Lookup Method: Based on the electromagnetic parameters (dielectric constant, conductivity) of building materials given by communication standards such as ITU-R P.2040, the loss is theoretically calculated, or the industry-standard fixed empirical loss values ​​are directly adopted. 2. Laboratory sample measurement method: Electromagnetic attenuation parameters of a single building material sample are measured in a shielded darkroom.

[0003] The loss values ​​obtained by the two methods mentioned above are static standard values, which only reflect ideal single material properties and cannot be adapted to real indoor environments. In actual scenarios, structures of the same material will be affected by construction technology, material aging, moisture content, multipath reflection, spatial standing waves, signal incident angle, and surrounding electromagnetic interference coupling, resulting in a large deviation between actual signal attenuation and standard static values. Existing models directly fix the dielectric loss as an inherent physical constant of the material, which cannot characterize the comprehensive attenuation characteristics in a scenario.

[0004] In addition, the few existing technologies that combine 3D point clouds with wireless signals are only used for indoor signal strength heat map visualization and do not have the functions of extracting the outline of structures or inverting the loss of individual structures; ray tracing simulation technology still relies on a preset material parameter library, and the simulation accuracy is limited by the completeness of the material library, and it cannot automatically solve the equivalent loss of the scene based on the actual measured data.

[0005] In summary, existing technologies lack a complete solution that can automatically, scenario-based, and structure-specifically extract comprehensive dielectric loss parameters of the actual indoor electromagnetic environment. technical terms

[0006] To clearly define the technical solution of this invention, the core terms are uniformly defined as follows.

[0007] Near-ground space region: The typical height range is 0 to 2.2m, covering all activity heights of humans in indoor standing, walking, and sitting postures, and is also the mainstream working height range of mobile phones and IoT terminals; the reference projection of the vehicle's base_link coordinate system is about 10cm above the ground, and the collected heights are all Z-axis offsets in this coordinate system. After conversion, the overall range falls within 0 to 2.2m. All measurement and calculation processes of this invention are carried out in this region, and the height values ​​are only preferred examples and do not constitute a limitation on the scope of protection.

[0008] 3D Raster Map: A spatial index structure obtained by discretizing the near-ground indoor space into cubic voxels of preset sizes. Each raster stores spatial coordinates and occupancy status; the preferred preset size is 20cm × 20cm × 20cm. The raster map is generated independently without downsampling or deleting the original point cloud, which is fully preserved for contour extraction. Smaller sizes (5cm, 10cm) can be selected to improve spatial registration accuracy, or larger sizes can be selected to reduce computational overhead, depending on the accuracy requirements of the indoor scene.

[0009] Point cloud clustering: The preferred Euclidean clustering algorithm is used to divide consecutive neighboring points into point cloud clusters of the same structure based on the spatial location and density characteristics of three-dimensional points.

[0010] Boundary detection: Extract the outer contour boundary points of the structure based on the clustered point cloud, and output the spatial coordinates, geometric dimensions, and surface normal vector of the structure.

[0011] Spatial reference point: The reference point for free space path loss calculation is selected with priority given to the near-ground geometric center of the measured area; if the geometric center of a super-large space such as a grand theater or stadium is outside the coverage of the measurement trajectory, it can be moved to the nearest reachable point of the measurement trajectory or the geometric centroid of the trajectory, which is only used as a distance calculation reference and does not change the calculation logic.

[0012] Spatial registration: Unify the coordinate system of multi-source data, convert the pose coordinates of wireless signal samples into grid coordinates of three-dimensional grid map, and establish the mapping relationship between samples and spatial grid.

[0013] Free-space path loss model: A wireless signal attenuation model determined solely by propagation distance and signal frequency in an ideal environment with no reflection, diffraction, or absorption. It is used to calculate the theoretical reference received power.

[0014] RSRP (Reference Signal Received Power), RSSI (Reference Signal Received Strength), RSRQ (Reference Signal Received Quality), SINR (Signal-to-Interference-plus-Noise Ratio), and frequency points: RSRP, RSSI, RSRQ, and SINR are common measurement indicators for mobile communication and wireless local area networks, used to characterize the measured signal power and link quality at the measurement point; the frequency point represents the center operating frequency of the wireless signal, used to distinguish different frequency bands, and provides a basis for frequency band solving and generating the dielectric loss parameter matrix in this invention.

[0015] Lumped equivalent attenuation parameter: The dielectric loss characteristic value output by this invention is not the material's inherent dielectric constant or conductivity physical constant. It is the sum of all additional attenuations after removing free space distance attenuation, including penetration absorption, multipath reflection cancellation, diffraction, standing wave superposition, and incident angle coupling. It is only applicable to the current measurement scenario propagation model calibration. Summary of the Invention

[0016] The calculation method of this invention is a technical solution.

[0017] To address the shortcomings of existing technologies, such as statically fixed dielectric loss parameters, inability to adapt to real indoor electromagnetic environments, low automation, and inability to retrieve scene-specific attenuation values ​​for individual structures, this invention provides a method and system for calculating the dielectric propagation loss of indoor structures based on 3D point cloud contour extraction and wireless signal samples. Relying on synchronous on-site measurement data at multiple heights, it automatically extracts multi-band lumped equivalent loss parameters for a single structure, accurately characterizing the real electromagnetic attenuation characteristics of the scene.

[0018] A method for calculating the propagation loss of indoor structures based on three-dimensional point cloud contour extraction and wireless signal samples includes the following steps.

[0019] Step 1: 3D Raster Map Construction. A mobile intelligent vehicle equipped with a depth camera and 2D LiDAR uses an indoor SLAM mapping solution to collect complete original 3D point cloud data. This embodiment preferably utilizes the open-source tool RTAB-Map within the ROS2 system ecosystem for point cloud mapping. Any 3D SLAM mapping tool, such as Cartographer or LOAM, can be used instead, as long as it can output complete and lossless original 3D point clouds. A cubic 3D raster map is constructed based on the spatial boundaries of the original point cloud. The preferred side length of the raster cube is 20cm, which can be adjusted as needed based on the deployment scene. The raster map is stored independently, and no modifications or deletions are made to the original point cloud throughout the process. The original point cloud resolution is fully preserved for subsequent structure outline extraction. The raster map is only used for spatial registration and sample association processing of wireless signal data.

[0020] Step Two: Structure Contour Extraction. The original 3D point cloud data acquired in Step One is directly used to extract the 3D contours of indoor structures through Euclidean clustering and boundary detection algorithms. Structures include walls, partitions, columns, furniture, door frames, etc. The contour output information includes the structure's spatial location, geometric dimensions, and surface normal vectors. Using the original point cloud for contour extraction avoids the accuracy loss caused by rasterization downsampling and accurately identifies the contours of small furniture, door frames, and other detailed structures.

[0021] Step 3: Spatial registration and sample association: 1. Within the near-ground space region of the vehicle coordinate system base_link, multiple sets of wireless signal measurement data are simultaneously collected at different heights. 20cm, 100cm, and 180cm are the preferred offsets for the Z-axis in the vehicle base_link coordinate system. The base_link reference is approximately 10cm above the ground. The collection height can be flexibly increased, decreased, or adjusted according to indoor ceiling height and equipment deployment scenarios. These three heights cover the low, middle, and high positions where conventional communication terminals are used by people indoors, allowing for the collection of samples under different signal incident angles. The collection height can be flexibly increased, decreased, or adjusted according to indoor ceiling height and equipment deployment scenarios. All collection points are limited to the near-ground space region. Simultaneous collection at multiple heights can obtain signal samples at different incident angles, and statistical analysis can reduce the measurement bias caused by single-layer collection and improve the stability of loss parameters. 2. The wireless signal sample pose coordinates are converted into 3D grid map grid coordinates, establishing a one-to-one mapping relationship between each measurement sample and the spatial grid; the 20cm grid granularity is consistent with the centimeter-level pose estimation accuracy of the vehicle and the screening threshold in step four, ensuring consistent spatial indexing logic; this invention does not require full coverage signal acquisition grid by grid, the vehicle collects data at low speed along the edge of the structure, allowing sufficient samples to accumulate at different orientations and heights of the structure, and suppressing measurement noise through statistical optimization. In this embodiment, a uniform speed of 0.2m / s is preferred, which matches the preferred 20cm grid size, ensuring that the vehicle can stably output multiple sets of point clouds and wireless signal samples within a single grid interval; when the grid size is adjusted to smaller voxels such as 5cm or 10cm, the driving speed can be reduced accordingly to ensure the data density of spatial registration; if a larger grid size is selected, the driving speed can be appropriately increased to improve acquisition efficiency; wireless signal measurement indicators include RSRP, RSSI, RSRQ, SINR, and frequency.

[0022] Step 4: Nearest Neighbor Sample Screening. For each structure whose contour has been extracted, a preset distance threshold (preferably 20cm, consistent with the preset size of the 3D raster map) is set based on the structure's surface. All wireless signal sample points within the threshold range are screened to generate a unique associated sample set for that structure.

[0023] Step 5, Dielectric Loss Calculation. The single-band dielectric loss is calculated using the associated sample set of the structure, the indoor space reference point, and the free-space path loss model. The indoor reference point serves as the calculation benchmark for the theoretical free-space received power, typically selected as the geometric center of the near-ground space region. In ultra-large spaces such as grand theaters and stadiums, if the geometric center exceeds the coverage area of ​​the vehicle's measurement trajectory, the reference point can be shifted to the nearest accessible location on the trajectory, or the geometric centroid of the measurement trajectory can be selected to ensure that the benchmark point is supported by measured data, avoiding a disconnect between theoretical calculations and actual spatial measurements.

[0024] The specific calculation logic is as follows: the difference between the measured received power of the sample and the theoretical received power of the reference point is taken as the total relative path loss. After deducting the free space loss caused by distance, the remaining value is the estimated value of the medium penetration loss of the sample point. For the loss estimates corresponding to all samples of the same structure, any one of the median, mean, weighted average, and M-estimation robust statistics is selected to complete the statistical optimization and obtain the medium loss characteristic value of the corresponding frequency band of the structure.

[0025] The dielectric loss characteristic value is the lumped equivalent attenuation parameter, which is not the same as the material's inherent physical constants such as dielectric constant and conductivity. It represents the sum of all additional attenuation of wireless signals other than free space geometric distance attenuation, covering the penetration absorption loss of structures, multipath reflection and diffraction cancellation loss, spatial standing wave superposition fluctuations, and signal incident angle coupling effect. It is only used as a comprehensive characterization of the current indoor electromagnetic environment and can only be used for the calibration of propagation models in the corresponding scenarios. It cannot be reused across scenarios as a general dielectric physical constant.

[0026] Step Six: Output Multi-Band Loss Parameters. Repeat the entire calculation process in Step Five by switching to different target frequency bands, and finally output the dielectric loss parameter matrix for all frequency bands corresponding to each structure; the target frequency bands include cellular mobile communication bands, Wi-Fi bands, and broadcast signal bands.

[0027] The present invention relates to a computing system technical solution.

[0028] An indoor media propagation loss calculation system based on 3D point cloud contour extraction and wireless signal samples, used to implement the above calculation method, includes: 1. 3D point cloud raw data acquisition module: mounted on a mobile vehicle, it acquires raw 3D point cloud data of the indoor environment; 2. 3D Raster Map Construction Module: Constructs a cubic 3D raster map based on the original point cloud spatial range, with a preferred grid side length of 20cm; 3. Structure outline extraction module: Based on the original 3D point cloud, the 3D outline of the structure is extracted through point cloud clustering and boundary detection algorithms; 4. Wireless signal acquisition module: Synchronously acquires wireless signal measurement data at multiple heights within the near-ground space of the vehicle body Z-axis from 0 to 2.2m. 20cm, 100cm, and 180cm are the preferred offsets of the Z-axis in the vehicle body base_link coordinate system. The base_link reference is about 10cm above the ground. 5. Spatial Registration Module: This module performs coordinate registration between wireless signal measurement samples and a 3D grid map, establishing a sample-grid mapping relationship. 6. Neighbor Sample Screening Module: Based on the surface of the structure, it filters the set of associated wireless signal samples according to a distance threshold of 20cm; 7. Dielectric Loss Calculation Module: Based on the associated sample set, spatial reference point, and free space path loss model, calculates the total equivalent dielectric loss characteristic value of the structure set in different frequency bands; 8. Output module: Integrates multi-band calculation results and outputs the dielectric loss parameter matrix of each structure.

[0029] Based on the above method and system, the present invention has the following beneficial effects: 1. High degree of automation: No manual labeling of structure location and material type is required throughout the entire process. The structure outline is extracted by automatic clustering and boundary detection based on the original point cloud, and the entire process is automated, which greatly improves work efficiency. 2. Strong scene adaptability: The output dielectric loss is the lumped equivalent attenuation parameter under the current indoor electromagnetic environment. The same material structure can output different loss values ​​under different layouts and multipath environment scenarios, which truly restores the on-site propagation attenuation characteristics, which is different from the traditional fixed static material parameters. 3. Wide range of structural compatibility: Supports the extraction and loss calculation of the outline of any geometric structure such as walls, columns, curved walls, irregular furniture, and door frames, without being limited by the preset standard material library; 4. Multi-band parallel adaptation: A single on-site data acquisition can complete the synchronous calculation of loss for multiple standards and multiple frequency bands, output a standardized parameter matrix, and support the unified calibration of multi-network integrated indoor propagation models; 5. Data is traceable and reproducible: The entire process of data acquisition, registration, screening, and calculation is standardized, the original point cloud and signal measurement samples are completely preserved, the calculation steps are reproducible, and the loss results have engineering verification basis. Attached Figure Description

[0030] This invention includes three accompanying drawings, which, together with the embodiments, fully illustrate the system architecture, hardware layout, and core computing process: Figure 1 is a schematic diagram of the overall system architecture, showing the connection relationship and complete data flow between the three-dimensional point cloud raw data acquisition module, the three-dimensional raster map construction module, the structure outline extraction module, the wireless signal acquisition module, the spatial registration module, the nearest neighbor sample screening module, the dielectric loss calculation module and the output module; Figure 2 is a schematic diagram of the three-dimensional layout of multiple signal acquisition devices in the vehicle coordinate system (base_link). Taking three wireless signal acquisition devices as an example, they are fixed at three preferred positions offset from the Z-axis of the vehicle base_link coordinate system by 20cm, 100cm, and 180cm respectively. The base_link reference projection is about 10cm above the ground. This is only an example of a multi-height deployment scheme. The height of the acquisition points can be changed as needed. Figure 3 is a schematic diagram of the dielectric loss calculation process, showing the complete two-layer loop calculation logic from input of a single structure-specific associated sample set to output of a multi-band dielectric loss parameter matrix; the attached figure includes the inner single structure sample traversal loop and the outer multi-band iterative loop. The process calculates independently for each single structure in batches, and does not use a globally unified mixed sample pool to complete the loss solution. Detailed Implementation

[0031] The technical solution of the present invention will be fully described below with reference to the accompanying drawings and three sets of typical embodiments. Unless otherwise specified, the transmission power and antenna gain of the wireless signal transmitter are read through the base station SIB system messages, or a standard signal source can be used instead. The terminal measurement values ​​already include the fixed losses of the antenna and feeder. The fixed deviations are automatically canceled when calculating the difference, which does not affect the final loss accuracy. All measurement and calculation operations are limited to the near-ground space area of ​​0 to 2.2m.

[0032] Example 1: Standard Office Scenario (Regular Structure) 1. Test scenario: A standard indoor office, including regular structures such as plasterboard partitions, concrete load-bearing columns, glass partitions, desks and chairs; 2. System Hardware Configuration: The mobile intelligent vehicle is equipped with an Astra RGBD depth camera, a Leishen N10P 2D LiDAR, and deploys a ROS 2 system; three commercial 5G smartphones serve as signal acquisition terminals, fixed at offset positions of 20cm, 100cm, and 180cm on the Z-axis of the vehicle's base_link coordinate system, respectively; the base_link reference is approximately 10cm above the ground. These three offsets represent the preferred deployment scheme in this embodiment, and the acquisition height and number of terminals can be adjusted as needed in actual engineering projects. 3. Data Acquisition Process: The trolley moves at a low and constant speed of 0.2 m / s along all walls and structure edges. This speed matches the 20 cm grid size in this embodiment. If other grid sizes are used, the speed can be adjusted synchronously to ensure sampling density. The depth camera and LiDAR continuously acquire the original 3D point cloud at a frequency of 10 Hz. Three mobile phones simultaneously acquire 4G, 5G, and Wi-Fi signal data at 1 second per acquisition, including RSRP, RSSI, RSRQ, SINR, and frequency points. 4. Data processing steps: 4.1 Raster Map Construction: Based on the complete original point cloud, a 20cm×20cm×20cm three-dimensional raster spatial index is generated. The original point cloud is completely preserved without any downsampling processing. 4.2 Contour Extraction: Euclidean clustering is performed on the original point cloud to distinguish point cloud clusters of different structures. Boundary detection algorithm is used to extract the 3D contours of walls, concrete columns, and desks, and output the spatial coordinates and surface normal vectors of each structure. 4.3 Spatial Registration: The coordinates of signal samples synchronously collected from three mobile phones at multiple altitudes are uniformly converted into raster map coordinates to establish a sample-raster mapping; 4.4 Nearest Neighbor Sample Screening: Using the surface of the concrete load-bearing column as a reference, all surrounding signal samples are screened with a distance threshold of 20cm to generate a unique associated sample set for the column; the trolley collects samples around the column, accumulating sufficient effective samples around the column and at a height of three floors. 4.5 Single-band loss calculation: Read the measured RSRP power value of the sample, select the geometric center of the office space as the spatial reference point, and calculate the theoretical received power by combining the free space path loss model; Formula for calculating the logarithmic field: In the formula: This is an estimate of the penetration loss of the sample medium, in dB. The measured received power of the sample is expressed in dBm. The theoretical free-space received power at the reference point, in dBm. Free space loss due to the distance between the sample and the reference point, in dB. By taking the median of all sample loss estimates and performing statistical optimization, the lumped equivalent dielectric loss characteristic value of concrete columns under a certain frequency band of a certain system is obtained. 4.6 Multi-band cyclic calculation: The repetitive loss calculation process of each frequency band under different standards is switched sequentially, and the multi-band loss parameter matrix of the concrete column is output. This embodiment verifies the automatic extraction and loss calculation capability of the present invention for conventional rectangular and block-shaped regular structures, and fully reproduces the complete set of acquisition-registration-screening-multi-band calculation process.

[0033] Example 2: Indoor corridor scene (irregularly shaped arc-shaped structure) 1. Test Scenario: A long corridor containing irregular structures such as curved walls and multiple doorways, used to verify the ability to adapt to irregular contours; 2. The acquisition, registration, and screening process is completely consistent with that of Example 1; the difference lies in the contour extraction step: this example does not require rectangular rule fitting of the curved wall surface. The original point cloud clustering + boundary detection algorithm directly outputs the complete contour point set of the curved surface. The 20cm distance threshold is used to screen the surrounding wireless signal samples along the normal of the curved surface, which is not limited by the geometry of the structure. After multi-band cyclic calculation, the multi-band dielectric loss parameter matrix corresponding to the curved wall surface and doorway partition is output, proving that the present invention can be adapted to any curved surface and irregular structure.

[0034] Example 3: Multi-standard synchronous measurement scenario 1. Test scenario: Indoor shopping mall, with multiple wireless signal sources including 4G macro base stations, 5G micro base stations, and 2.4G and 5G Wi-Fi deployed simultaneously; 2. Several independent acquisition terminals can be deployed at multiple height levels from 0 to 2.2m on the vehicle body. Different terminals lock onto different wireless standards and acquire data synchronously. Multiple standard signal samples can be acquired in parallel during a single test run. There is no need to enter the site multiple times for acquisition. All standard loss calculations are completed by looping through the back-end frequency band, and a complete loss parameter matrix is ​​output at once, which greatly improves the efficiency of scenario testing.

[0035] Supplementary Explanation The structural medium loss values ​​output by the three sets of embodiments are all scene-specific lumped equivalent attenuation parameters; the loss values ​​of the same concrete wall are significantly different in open glass partition offices and dense brick wall offices. This difference is an objective result of the coupling of the real electromagnetic environment and not a measurement error; the output parameters of this invention are only used for the calibration of the corresponding indoor scene propagation model and cannot be directly used as general physical constants of materials for cross-scene reuse.

Claims

1. A method for calculating the dielectric loss of indoor structures based on three-dimensional point cloud contours and wireless signal samples, characterized in that, Includes the following steps: Step 1: Acquire raw 3D point cloud: A mobile vehicle equipped with a depth camera and a 2D LiDAR acquires raw 3D point cloud data indoors. Step 2, construct a 3D raster map: construct a cubic 3D raster map based on the original point cloud, the raster is stored independently and the original point cloud is not downsampled or deleted; Step 3, extract the structure outline: Based on the original 3D point cloud, extract the 3D outline of each structure through point cloud clustering and boundary detection algorithms, and output the spatial position, geometric dimensions and surface normal vector of the structure. Step 4, Spatial registration and sample association: Set up multiple height layers to synchronously collect wireless signals in the near-ground space area of ​​0 to 2.2m in the Z-axis direction of the vehicle base_link coordinate system, unify the signal sample coordinates and the grid coordinate system, and establish a one-to-one mapping relationship between samples and grids; Step 5, Nearest Neighbor Sample Screening: Set a preset distance threshold based on the surface of the structure, screen all signal samples within the threshold range, and generate a unique associated sample set for the structure. Step 6, Dielectric loss calculation: Based on the associated sample set, indoor space reference point and free space path loss model, calculate the estimated dielectric penetration loss of each sample, perform statistical optimization on all loss values ​​of the same structure, and obtain the lumped equivalent dielectric loss characteristic value under the current frequency band. Step 7, Multi-band output: Switch to different target frequency bands and repeat step 6 to output the dielectric loss parameter matrix for all frequency bands corresponding to each structure.

2. The method according to claim 1, characterized in that, The side length of the cubic grid can be flexibly set according to the needs of the scene, with 20cm being the preferred size for engineering projects.

3. The method according to claim 1, characterized in that, The multiple height layers include three height layers offset by 20cm, 100cm, and 180cm in the Z-axis direction of the vehicle body base_link coordinate system; the offset values ​​and the number of height layers can be adjusted according to the needs of the scenario.

4. The method according to claim 1, characterized in that, The origin of the vehicle body base_link coordinate system is located on the plane of the mobile vehicle chassis, with a reference height of 10cm above the ground. The absolute height of each signal acquisition layer above the ground is determined by summing the reference height above the ground and the Z-axis offset.

5. The method according to claim 1, characterized in that, The distance threshold is equal to the grid side length of the 3D grid map; when the grid side length is 20cm, the distance threshold is 20cm.

6. The method according to claim 1, characterized in that, Wireless signal measurement metrics include RSRP, RSSI, RSRQ, SINR, and frequency points. Frequency points are used to distinguish different frequency bands in order to generate a multi-band loss matrix.

7. The method according to claim 1, characterized in that, The moving car's speed is adapted to the grid size; when the grid side length is 20cm, the corresponding constant speed is 0.2m / s.

8. The method according to claim 1, characterized in that, The statistical optimization uses any one of the following robust statistics: median, mean, weighted average, or M-estimation.

9. The method according to claim 1, characterized in that, Spatial reference points are preferentially selected from near-Earth spatial geometric centers; when the geometric center is outside the acquisition trajectory, the nearest reachable point or the geometric centroid of the trajectory is selected.

10. The method according to claim 1, characterized in that, The original 3D point cloud was acquired using a SLAM scheme; the SLAM scheme can run under the ROS2 software framework and can use any type of 3D SLAM function package such as RTAB-Map or Cartographer to achieve 3D mapping.

11. The method according to claim 1, characterized in that, The target frequency bands include cellular mobile communication bands, Wi-Fi bands, and broadcast signal bands.

12. A system for calculating the dielectric loss of indoor structures based on three-dimensional point cloud contours and wireless signal samples, characterized in that, A method for implementing any one of claims 1 to 11, comprising: A 3D point cloud raw data acquisition module, mounted on a mobile vehicle, is used to acquire raw 3D point clouds indoors. The 3D raster map building module is used to build a cubic 3D raster map based on the original point cloud. The structure outline extraction module is used to extract the 3D outline of each structure based on the original 3D point cloud. The wireless signal acquisition module is used to simultaneously acquire wireless signals at multiple altitudes within the near-ground space area of ​​the vehicle body. The spatial registration module is used to unify the coordinates of signal samples with the grid coordinate system and establish the mapping relationship between samples and grids. The nearest neighbor sample filtering module is used to filter the set of associated signal samples based on the surface of the structure; The dielectric loss calculation module is used to calculate the total equivalent dielectric loss characteristic value of each structure set in different frequency bands based on the associated sample set, spatial reference point and free space path loss model. The output module is used to integrate the multi-band loss values ​​of all structures and output the dielectric loss parameter matrix.

13. The system according to claim 12, characterized in that, The grid side length of the three-dimensional raster map construction module is 20cm; the distance threshold of the nearest neighbor sample screening module is 20cm.

14. The system according to claim 12, characterized in that, The wireless signal acquisition module simultaneously acquires wireless signals at three height levels offset from the Z-axis in the vehicle's base_link coordinate system: 20cm, 100cm, and 180cm. The reference ground clearance of the mobile vehicle is 10cm.

Images