Surveying and mapping system and surveying and mapping method for outdoor geographic surveying and mapping

Abstract

The invention discloses a surveying and mapping system and method for outdoor geographic surveying and mapping, and particularly relates to the field of outdoor geographic surveying and mapping, and the system comprises a preparation module, a constraint module, a model optimization module, a space-time registration module, a verification module and an evaluation module. The preparation module is used for deploying an unmanned aerial vehicle LiDAR scanning unit, a ground mobile surveying and mapping module and a distributed base station in a target area, constructing a space-time synchronous sensor network, and establishing a dynamic three-dimensional reference framework with a timestamp based on a measurement area control point; the constraint module is used for constructing a multi-threshold constraint matrix according to the terrain roughness, the vegetation coverage and the sensor performance parameters; the model optimization module is used for establishing a two-dimensional optimization model integrating spatial precision and data timeliness and forming a dynamic surveying and mapping decision function; according to the method, by introducing the dynamic weight and G-N iterative optimization, the root mean square error of registration is reduced, the convergence times are reduced, and the problems that multi-sensor registration is prone to local optimum and low in efficiency are solved.

Description

Technical Field

[0001] The present invention relates to the field of geographic surveying and mapping technology, and more particularly, to a surveying and mapping system and a surveying and mapping method for outdoor geographic surveying and mapping. Background Art

[0002] In the field of outdoor geographic surveying and mapping, traditional static surveying methods face the dual challenges of insufficient accuracy and poor dynamic adaptability in complex environments. Existing technologies mostly rely on a single sensor (such as a total station or a single LiDAR) to implement single-point discrete measurements. In forest areas with high vegetation coverage or mountainous areas with large standard deviations of terrain undulation, the plane positioning error is high and the elevation error is large, making it difficult to meet high-precision requirements. For example, although traditional drone LiDAR scanning can obtain point cloud data, the ground point cloud acquisition rate is insufficient in areas blocked by vegetation, and there is a lack of a dynamic parameter adjustment mechanism based on penetration rate, resulting in a high rate of missed detection of ground objects.

[0003] While multi-sensor collaborative mapping technology has been applied, it faces three major technical bottlenecks: First, when the data time synchronization error is extremely large (such as the time difference between LiDAR and imagery), the root mean square error of the registration increases sharply. Existing linear interpolation methods do not combine the dynamic characteristics of the sensor for error control, making it difficult to meet the ISO17123-10 requirements for temporal consistency of mobile measurement systems. Second, the traditional ICP registration algorithm does not introduce dynamic weight constraints for terrain roughness and vegetation penetration, and is prone to falling into local optimality in complex terrain areas. The average number of registration convergence times is high, and more efficient algorithms are time-consuming. Third, spatiotemporal consistency processing relies on static thresholds and does not consider the nonlinear characteristics of terrain deformation (such as the secondary acceleration change of landslides). The error in later dynamic mapping is large, exceeding the timeliness requirements of GB50328-2014 for deformation monitoring. These problems make it difficult for traditional methods to achieve high-precision modeling with a unified spatiotemporal benchmark in dynamic scenarios such as emergency mapping and geological disaster monitoring.

[0004] Therefore, there is an urgent need for a surveying and mapping system and method for outdoor geographic surveying. By constructing a heterogeneous sensor network and introducing a dynamic constraint matrix and adaptive fusion algorithm based on measured data, the core problems of the existing technology, namely "poor environmental adaptability, weak temporal and spatial consistency, and extensive parameter design", are solved. Summary of the Invention

[0005] In order to overcome the above-mentioned defects of the prior art, the present invention provides a surveying and mapping system and a surveying and mapping method for outdoor geographic surveying, which solve the problems raised in the above-mentioned background technology through the following scheme.

[0006] To achieve the above-mentioned purpose, the present invention provides the following technical solution: a surveying and mapping system for outdoor geographic surveying and mapping, comprising:

[0007] Preparation module: Deploy UAV LiDAR scanning units, ground mobile mapping modules and distributed reference stations in the target area to build a temporally and spatially synchronized sensor network and establish a dynamic 3D reference frame with timestamps based on the control points in the survey area;

[0008] Constraint module: Constructs a multi-threshold constraint matrix based on terrain roughness, vegetation coverage and sensor performance parameters;

[0009] Model optimization module: establishes a two-dimensional optimization model integrating spatial accuracy and data timeliness to form a dynamic surveying and mapping decision function;

[0010] Spatiotemporal registration module: Designs a multi-source data fusion process with adaptive weights to achieve spatiotemporal registration of point cloud, imagery, and inertial navigation data;

[0011] Verification module: performs multi-dimensional consistency verification on the generated 3D terrain model and optimizes model details based on iterative convergence conditions;

[0012] Evaluation module: Complete the surveying and mapping results evaluation by calculating the spatiotemporal deviation of feature points and comparing it with the dynamic threshold.

[0013] Preferably, the UAV LiDAR system uses a 1550nm laser wavelength, a scanning frequency adjustable from 200 to 500Hz, and a point cloud density of 50 points / n in a flat area. 2 Mountain 200 points / m 2 ; The ground mobile module integrates a 2-megapixel optical camera and a fiber optic IMU; the distributed base station adopts the BeiDou-3 differential module, with a plane positioning accuracy of ±1cm and an elevation accuracy of ±2cm. The distance between adjacent base stations is 1-2km in plain areas and 500-800m in mountainous areas; the dynamic three-dimensional reference frame is constructed as follows: the control point in the upper left corner of the survey area is used as the origin O(0, 0, 0), the east direction is the X-axis, the north direction is the Y-axis, and the sky direction is the Z-axis. Each surveying and mapping node is attached with a timestamp t, and the coordinates are expressed as P(x, y, z, t). The clock deviation is synchronized in real time through the base station.

[0014] Preferably, the multi-threshold constraint matrix includes terrain roughness constraint, vegetation penetration constraint and data synchronization constraint; the terrain roughness constraint in represents the local terrain relief rate of the (x, y, z) coordinate point at time t, represents the average elevation of the (x, y) plane area at time t, δ(t) represents the time attenuation factor; the vegetation penetration constraint where N g Indicates the number of ground point clouds, N t Indicates the total number of point clouds, V c Indicates vegetation coverage; the data synchronization constraint Where Ts Denote the sensor time difference, Δt L-I Denote the time difference between the LiDAR point cloud and the optical image, Δt I-G Denote the time difference between the IMU inertial data and the positioning data.

[0015] Preferably, the two-dimensional optimization model includes a spatial accuracy function and a data timeliness function; the spatial accuracy function where σ x and σ y and σ z represent the coordinate component mean errors; the data timeliness function where t0 represents the initial scanning time and λ represents the timeliness attenuation coefficient.

[0016] Preferably, the multi-source data fusion process includes timestamp alignment, feature pyramid construction, cross-modal feature matching, dynamic weight calculation, and iterative registration optimization; the timestamp alignment: use the linear interpolation method to unify the LiDAR point cloud, image, and IMU data to a 1ms time reference, and the interpolation formula is: where P(t) represents the sensor coordinate value corresponding to the target timestamp t, P(t1), P(t2) represent the known coordinate values corresponding to the adjacent timestamps t1, t2 and t1 < t < t2, t represents the target time to be interpolated, and t1, t2 represent the front and back timestamps adjacent to t; the feature pyramid construction: build an 8-layer Gaussian pyramid for the image and a voxel pyramid for the point cloud; the cross-modal feature matching: use the SuperPoint operator to extract the key points of the image, with a response threshold of 0.01, retain the top 1000 key points through non-maximum suppression, combine the point cloud normal vector mutation points, and establish corresponding point pairs through the descriptor Hamming distance, i.e., a threshold of 32; the dynamic weight calculation ​​​​​​​​​​​​​​​​​​​​​​​​​Preferably, the multi-dimensional consistency check includes spatiotemporal consistency check, scale consistency check and semantic consistency check; the spatiotemporal consistency check: checks the coordinate deviation of the same-name points in different time phases, and triggers the time phase correction when the plane deviation is greater than 0.15m or the elevation deviation is greater than 0.2m. The correction model is: ΔP = a×(t-t0)+b×(t-t0) 2 , where a and b represent correction coefficients, and t0 represents the initial scanning time; the scale consistency check: in a 1:500 scale survey map, measure 20 known side lengths and calculate the scale factor And it is required that |κ|<=0.001, where represents the side length value measured by the model, Represents the standard side length value measured in the field; the semantic consistency check: the feature category is identified by the CNNResNet-50 model, and 10,000 annotated images of 10 categories of features are used for training. The training parameters are: learning rate 0.001, batch size 32, number of iterations 200 rounds, validation set accuracy ≥ 85%, checking the consistency of point cloud classification and image interpretation, and re-annotating when the conflict rate is >15%; the iteration termination condition is: in 5 consecutive iterations, the spatiotemporal deviation attenuation rate is >50% and the semantic conflict rate decreases by >10%.

[0018] Preferably, the spatiotemporal deviation of the feature points Where (x p ,y p ,z p ,t p ) represents the spatial and temporal coordinates of the measured feature points, (x m ,y m ,z m ,t m ) represents the coordinates of the corresponding points of the model, D scale Indicates the denominator of the mapping scale; the dynamic threshold STD thr (t) = STD0 × (1 + 0.005 (t-t0)), where STD0 represents the initial threshold; the surveying and mapping results evaluation method is: when STD thr When (t)>STD, a normal signal is issued, which indicates that the surveying and mapping results are good; when STD thr When (t)<=STD, an early warning signal is issued, which means that the surveying and mapping results are not ideal and adjustments are required by relevant technical personnel.

[0019] Preferably, a surveying and mapping method for outdoor geographic surveying and mapping comprises:

[0020] S1. Deploy UAV LiDAR scanning units, ground mobile mapping modules, and distributed reference stations in the target area to build a spatiotemporally synchronized sensor network and establish a dynamic 3D reference frame with timestamps based on the control points in the survey area.

[0021] S2, construct a multi-threshold constraint matrix based on terrain roughness, vegetation coverage and sensor performance parameters;

[0022] S3. Establish a two-dimensional optimization model that integrates spatial accuracy and data timeliness to form a dynamic surveying and mapping decision function;

[0023] S4. Design a multi-source data fusion process with adaptive weights to achieve spatiotemporal registration of point cloud, imagery, and inertial navigation data.

[0024] S5. Perform multi-dimensional consistency check on the generated three-dimensional terrain model and optimize the model details according to the iterative convergence conditions;

[0025] S6. Complete the surveying and mapping results evaluation by calculating the spatiotemporal deviation of feature points and comparing it with the dynamic threshold.

[0026] Technical effects and advantages of the present invention:

[0027] 1. This invention deploys UAV LiDAR, ground mobile modules, and BeiDou reference stations, and combines terrain roughness constraints with vegetation penetration constraints to improve the ground point cloud acquisition rate in areas with high vegetation coverage and reduce the standard deviation of terrain undulation. This solves the problem of "insufficient accuracy of a single sensor in complex environments" in background technologies and enables detailed modeling of complex scenes such as forests and mountains.

[0028] 2. By introducing dynamic weights and GN iterative optimization, the present invention reduces the root mean square error of registration and the number of convergence times, thus solving the problem of "multi-sensor registration easily falling into local optimality and low efficiency" and ensuring the geometric consistency of the 3D model;

[0029] 3. The present invention reduces the late terrain deformation error and the temporal and spatial deviation evaluation error of feature points through the quadratic polynomial correction model and the time attenuation factor, thus solving the problem that "traditional static thresholds are difficult to adapt to dynamic changes in terrain". BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 Schematic diagram of the system structure of the present invention;

[0031] Figure 2 Schematic diagram of the method structure of the present invention. DETAILED DESCRIPTION

[0032] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0033] like Figure 1 A surveying and mapping system for outdoor geographic surveying and mapping is shown, comprising a preparation module, a constraint module, a model optimization module, a spatiotemporal registration module, a verification module, and an evaluation module.

[0034] The preparation module deploys UAV LiDAR scanning units, ground mobile mapping modules and distributed reference stations in the target area, builds a temporally and spatially synchronized sensor network, and establishes a dynamic three-dimensional reference frame with timestamps based on the control points of the survey area;

[0035] In this embodiment, it should be specifically noted that the UAV LiDAR system uses a 1550nm laser wavelength, a scanning frequency adjustable from 200 to 500Hz, and a point cloud density of 50 points / m in a flat area. 2 Mountain 200 points / m 2 ; The ground mobile module integrates a 2-megapixel optical camera and a fiber optic IMU; the distributed base station adopts the BeiDou-3 differential module, with a plane positioning accuracy of ±1cm and an elevation accuracy of ±2cm. The distance between adjacent base stations is 1-2km in plain areas and 500-800m in mountainous areas; the dynamic three-dimensional reference frame is constructed as follows: the control point in the upper left corner of the survey area is used as the origin O(0, 0, 0), the east direction is the X-axis, the north direction is the Y-axis, and the sky direction is the Z-axis. Each surveying and mapping node is attached with a timestamp t, and the coordinates are expressed as P(x, y, z, t). The clock deviation is synchronized in real time through the base station.

[0036] The constraint module constructs a multi-threshold constraint matrix based on terrain roughness, vegetation coverage, and sensor performance parameters. The threshold values ​​are based on GB / T50138-2011 "Engineering Survey Standard" and verified by 100 sets of typical terrain measured data.

[0037] In this embodiment, it should be specifically explained that: the multi-threshold constraint matrix includes terrain roughness constraint, vegetation penetration constraint and data synchronization constraint; the terrain roughness constraint in represents the local terrain relief rate of the (x, y, z) coordinate point at time t, It represents the average elevation of the (x, y) plane area at time t, δ(t) represents the time attenuation factor, δ = 1 within 24 hours, δ = 1.5 over 72 hours, based on 100 sets of slope deformation monitoring data statistics, the probability of terrain change after 72 hours is ≥ 15%; the numerator uses the extreme difference of the elevation gradient to quantify the severity of the local terrain fluctuation and avoid the one-sidedness of a single gradient value; the denominator is normalized with the average elevation to eliminate the interference of absolute elevation on roughness (such as the comparison between plateau areas and plain areas); the time factor δ(t) introduces dynamic correction to make the constraint conditions adapt to the timeliness requirements and meet the time accuracy requirements of deformation monitoring in GB / T50138-2011; the vegetation penetration constraint where N g Indicates the number of ground point clouds, N t Indicates the total number of point clouds, V c Indicates vegetation coverage, the threshold of 60% is based on LiDAR penetration experiment (when V c >60%, the ground point cloud acquisition rate drops by 50%); the left-hand fraction represents the proportion of ground point cloud, which directly reflects the ability of LiDAR to penetrate vegetation; the right-hand exponential function simulates the nonlinear attenuation of penetration rate after the vegetation coverage exceeds 60% (experimental data show that V c >60%, the ground point cloud acquisition rate drops by more than 50%); the formula as a whole converts vegetation occlusion into a quantifiable constraint factor, providing a basis for sensor parameter adjustment (such as scanning frequency, flight altitude); the data synchronization constraint Where T s represents the sensor time difference. According to Shannon sampling theorem, the sensor attitude change within 50ms is ≤0.5°, and the time synchronization error can be ignored; Δt L-I Represents the time difference between the LiDAR point cloud and the optical image, Δt I-G The time difference between the IMU inertial data and the positioning data is represented. The time difference is normalized by the ratio of the time difference to the threshold, making the time synchronization errors of different sensors comparable. The summation operation comprehensively evaluates the temporal consistency of multi-source data to avoid registration errors caused by the delay of a single sensor. The 50ms threshold setting meets the time synchronization requirements of ISO17123-10 for mobile measurement systems.

[0038] The model optimization module: establishes a two-dimensional optimization model integrating spatial accuracy and data timeliness to form a dynamic surveying and mapping decision function;

[0039] In this embodiment, it should be specifically explained that: the two-dimensional optimization model includes a spatial accuracy function and a data timeliness function; the spatial accuracy function Among them, σ x , σ y , σ zIt represents the mean error of the coordinate component, verified by 200 sets of measured data, and the mean error is calculated using the Bessel formula; the data time function Where t0 represents the initial scanning time, and λ represents the time-dependent decay coefficient (based on geological hazard monitoring data, the mean square error of terrain change after 100 hours is 0.2 m, corresponding to λ = 0.001). The root sum of squares of the mean square error in the denominator conforms to the error propagation law and represents the comprehensive accuracy of the three-dimensional space. The product term on the right side introduces a time synchronization penalty factor (0.5×DSC). When DSC = 1, the accuracy is reduced by 50%, reflecting the significant impact of time asynchrony on spatial registration. The output value of the formula is positively correlated with the accuracy, facilitating the maximization of the objective function in the optimization model. The exponential decay model conforms to the time-dependent laws of terrain change (for example, the deformation rate of landslides and debris flows decreases with time). The value of λ is fitted by measured data to ensure that the Tim value is ≤ 0.93 after 72 hours, corresponding to a 7% increase in terrain change risk, which is consistent with the deformation monitoring period requirements of GB50328-2014. The function output is used for dynamic threshold correction to appropriately reduce the accuracy requirements of old data, meeting actual surveying and mapping needs.

[0040] The spatiotemporal registration module designs a multi-source data fusion process with adaptive weights to achieve spatiotemporal registration of point cloud, imagery, and inertial navigation data.

[0041] In this embodiment, it should be specifically explained that the multi-source data fusion process includes timestamp alignment, feature pyramid construction, cross-modal feature matching, dynamic weight calculation, and iterative registration optimization; the timestamp alignment uses linear interpolation to unify the LiDAR point cloud, image, and IMU data to a 1ms time base. The interpolation formula is: Where P(t) represents the sensor coordinate value corresponding to the target timestamp t, P(t1) and P(t2) represent the known coordinate values corresponding to adjacent timestamps t1 and t2, and t1 < t < t2. t represents the target time to be interpolated, and t1 and t2 represent the previous and subsequent timestamps adjacent to t. The interpolation formula satisfies the rationality of the linear assumption and the multi-sensor synchronization requirements. Rationality of the linear assumption: Within a short time interval (such as 10 ms for LiDAR), the movement of the sensor (such as the flight of an unmanned aerial vehicle or the driving of a ground vehicle) can be approximated as a uniform linear motion, and the attitude change (such as the heading angle of an IMU) can be approximated as a linear change. Therefore, linear interpolation can meet the millimeter-level accuracy requirements (actual measurements show that within 10 ms, the displacement error of the unmanned aerial vehicle ≤ 0.01 m, and the attitude angle error ≤ 0.05°). Multi-sensor synchronization requirements: The time bases of LiDAR (sampling at 10 ms), images (captured at 50 ms), and IMU (sampling at 100 Hz) are different, and they need to be unified to a 1-ms time grid through interpolation. This formula achieves a smooth transition of coordinate values through time ratio allocation, avoiding registration misalignment caused by time asynchronization. The construction of the feature pyramid: An 8-layer Gaussian pyramid is established for the image (the bottom layer resolution is 4000×3000 pixels, and the scale factor for each layer is 1.5, conforming to the Lindeberg scale space theory), and a voxel pyramid is established for the point cloud (the minimum voxel is 0.05 m 3 , based on the optimal matching experiment of point cloud density and voxel size); The cross-modal feature matching: The SuperPoint operator is used to extract the key points of the image, with a response threshold of 0.01. The top 1000 key points are retained through non-maximum suppression. Combining with the sudden change points of the point cloud normal vector (curvature threshold 0.02, calculated based on PCA principal component analysis), corresponding point pairs are established through the Hamming distance of the descriptor, that is, the threshold is 32; The calculation of the dynamic weight where w lidar , w image , w imu represent the fusion weights of LiDAR, image, and IMU data respectively. The VPC threshold of 0.3 corresponds to an effective penetration rate of 30% for LiDAR ground point clouds (actual measurement data shows that when VPC > 0.3, the accuracy of image feature matching > 85%). The TRC threshold of 0.4 corresponds to a terrain起伏 standard deviation of 0.4 m (when exceeding this value, the influence of the IMU attitude is significant); The Sigmoid function is used to achieve the non-linear switching of weights. When VPC > 0.3, w lidar jumps from < 0.5 to w lidar>0.9, which is consistent with the positive correlation between vegetation penetration and LiDAR reliability; the TRC threshold of 0.4 distinguishes between flat terrain (TRC<0.4) and complex terrain (TRC≥0.4). In complex terrain, the IMU weight is increased to strengthen attitude correction; the weight adaptation mechanism avoids the dominance of a single data source, increasing the image weight in dense vegetation areas (low VPC) and the IMU weight in steep mountainous areas (high TRC); the iterative registration optimization: the GN algorithm is used to minimize the objective function: where w i Indicates the fusion weight of the i-th pair of points with the same name, represented by w lidar 、w image 、w imu Dynamic allocation, represents the three-dimensional coordinates of the i-th point in the LiDAR point cloud, represents the coordinates and posture parameters of the i-th feature point corresponding to the image and IMU, T represents the transformation matrix, and the iteration termination condition is that the error change rate is less than 0.1% (verified by 20 iterations, the root mean square error of the registration under this condition is ≤ 0.05m); the significance of the GN algorithm minimizing the objective function is: weight adaptive optimization: through w i Vegetation penetration constraint (VPC) and terrain roughness constraint (TRC) are introduced to enable the algorithm to automatically reduce the LiDAR weight in dense vegetation areas and strengthen the IMU attitude correction in complex terrain areas to avoid the registration deviation caused by the dominance of a single data source; Applicability to nonlinear problems: The coordinate transformation of sensor data is essentially a nonlinear problem (such as the orthogonality constraint of the rotation matrix). The GN algorithm linearizes it through Taylor expansion, solves the optimal parameter increment in each iteration, and converges quickly (the root mean square error within 20 iterations is measured to be ≤0.05m); Difference from existing technologies: The traditional ICP algorithm does not introduce dynamic weights and multi-sensor constraints. This solution uses w i The joint optimization of and T() improves the registration accuracy by 30% (RMSE is reduced from 0.08m to 0.05m in plain areas and from 0.12m to 0.08m in mountainous areas).

[0042] The verification module performs multi-dimensional consistency verification on the generated three-dimensional terrain model and optimizes the model details according to the iterative convergence conditions;

[0043] In this embodiment, it should be specifically explained that the multi-dimensional consistency check includes spatiotemporal consistency check, scale consistency check, and semantic consistency check; the spatiotemporal consistency check checks the coordinate deviations of the same-name points in different time phases. When the plane deviation is greater than 0.15m or the elevation deviation is greater than 0.2m, a time phase correction is triggered. The correction model is: ΔP = a × (t-t0) + b × (t-t0) 2, where a and b represent correction coefficients, which are determined by fitting 20 sets of time phase data using the least squares method. The goodness of fit R 2 >0.95, t0 represents the initial scanning time; the scale consistency check: in the 1:500 scale map, measure 20 known side lengths and calculate the scale factor And it is required that |κ|<=0.001, according to the GB / T20257.1-2017 map illustration standard, where represents the side length value measured by the model, Represents the standard side length value measured in the field; the semantic consistency check: the feature category is identified by the CNNResNet-50 model, and 10,000 annotated images of 10 categories of features are used for training. The training parameters are: learning rate 0.001, batch size 32, number of iterations 200 rounds, verification set accuracy ≥ 85%, checking the consistency of point cloud classification and image interpretation, and re-annotating when the conflict rate is > 15%; the iteration termination condition is: in 5 consecutive iterations, the spatiotemporal deviation attenuation rate is > 50% and the semantic conflict rate decreases by > 10%. The model converges under this condition, verified by 30 sets of complex terrain data.

[0044] The evaluation module completes the surveying and mapping results evaluation by calculating the spatiotemporal deviation of feature points and comparing it with the dynamic threshold.

[0045] In this embodiment, it should be specifically explained that: the spatiotemporal deviation of the feature points Where (x p ,y p ,z p ,t p ) represents the spatial and temporal coordinates of the measured feature points, (x m ,y m ,z m ,t m ) represents the coordinates of the corresponding points of the model, D scale Indicates the denominator of the mapping scale; the square root of the numerator contains the three-dimensional spatial coordinate difference and the time difference, forming a four-dimensional deviation measurement, which meets the spatiotemporal consistency requirements of dynamic mapping; the denominator uses the scale denominator to standardize the spatial deviation (for example, in 1:500 mapping, 1m of field distance corresponds to 0.002m of the model), and (1+Tim) is used to weaken the deviation influence of old data; the output value of the formula is independent of the scale, which is convenient for unified evaluation of projects with different accuracy requirements (for example, 1:500 and 1:1000 mapping can share the same STD threshold standard) as the accuracy magnification factor. When the Acc value is high (for example, Acc>10m -1 , corresponding to the three-dimensional error ≤ 0.1m), the denominator increases and the STD value decreases. When the characterization model accuracy is higher, the allowed deviation threshold is lower; the dynamic threshold STD thr(t) = STD0 × (1 + 0.005 (t-t0)), where STD0 represents the initial threshold, STD0 = 0.08m in 1:500 mapping, based on the national basic scale topographic map accuracy requirements, the coefficient 0.005 / hour is determined by analyzing the error growth curve of 100 sets of mapping data at different time phases (the average error growth after 72 hours is 0.012m / hour); the surveying and mapping results evaluation method is: when STD thr When (t)>STD, a normal signal is issued, which indicates that the surveying and mapping results are good; when STD thr When (t)<=STD, an early warning signal is issued, which means that the surveying and mapping results are not ideal and adjustments are required by relevant technical personnel.

[0046] Based on the above scheme and Figure 2 The present invention also provides a surveying and mapping method for outdoor geographic surveying and mapping, comprising:

[0047] S1. Deploy UAV LiDAR scanning units, ground mobile mapping modules, and distributed reference stations in the target area to build a spatiotemporally synchronized sensor network and establish a dynamic 3D reference frame with timestamps based on the control points in the survey area.

[0048] S2, construct a multi-threshold constraint matrix based on terrain roughness, vegetation coverage and sensor performance parameters;

[0049] S3. Establish a two-dimensional optimization model that integrates spatial accuracy and data timeliness to form a dynamic surveying and mapping decision function;

[0050] S4. Design a multi-source data fusion process with adaptive weights to achieve spatiotemporal registration of point cloud, imagery, and inertial navigation data.

[0051] S5. Perform multi-dimensional consistency check on the generated three-dimensional terrain model and optimize the model details according to the iterative convergence conditions;

[0052] S6. Complete the surveying and mapping results evaluation by calculating the spatiotemporal deviation of feature points and comparing it with the dynamic threshold.

[0053] Secondly: The drawings of the embodiments disclosed in the present invention only involve structures related to the embodiments disclosed in the present invention. Other structures may refer to conventional designs. The same embodiment and different embodiments of the present invention may be combined with each other without conflict.

[0054] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A surveying and mapping system for outdoor geographic surveying, characterized in that: include: Preparation module: Deploy UAV LiDAR scanning units, ground mobile mapping modules and distributed reference stations in the target area to build a temporally and spatially synchronized sensor network and establish a dynamic 3D reference frame with timestamps based on the control points in the survey area; Constraint module: Constructs a multi-threshold constraint matrix based on terrain roughness, vegetation coverage and sensor performance parameters; Model optimization module: establishes a two-dimensional optimization model integrating spatial accuracy and data timeliness to form a dynamic surveying and mapping decision function; Spatiotemporal registration module: Designs a multi-source data fusion process with adaptive weights to achieve spatiotemporal registration of point cloud, imagery, and inertial navigation data; Verification module: performs multi-dimensional consistency verification on the generated 3D terrain model and optimizes model details based on iterative convergence conditions; Evaluation module: Complete the surveying and mapping results evaluation by calculating the spatiotemporal deviation of feature points and comparing it with the dynamic threshold.

2. A surveying and mapping system for outdoor geographic surveying and mapping according to claim 1, characterized in that: The sensor network deployment parameters are as follows: the UAV LiDAR system uses a 1550nm laser wavelength, a scanning frequency adjustable from 200 to 500Hz, and a point cloud density of 50 points / m in a flat area. 2 Mountain 200 points / m 2 ; The ground mobile module integrates a 2-megapixel optical camera and a fiber optic IMU; the distributed base station adopts the BeiDou-3 differential module, with a plane positioning accuracy of ±1cm and an elevation accuracy of ±2cm. The distance between adjacent base stations is 1-2km in plain areas and 500-800m in mountainous areas; the dynamic three-dimensional reference frame is constructed as follows: the control point in the upper left corner of the survey area is used as the origin O(0, 0, 0), the east direction is the X-axis, the north direction is the Y-axis, and the sky direction is the Z-axis. Each surveying and mapping node is attached with a timestamp t, and the coordinates are expressed as P(x, y, z, t). The clock deviation is synchronized in real time through the base station.

3. The outdoor geographic surveying and mapping system according to claim 1, characterized in that: The multi-threshold constraint matrix includes terrain roughness constraint, vegetation penetration constraint and data synchronization constraint; the terrain roughness constraint in represents the local terrain relief rate of the (x, y, z) coordinate point at time t, represents the average elevation of the (x, y) plane area at time t, δ(t) represents the time attenuation factor; the vegetation penetration constraint where N g Indicates the number of ground point clouds, N t Indicates the total number of point clouds, V c Indicates vegetation coverage; the data synchronization constraint Where T s Indicates the sensor time difference, Δt L-I Represents the time difference between the LiDAR point cloud and the optical image, Δt I-G Indicates the time difference between IMU inertial data and positioning data.

4. The outdoor geographic surveying and mapping system according to claim 1, characterized in that: The two-dimensional optimization model includes a spatial precision function and a data timeliness function; the spatial precision function Among them, σ x , σ y , σ z Represents the error in the coordinate component; the data time function Where t0 represents the initial scanning time and λ represents the time-dependent attenuation coefficient.

5. The outdoor geographic surveying and mapping system according to claim 1, characterized in that: The multi-source data fusion process includes timestamp alignment, feature pyramid construction, cross-modal feature matching, dynamic weight calculation, and iterative registration optimization; the timestamp alignment: using the linear interpolation method to unify the LiDAR point cloud, image, and IMU data to a 1ms time reference, and the interpolation formula is: where P(t) represents the sensor coordinate value corresponding to the target timestamp t, P(t1), P(t2) represent the known coordinate values corresponding to adjacent timestamps t1, t2 and t1 < t < t2, t represents the target time to be interpolated, and t1, t2 represent the front and back timestamps adjacent to t; the feature pyramid construction: building an 8-layer Gaussian pyramid for the image and a voxel pyramid for the point cloud; the cross-modal feature matching: using the SuperPoint operator to extract the key points of the image, with a response threshold of 0.01, retaining the top 1000 key points through non-maximum suppression, combining the point cloud normal vector mutation points, and establishing corresponding point pairs through the descriptor Hamming distance, i.e., a threshold of 32; the dynamic weight calculation w imu =1 - w lidar - w image where w lidar 、w image 、w imu respectively represent the fusion weights of the LiDAR, image, and IMU data; the iterative registration optimization: using the G-N algorithm to minimize the objective function: where w i represents the fusion weight of the i-th corresponding point pair, dynamically allocated by w lidar 、w image 、w imu , represents the three-dimensional coordinate of the i-th point in the LiDAR point cloud, represents the coordinate and pose parameters of the i-th feature point corresponding to the image and IMU.

6. The outdoor geographic surveying and mapping system according to claim 1, characterized in that: The multi-dimensional consistency check includes spatiotemporal consistency check, scale consistency check, and semantic consistency check. The spatiotemporal consistency check checks the coordinate deviation of the same-name points in different time phases. When the plane deviation is greater than 0.15m or the elevation deviation is greater than 0.2m, a time phase correction is triggered. The correction model is: ΔP = a × (t-t0) + b × (t-t0) 2 , where a and b represent correction coefficients, and t0 represents the initial scanning time; The scale consistency check: In the 1:500 scale survey map, measure 20 known side lengths and calculate the scale factor And it is required that |κ|<=0.001, where represents the side length value measured by the model, Represents the standard side length value measured in the field; the semantic consistency check: the feature category is identified by the CNNResNet-50 model, and 10,000 annotated images of 10 categories of features are used for training. The training parameters are: learning rate 0.001, batch size 32, number of iterations 200 rounds, validation set accuracy ≥ 85%, checking the consistency of point cloud classification and image interpretation, and re-annotating when the conflict rate is >15%; the iteration termination condition is: in 5 consecutive iterations, the spatiotemporal deviation attenuation rate is >50% and the semantic conflict rate decreases by >10%.

7. The outdoor geographic surveying and mapping system according to claim 1, characterized in that: The spatiotemporal deviation of the feature points Where (x p ,y p ,z p ,t p ) represents the spatial and temporal coordinates of the measured feature points, (x m ,y m ,z m ,t m ) represents the coordinates of the corresponding points of the model, D scale Indicates the denominator of the mapping scale; the dynamic threshold STD thr (t) = STD0 × (1 + 0.005 (t-t0)), where STD0 represents the initial threshold; the surveying and mapping results evaluation method is: when STD thr When (t)>STD, a normal signal is issued, which indicates that the surveying and mapping results are good; when STD thr When (t)<=STD, an early warning signal is issued, which means that the surveying and mapping results are not ideal and adjustments are required by relevant technical personnel.

8. A surveying and mapping method for outdoor geographic surveying and mapping according to claim 1, used to implement a surveying and mapping system for outdoor geographic surveying and mapping according to any one of claims 1 to 7, characterized in that: include: S1. Deploy UAV LiDAR scanning units, ground mobile mapping modules, and distributed reference stations in the target area to build a spatiotemporally synchronized sensor network and establish a dynamic 3D reference frame with timestamps based on the control points in the survey area. S2, construct a multi-threshold constraint matrix based on terrain roughness, vegetation coverage and sensor performance parameters; S3. Establish a two-dimensional optimization model that integrates spatial accuracy and data timeliness to form a dynamic surveying and mapping decision function; S4. Design a multi-source data fusion process with adaptive weights to achieve spatiotemporal registration of point cloud, imagery, and inertial navigation data. S5. Perform multi-dimensional consistency check on the generated three-dimensional terrain model and optimize the model details according to the iterative convergence conditions; S6. Complete the surveying and mapping results evaluation by calculating the spatiotemporal deviation of feature points and comparing it with the dynamic threshold.

Images