Photogrammetry integrated control method and system

Abstract

The invention provides a photogrammetry integrated control method and system, and relates to the technical field of photogrammetry control, and the method comprises the steps: obtaining multi-source original data, such as a sequence image, an initial point cloud and a real-time attitude parameter; carrying out combined preprocessing of timestamp alignment, point cloud denoising and image distortion correction, and uploading key logs and quality indexes through a consensus mechanism for evidence storage; performing bundle adjustment to solve internal and external orientation elements, and calling block chain historical control points for verification and dynamic adjustment; generating an initial three-dimensional point cloud through dense matching; verifying the point cloud precision by using a high-precision check point, and feeding back a correction instruction when the point cloud exceeds the limit; gridding and texture mapping are carried out on the effective point cloud to generate a three-dimensional model, and achievements are partitioned and stored in edge nodes and are chained for evidence storage; the system comprises a multi-source data acquisition module, a data preprocessing module, an adjustment resolving module, a point cloud generation module, a point cloud verification module and a three-dimensional modeling module. The efficiency, precision and reliability of photogrammetry are effectively improved.

Description

Technical Field

[0001] This invention relates to the field of photogrammetry control technology, and in particular to an integrated photogrammetry control method and system. Background Technology

[0002] Photogrammetry, which processes image data to obtain geometric information and texture of objects, has been widely used in urban planning, geological disaster monitoring, digital cultural heritage protection, and large-scale engineering construction. However, as application scenarios increasingly demand real-time performance, accuracy, and data reliability, existing technologies face severe challenges in practical implementation. In traditional workflows, data acquisition, preprocessing, aerial triangulation, and 3D reconstruction typically rely on centralized post-processing platforms. For example, in geological disaster emergency monitoring, massive amounts of imagery and point cloud data collected by drones must be transmitted back to a cloud computing center for processing. Due to network bandwidth limitations, data transmission delays can reach several hours, making it difficult to provide real-time support for disaster relief decisions. In the periodic deformation monitoring of large bridges and tunnels, quality control at each stage of data processing relies heavily on manual intervention and post-event checks, lacking a continuous automated accuracy verification and feedback mechanism. This can lead to error accumulation and make it difficult to meet the requirements for millimeter-level high-precision monitoring. Furthermore, the authenticity and integrity of the final results lack effective technical means for auditing and traceability, making measurement results susceptible to questioning in scenarios such as engineering acceptance or forensic appraisal. The core challenges that existing technologies struggle to address lie primarily in three areas: First, the timeliness of data processing is a bottleneck; the centralized transmission and processing of massive datasets cannot meet the low-latency requirements of scenarios such as emergency response and real-time monitoring. Second, quality control is disconnected; the accuracy control of adjustment calculations and 3D reconstruction is relatively independent, lacking a closed-loop verification and dynamic adjustment mechanism based on measured checkpoints, making it difficult to detect and correct system errors in a timely manner. Third, there is a lack of reliable data assurance throughout the entire process; from raw data and intermediate results to the final model, the processing procedures, quality indicators, and version history lack tamper-proof evidence preservation and traceability capabilities, making it difficult to meet compliance requirements. Therefore, there is an urgent need in the market for a novel integrated photogrammetric control method and system. Summary of the Invention

[0003] The purpose of this invention is to provide an integrated photogrammetric control method and system to solve the problems in the prior art, such as poor timeliness caused by centralized data processing in the cloud, difficulty in dynamically correcting accumulated errors caused by the disconnect in quality control of various processing links (such as adjustment calculation and 3D reconstruction), and lack of a reliable evidence storage mechanism to prevent tampering of data and results throughout the entire process. The specific technical solution is as follows: This invention provides an integrated photogrammetric control method, comprising: S011, multi-source raw data of the area to be measured is synchronously acquired through an integrated data acquisition terminal. The multi-source raw data includes sequential image data, initial point cloud data and real-time attitude parameter data. The multi-source raw data is jointly preprocessed to generate preprocessed data. The edge computing node uploads the key logs and quality evaluation indicators of the preprocessing process to the blockchain network for evidence storage through a consensus mechanism. S012, based on the preprocessed data, the initial interior and exterior orientation elements of the sequence image are solved by bundle adjustment at the edge computing node; the database of historical control points in the same area stored in the blockchain network is called for verification. If the coordinate deviation value exceeds the preset threshold, the adjustment weight coefficient is dynamically adjusted and recalculated. If the coordinate deviation value is less than or equal to the preset threshold, the interior and exterior orientation elements currently solved are determined to be the final interior and exterior orientation elements. The data summary of the solution results of the final interior and exterior orientation elements is uploaded to the blockchain network. S013, combining the final interior and exterior orientation elements with the preprocessed data, constructs an association mapping model of "image pixel - point cloud coordinates" at the edge computing node, performs dense matching, and generates the initial three-dimensional point cloud; S014: Using high-precision checkpoint data measured in the field, verify the coordinate error of the initial 3D point cloud; if the error exceeds the limit, generate an accuracy correction instruction and feed it back to the dense matching step to regenerate the point cloud; after verification, store the data summary and accuracy report of the effective 3D point cloud on the blockchain. S015. The effective 3D point cloud is meshed, and the preprocessed image data is mapped as a texture map to generate a 3D visualization model; the final result data package is split and stored in multiple edge nodes, and its index address and integrity verification value are stored in the blockchain network.

[0004] Furthermore, edge computing nodes deployed near the data acquisition terminal are used to perform joint preprocessing on the multi-source raw data, including timestamp alignment, point cloud denoising, and image distortion correction, to generate preprocessed data; the preprocessed data includes preprocessed image data and preprocessed point cloud data.

[0005] Furthermore, the joint preprocessing includes: Timestamp alignment is based on hardware synchronization signals, which unifies the timestamps of image data, point cloud data and attitude data to the Global Positioning System. Point cloud denoising employs a statistical filtering algorithm to calculate the local density of the point cloud and remove outliers. Image distortion correction employs the Brown-Conrad model, using camera calibration parameters for geometric correction.

[0006] Furthermore, in step S012, the bundle adjustment uses the collinearity condition equation to construct the error equation, with the image point observation value as the observation value and the exterior orientation element and object point coordinates as unknowns, and the Levenburg-Marquardt algorithm is used for iterative solution; the historical control point database contains at least three uniformly distributed control points, and the preset thresholds include the plane position deviation threshold and the elevation deviation threshold.

[0007] Furthermore, step S013 involves constructing a mapping model between image pixels and point cloud coordinates, including: Based on the final interior and exterior orientation elements and the preprocessed image data, key point detection and descriptor generation operations are performed using the scale-invariant feature transformation algorithm to obtain the image feature point set. Based on the image feature point set, a random sampling consensus algorithm is used to remove mismatches and obtain optimized matching point pairs. Based on the optimized matching point pairs and the final interior and exterior orientation elements, the three-dimensional coordinate calculation operation is performed through the forward intersection algorithm to obtain a sparse three-dimensional point cloud; Based on sparse 3D point cloud and preprocessed image data, a pixel-level disparity calculation operation is performed using a semi-global matching algorithm to generate a dense 3D point cloud as the initial 3D point cloud.

[0008] Furthermore, the verification of the coordinate error of the initial 3D point cloud in step S014 includes: Based on the initial 3D point cloud and the high-precision checkpoint data measured in the field, the point cloud and checkpoints are registered using the nearest neighbor search algorithm to obtain a set of registered point pairs. Based on the set of registration point pairs, coordinate system one operation is performed through a seven-parameter coordinate transformation model to obtain the transformed checkpoint coordinates; Based on the transformed checkpoint coordinates and the corresponding initial 3D point cloud coordinates, the coordinate residual sequence is calculated using Euclidean distance. Based on the coordinate residual sequence, error distribution analysis is performed through statistical hypothesis testing. When the error exceeds the limit, a precision correction command containing correction parameters is generated.

[0009] Furthermore, step S015 involves meshing and texture mapping of the effective 3D point cloud, including: Based on the effective 3D point cloud, the point cloud density field is calculated and the isosurface is extracted using the Poisson surface reconstruction algorithm to obtain the initial triangular mesh model. Based on the initial triangular mesh model, a simplified mesh model is obtained by optimizing the number of vertices through a mesh simplification algorithm based on edge folding. Based on the simplified mesh model, the mesh surface is smoothed using the Laplacian smoothing algorithm to obtain the optimized mesh model; Based on the optimized mesh model and preprocessed image data, image registration and color fusion operations are performed using a multi-view texture mapping algorithm to generate a textured 3D model. The distributed storage of the final result data package in step S015 includes: Based on the textured 3D model, the model data is segmented using a spatial block algorithm to generate multiple model data blocks. Multiple model data blocks are distributed and stored on edge nodes, and the integrity check value of each data block is calculated. The index address and integrity verification value of the model data block are stored on the blockchain network through a smart contract.

[0010] The present invention also provides an integrated photogrammetric control system for implementing the aforementioned integrated photogrammetric control method, the system comprising: The multi-source data acquisition module is used to synchronously acquire multi-source raw data of the area to be measured through an integrated data acquisition terminal; The data preprocessing module is used to perform joint preprocessing on multi-source raw data using edge computing nodes to generate preprocessed data; The adjustment module is used to perform initial interior and exterior orientation elements of the image sequence using bundle adjustment based on the preprocessed data, and to verify and adjust them. The point cloud generation module is used to combine the final interior and exterior orientation elements with the preprocessed data to build an association mapping model and generate an initial 3D point cloud. The point cloud verification module is used to verify the coordinate errors of the initial 3D point cloud using high-precision checkpoint data and generate accuracy correction instructions. The 3D modeling module is used to perform meshing and texture mapping on effective 3D point clouds to generate 3D visualization models.

[0011] The present invention also provides an electronic device, including a processor, a memory, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the integrated photogrammetric control method described above.

[0012] The present invention also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the integrated photogrammetric control method.

[0013] The beneficial effects of this invention are as follows: The integrated photogrammetric control method and system provided by this invention achieve collaborative processing of data acquisition, preprocessing and adjustment calculation through edge computing nodes, and combine blockchain technology to perform distributed storage of key data and accuracy indicators, thus constructing a closed-loop control system for the entire process from data acquisition to 3D modeling. This significantly improves the automation level and reliability of photogrammetric data processing, effectively solves the problems of process fragmentation, lagging accuracy control and low data credibility in traditional methods, and achieves efficient 3D reconstruction with millimeter-level accuracy and traceable quality management.

[0014] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0015] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings: Figure 1 This is a schematic diagram of the steps of the integrated photogrammetry control method of the present invention; Figure 2 This is a schematic diagram of the structure of an integrated photogrammetry control system according to the present invention. Detailed Implementation

[0016] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0017] It should be noted that all directional indications (such as up, down, left, right, front, back, etc.) in the embodiments of the present invention are only used to explain the relative positional relationship and movement of each component in a certain specific posture (as shown in the figure). If the specific posture changes, the directional indication will also change accordingly.

[0018] It should also be noted that when a component is described as "fixed to" or "set on" another component, it can be directly on the other component or there may be an intervening component present. When a component is described as "connected to" another component, it can be directly connected to the other component or there may be an intervening component present.

[0019] Furthermore, the use of terms such as "first" and "second" in this invention is for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined with "first" or "second" may explicitly or implicitly include at least one of that feature. Additionally, the technical solutions of the various embodiments can be combined with each other, but only on the basis of being achievable by those skilled in the art. When the combination of technical solutions is contradictory or impossible to implement, such a combination of technical solutions should be considered non-existent and not within the scope of protection claimed by this invention.

[0020] In an embodiment of the present invention, an integrated photogrammetric control method is provided; please refer to [link to relevant documentation]. Figure 1 The method includes the following steps: S011, multi-source raw data of the area to be measured is synchronously acquired through an integrated data acquisition terminal. The multi-source raw data includes sequential image data, initial point cloud data, and real-time attitude parameter data. Edge computing nodes deployed near the data acquisition terminal perform joint preprocessing on the multi-source raw data, including timestamp alignment, point cloud denoising, and image distortion correction, to generate preprocessed data. The preprocessed data includes preprocessed image data and preprocessed point cloud data. The edge computing nodes upload key logs and quality evaluation indicators of the preprocessing process to a blockchain network for evidence storage through a consensus mechanism.

[0021] Specifically, the integrated data acquisition terminal includes a high-resolution digital aerial camera (e.g., ≥24 million pixels), a laser scanner (e.g., scanning frequency ≥100kHz), and an inertial measurement unit (IMU) / global positioning system (GPS) integrated navigation system (e.g., positioning accuracy ≤2cm). During data acquisition, the terminal flies along the planned route at a preset altitude (e.g., 100 meters) and heading overlap (e.g., 80%), simultaneously triggering the camera to acquire sequential images (e.g., 2 frames per second), the laser scanner to acquire initial point clouds, and the IMU / GPS to record real-time attitude parameters (including 3D position, roll angle, pitch angle, and yaw angle). Edge computing nodes employ high-performance embedded processors (such as the ARM Cortex-A77 architecture) and are deployed near the data acquisition terminal (e.g., connected via a 5G low-latency link). They are used to complete time-sensitive tasks such as data preprocessing, real-time adjustment calculations, and lightweight point cloud generation. For computationally intensive, detailed 3D reconstruction and modeling tasks, the data can be completed at the edge node or back to the cloud computing center, depending on network conditions. Joint preprocessing begins with timestamp alignment: based on hardware synchronization signals (such as PPS pulses), the timestamps of image data, point cloud data, and attitude data are unified to the Global Positioning Time System (such as GPST), with deviations controlled within milliseconds (e.g., <10ms). Point cloud denoising employs statistical filtering algorithms: calculating local point cloud density (e.g., a spherical neighborhood with a radius of 0.5 meters), and removing outliers (e.g., points with density less than twice the standard deviation of the average density). Image distortion correction uses a Brown-Conrad model, employing camera calibration parameters (such as focal length, principal point, radial distortion coefficients k1, k2, and tangential distortion coefficients p1, p2) for geometric correction to eliminate lens distortion (e.g., distortion residual ≤0.1 pixels). Key logs include preprocessing timestamps, data size, and processing status; quality evaluation metrics include image signal-to-noise ratio (SNR ≥ 30dB), point cloud integrity (missing rate ≤ 5%), and geometric accuracy (root mean square error of reprojection ≤ 0.1 pixels). The consensus mechanism employs the Practical Byzantine Fault Tolerance (PBFT) algorithm, which uploads the hash values ​​of logs and metrics (such as SHA-256 digests) to a blockchain network (such as a private chain based on Ethereum) to ensure that the data is immutable.

[0022] S012, based on the preprocessed data, the initial interior and exterior orientation elements of the image sequence are solved by bundle adjustment at the edge computing node; the database of historical control points in the same area stored in the blockchain network is called for verification. If the coordinate deviation value exceeds the preset threshold, the adjustment weight coefficient is dynamically adjusted and recalculated. If the coordinate deviation value is less than or equal to the preset threshold, the currently solved interior and exterior orientation elements are determined to be the final interior and exterior orientation elements, and the data digest (such as SHA-256 hash value) of the solution result of the final interior and exterior orientation elements is uploaded to the blockchain network.

[0023] Specifically, bundle adjustment uses collinearity condition equations to construct error equations, with image point observations (such as feature point coordinates) as observations and exterior orientation elements (six parameters for each image: three-dimensional position and three attitude angles) and object point coordinates as unknowns. The Levenburg-Marquardt algorithm is used for iterative solution (e.g., a maximum of 100 iterations and a convergence threshold ≤ 1e-6). Initial values ​​are provided by IMU / GPS data (e.g., position error ≤ 5cm). The historical control point database is retrieved from a blockchain network and contains at least three uniformly distributed control points (e.g., high-precision coordinates obtained through previous measurements, with a planar accuracy ≤ 5cm). During verification, the calculated object point coordinates are compared with the historical control point coordinates to calculate the deviation (e.g., Euclidean distance is used for planar position deviation, and absolute difference is used for elevation deviation). A preset threshold is set according to measurement accuracy requirements (e.g., planar position deviation threshold ≤ 10cm). If the deviation exceeds the limit, the adjustment weighting coefficient is dynamically adjusted: based on the magnitude of the deviation, the weight of the control points is increased (e.g., the weighting coefficient is adjusted from 1.0 to 2.0) or additional parameters (e.g., lens distortion correction terms) are introduced, and the calculation is repeated until the deviation meets the requirements. The final internal and external orientation elements include camera internal orientation elements (focal length, principal point coordinates) and external orientation elements (X, Y, Z, ω, ..., ...) for each image. The solution results and their hash values ​​are uploaded to the blockchain via smart contracts (such as those written in Solidity) to ensure data traceability.

[0024] S013, combining the final interior and exterior orientation elements with the preprocessed data, constructs an association mapping model of "image pixel-point cloud coordinates" at the edge computing node, performs dense matching, and generates the initial three-dimensional point cloud.

[0025] In one specific implementation, the step of constructing an association mapping model of "image pixels-point cloud coordinates," performing dense matching, and generating an initial 3D point cloud includes: Based on the final interior and exterior orientation elements and the preprocessed image data, key point detection and descriptor generation operations are performed using the scale-invariant feature transformation algorithm to obtain the image feature point set. Based on the image feature point set, a random sampling consensus algorithm is used to perform mismatch removal operations to obtain optimized matching point pairs; Based on the optimized matching point pairs and the final interior and exterior orientation elements, a sparse 3D point cloud is obtained by performing a forward intersection algorithm to calculate the 3D coordinates. Based on the sparse 3D point cloud and the preprocessed image data, a pixel-level disparity calculation operation is performed using a semi-global matching algorithm to generate a dense 3D point cloud as the initial 3D point cloud.

[0026] Specifically, the preprocessed image data is first processed using the Scale Invariant Feature Transform (SIFT) algorithm: A Gaussian pyramid scale space is constructed using each image as input (e.g., the number of scale layers is set to 4, and the Gaussian blur coefficient for each layer increases from a base of 1.6), and local extrema are detected as candidate keypoints. Keypoint localization accuracy is achieved by fitting sub-pixel positions using a 3D quadratic function (offset tolerance less than 0.5 pixels), and low-contrast points (contrast threshold dynamically set based on the image signal-to-noise ratio, e.g., a threshold of 0.03 when the signal-to-noise ratio is ≥30dB) and edge response points (principal curvature ratio threshold set to 10) are removed. A 128-dimensional descriptor is generated for each keypoint: a 16×16 pixel region centered on the keypoint is selected, an 8-directional gradient histogram is calculated, and normalized to unit length to enhance illumination invariance. The final output is a set of image feature points containing the keypoint coordinates, scale, orientation, and descriptor. Next, mismatch removal is performed: based on the image feature point set, the k-nearest neighbor algorithm (k=2) is used for initial matching, calculating the descriptor Euclidean distance ratio (the ratio of the nearest neighbor distance to the second nearest neighbor distance). If the ratio is less than a preset threshold (determined to be 0.7 through statistical matching reliability experiments), it is retained as a candidate matching point pair. Subsequently, the Random Sample Consensus (RANSAC) algorithm is applied for optimization: a minimum sample set (e.g., 4 pairs of matching points) is randomly selected to calculate the fundamental matrix, and the number of iterations is adaptively adjusted according to the inlier ratio (e.g., the confidence level is set to 0.99, and the initial estimate of the inlier ratio is 0.5). In each iteration, the distance from the matching point pair to the epipolar line is calculated. If the distance is less than a threshold (e.g., 1 pixel), it is marked as an inlier. Finally, the matching point pair corresponding to the model with the highest inlier ratio is retained, and the optimized matching point pair is output.

[0027] Based on optimized matching point pairs and final interior / exterior orientation elements, forward intersection calculation is performed: For each pair of matching points, a collinearity condition equation is constructed, associating the image point coordinates (x, y) with the object coordinates (X, Y, Z). The least squares method is used to solve the overdetermined equation system (e.g., requiring matching points from at least two images), iteratively solving for the object coordinates (e.g., using the Levenberg-Marquardt algorithm), with the convergence condition set at a coordinate update amount less than 1e-6 meters. Simultaneously, the reprojection error is calculated; if the error exceeds a threshold (e.g., 2 pixels), the point is discarded. A sparse 3D point cloud is output, with the point cloud density determined by the number of matching points (e.g., extracting an average of 500 feature points per image, generating approximately 10^4 sparse points). Finally, a semi-global matching (SGM) algorithm is executed to generate a dense point cloud: using the sparse 3D point cloud as initial constraints, a disparity space image is constructed from the preprocessed image data. The matching cost calculation employs a fusion of Census transform and gradient information: the Census transform calculates the local binary pattern using a 9×7 pixel window, and gradient information is extracted using the Sobel operator (with a gradient weight set to 0.5). Cost aggregation is performed along eight paths (0°, 45°, 90°, 135°, and the opposite direction), with penalty coefficients P1 and P2 set according to image texture richness (e.g., P1=5, P2=20 for texture-rich regions; P1=2, P2=10 for smooth regions). A winner-takes-all strategy is used for disparity calculation, with sub-pixel accuracy optimized through quadratic interpolation. The disparity range is dynamically set based on the depth prior provided by the sparse point cloud (e.g., depth range ±10% as the disparity search interval). The disparity map is converted to 3D coordinates, generating a dense 3D point cloud (point spacing 1-2 times the image ground sampling distance) as the initial 3D point cloud.

[0028] This step, through cascaded processing of feature matching and dense reconstruction, ensures that the point cloud covers the entire measurement area while maintaining geometric accuracy, providing foundational data for subsequent accuracy verification. For example, sparse point clouds are used to stabilize the model framework, while dense point clouds are used to refine the surface morphology; the combination of the two improves the reliability and completeness of the 3D reconstruction.

[0029] S014: Using high-precision checkpoint data measured in the field, verify the coordinate error of the initial 3D point cloud; if the error exceeds the limit, generate an accuracy correction instruction and feed it back to the dense matching step to regenerate the point cloud; after verification, store the data digest (such as SHA-256 hash value) and accuracy report of the valid 3D point cloud on the blockchain.

[0030] In one specific implementation, verifying the coordinate error of the initial 3D point cloud using high-precision checkpoint data measured in the field includes: Based on the initial 3D point cloud and the high-precision checkpoint data measured in the field, a point cloud-checkpoint registration operation is performed using the nearest neighbor search algorithm to obtain a set of registration point pairs. Based on the set of registration point pairs, coordinate system one operation is performed through a seven-parameter coordinate transformation model to obtain the transformed checkpoint coordinates; Based on the transformed checkpoint coordinates and the corresponding initial 3D point cloud coordinates, the coordinate residual sequence is calculated using Euclidean distance. Based on the coordinate residual sequence, error distribution analysis is performed through statistical hypothesis testing, and when the error exceeds the limit, a precision correction instruction containing correction parameters is generated. According to the accuracy correction instruction, adjust the weight of the matching cost function of the dense matching algorithm and regenerate the point cloud; Once verified, the valid 3D point cloud and its accuracy report will be stored on the blockchain network via a smart contract.

[0031] Specifically, the first step is to perform point cloud-checkpoint registration: using a kd-tree nearest neighbor search algorithm, the high-precision checkpoint coordinates (obtained through total station or RTK measurements, with a horizontal accuracy ≤2cm and an vertical accuracy ≤3cm) are used as query points to search for nearest neighbors in the initial 3D point cloud. The search radius is set according to the point cloud density (e.g., 3 times the average point spacing) to ensure that each checkpoint can find a corresponding point cloud point. For successfully matched point pairs, the distance between the two points is calculated. If the distance exceeds the tolerance threshold (e.g., 5 times the point cloud accuracy), it is considered an invalid match and is discarded. The final output is a set of registered point pairs containing the checkpoint coordinates and the corresponding point cloud point coordinates. Next, coordinate system one is performed: using a seven-parameter Helmholtz coordinate transformation model (including 3 translation parameters, 3 rotation parameters, and 1 scale parameter), the transformation parameters are solved using the least squares method based on the set of registered point pairs. The transformation accuracy is evaluated using the root mean square error (RMSE). If the RMS error exceeds a threshold (e.g., 1.5 times the accuracy of the checkpoints), an iterative weighted least squares method is used to dynamically adjust the weights of point pairs based on the residual magnitude (the weights of point pairs with larger residuals are reduced). After completing coordinate system one, the transformed checkpoint coordinates are output (consistent with the point cloud coordinate system). Then, the coordinate residual sequence is calculated: for each registered point pair, the Euclidean distance between the transformed checkpoint coordinates and the corresponding point cloud point coordinates is calculated, generating a coordinate residual sequence. The residual calculation covers both planar position residuals and elevation residuals, and their average, maximum, and standard deviations are calculated respectively.

[0032] Error distribution analysis is performed based on the coordinate residual sequence: the Kolmogorov-Smirnov test is used to determine whether the residuals follow a normal distribution. If the test passes (p-value > 0.05), the error range of the 95% confidence interval is calculated as the accuracy index; if the test fails, the Chebyshev inequality is used to determine the error boundary. The calculated error range is compared with a preset accuracy threshold (for example, in topographic surveying, the horizontal position error threshold is set to 0.1m, and the elevation error threshold is set to 0.15m). If the error in any direction exceeds the limit, a precision correction instruction is generated based on the spatial distribution characteristics and statistical properties of the coordinate residual sequence. The precision correction instruction is used to guide the point cloud generation module to adjust the parameter configuration of the dense matching algorithm (such as the weight of the matching cost function, the disparity search range, etc.) or trigger the re-acquisition and processing of data in a specific area. Specifically, if the error in any direction exceeds the limit, a precision correction instruction is generated, and the instruction content includes: the image area to be matched (located according to the residual distribution), the adjustment ratio of the gray-level similarity weight and the geometric constraint weight in the matching cost function (for example, the geometric constraint weight is increased by 20% in areas with large residuals), and the correction value of the disparity search range (dynamically shrinking or expanding according to the error size).

[0033] The dense matching process is adjusted according to the accuracy correction instructions: the cost function of the semi-global matching algorithm is modified, for example, by increasing the weight coefficient of geometric continuity constraints in the error-exceeding region (adjusting it from the default value of 1.0 to 1.2), while simultaneously narrowing the disparity search range (determining the search interval offset based on the error vector direction). Dense matching is re-executed to generate the corrected 3D point cloud. After successful verification, the valid 3D point cloud and its accuracy report (including indicators such as plane mean square error, elevation mean square error, and maximum residual) are stored as evidence via a smart contract. The smart contract automatically generates a data hash (using the SHA-256 algorithm) and binds the hash value to a storage address (such as an IPFS content identifier) ​​before writing it to the blockchain. A multi-node verification mechanism (such as PBFT consensus) is triggered during evidence storage to ensure the data is immutable.

[0034] This step achieves closed-loop control of point cloud accuracy through statistical testing and dynamic parameter adjustment. For example, adaptive weight adjustment based on residual distribution effectively suppresses the accumulation of systematic errors, ensuring that the 3D model meets engineering accuracy requirements. The generated accuracy report provides a reliable basis for subsequent applications, such as determining the sensitivity threshold for structural deformation monitoring in civil engineering based on point cloud accuracy.

[0035] S015, the effective 3D point cloud is meshed, and the preprocessed image data is mapped as a texture map to generate a 3D visualization model; the final result data package is split and stored in multiple edge nodes, and its index address and integrity verification value are stored in the blockchain network to realize distributed management and tamper-proof traceability of the results.

[0036] In one specific implementation, the meshing and texture mapping of the effective 3D point cloud includes: Based on the effective 3D point cloud, the point cloud density field is calculated and the isosurface is extracted using the Poisson surface reconstruction algorithm to obtain the initial triangular mesh model. Based on the initial triangular mesh model, a simplified mesh model is obtained by optimizing the number of vertices using a mesh simplification algorithm based on edge folding. Based on the simplified mesh model, the mesh surface is smoothed using the Laplace smoothing algorithm to obtain the optimized mesh model; Based on the optimized mesh model and the preprocessed image data, image registration and color fusion operations are performed using a multi-view texture mapping algorithm to generate a textured 3D model. Based on the textured 3D model, the model data is segmented using a spatial block algorithm to generate multiple model data blocks. The multiple model data blocks are distributed and stored on edge nodes, and the integrity check value of each data block is calculated. The index address and integrity verification value of the model data block are stored on the blockchain network through a smart contract.

[0037] Specifically, Poisson surface reconstruction is first performed based on the effective 3D point cloud: an octree data structure is used to hierarchically divide the point cloud space, with the depth adaptively set according to the point cloud density (e.g., a depth of 8 when the average point spacing is 0.1m). The point cloud density field is calculated by solving the Poisson equation (e.g., using the conjugate gradient method for iterative solution) to obtain the scalar field. The isosurface threshold is set according to the point cloud noise level (e.g., determined by multiplying the point cloud standard deviation by a coefficient of 0.5). The isosurface is extracted to generate an initial triangular mesh model, ensuring surface closure (e.g., no holes or defects). The mesh resolution is controlled by the size of the octree leaf nodes, for example, set to 1.5 times the average point spacing of the point cloud. Next, the initial triangular mesh model is simplified: an edge-folding-based algorithm is used to calculate the folding cost of each edge (e.g., based on a quadratic error metric), prioritizing the folding of edges with the lowest cost. The simplification objective is controlled by a threshold for the number of mesh faces (e.g., simplifying to less than 500,000 faces when the original mesh faces exceed 1 million). During the folding process, feature edges (such as edges with curvature changes greater than 0.1) are preserved to avoid the loss of important geometric details. A simplified mesh model is output, keeping geometric errors within acceptable limits (e.g., Hausdorff distance less than twice the point cloud accuracy). Then, mesh smoothing is performed: a Laplacian smoothing algorithm is applied, iteratively updating the position of each vertex (e.g., setting the iteration count to 10). Smoothing weights are allocated based on the inverse distance to neighboring vertices, while feature protection constraints are introduced (e.g., setting the normal vector change threshold to 15°) to prevent over-smoothing from blurring sharp edges. An optimized mesh model is output, significantly reducing surface roughness (e.g., reducing the average curvature variance by more than 30%). Texture mapping is performed based on the optimized mesh model: first, preprocessed image data is aligned with the mesh model through multi-view image registration. A feature-point-based registration method (e.g., SIFT feature matching) is used to calculate the mapping relationship between the image and the mesh UV coordinates (e.g., using a perspective transformation model). For each triangular facet, the image with the view angle closest to vertical is selected as the texture source (e.g., view angle less than 30°). If multiple images cover the same patch, color fusion is performed (e.g., weighted averaging, with weights calculated based on image resolution and lighting consistency). Texture resolution is set according to the grid size (e.g., 512×512 pixels per meter) to generate a textured 3D model. The textured 3D model is then divided into data blocks: a spatial partitioning algorithm (e.g., octree partitioning) is used to divide the model into uniformly sized data blocks (e.g., each block covers a 10m×10m×10m space). The block size is dynamically adjusted based on the storage capacity of the edge nodes (e.g., when the node capacity is 1GB, the block size is set to 100MB). Each data block contains geometric data (vertex coordinates, patch indexes) and texture data (image maps), and block metadata (e.g., bounding box coordinates, data size) is generated. The partitioned model data blocks are distributed and stored across multiple edge nodes (e.g., storage nodes are allocated based on a consistent hashing algorithm).Calculate the integrity check value (e.g., SHA-256 hash value) for each data block and record the index address of the storage node (e.g., IPFS content identifier). Package the index address and hash value into a transaction via a smart contract and submit it to the blockchain network (e.g., using the PBFT consensus mechanism). Multi-node verification is triggered during notarization to ensure the data is immutable.

[0038] This step ensures the geometric accuracy of the model through Poisson reconstruction and mesh optimization, enhances visual realism through multi-view texture mapping, and improves data reliability through distributed storage combined with blockchain notarization. For example, the mesh simplification rate is controlled within 50% to ensure the accuracy requirements of engineering applications, texture fusion eliminates image stitching gaps, block storage supports rapid loading of large-scale models, and integrity check values ​​prevent data transmission corruption, thus achieving efficient and reliable 3D result management.

[0039] Accordingly, such as Figure 2 As shown, based on an integrated photogrammetric control method, this embodiment of the invention provides an integrated photogrammetric control system to implement the integrated photogrammetric control method of this embodiment. The system includes: The multi-source data acquisition module is used to synchronously acquire multi-source raw data of the area to be measured through an integrated data acquisition terminal. The multi-source raw data includes sequential image data, initial point cloud data, and real-time attitude parameter data. The data preprocessing module is used to perform joint preprocessing of the multi-source raw data, including timestamp alignment, point cloud denoising, and image distortion correction, using edge computing nodes deployed near the data acquisition terminal, to generate preprocessed data, and upload the key logs and quality evaluation indicators of the preprocessing process to the blockchain network for evidence storage through a consensus mechanism. The adjustment and solution module is used to perform initial interior and exterior orientation elements of the image sequence using bundle adjustment at the edge computing node based on the preprocessed data. It calls the database of historical control points in the same area stored in the blockchain network for verification. If the coordinate deviation value exceeds the preset threshold, the adjustment weight coefficient is dynamically adjusted and the solution is recalculated. If the coordinate deviation value is less than or equal to the preset threshold, the currently calculated interior and exterior orientation elements are determined to be the final interior and exterior orientation elements, and the solution results and hash values ​​of the final interior and exterior orientation elements are uploaded to the blockchain network. The point cloud generation module is used to combine the final interior and exterior orientation elements with the preprocessed data to build an association mapping model of "image pixel - point cloud coordinates" at the edge computing node, perform dense matching, and generate the initial 3D point cloud. The point cloud verification module is used to verify the coordinate error of the initial 3D point cloud using high-precision checkpoint data measured in the field. If the error exceeds the limit, a precision correction instruction is generated and fed back to the point cloud generation module to regenerate the point cloud. After the verification is passed, the valid 3D point cloud and its precision report are stored in the blockchain network. A 3D modeling module is used to mesh the effective 3D point cloud and map the preprocessed image data as texture maps to generate a 3D visualization model. The final result data package is split and stored on multiple edge nodes, and its index address and integrity verification value are stored on a blockchain network to achieve distributed management and tamper-proof traceability of the results. It should be noted that the photogrammetric integrated control system provided in this embodiment of the invention executes all the process steps of the photogrammetric integrated control method described in the above embodiments. The working principles and beneficial effects of both correspond one-to-one, and therefore will not be elaborated further. This embodiment of the invention also provides an electronic device. The electronic device includes: a processor, a memory, and a computer program stored in the memory and executable on the processor, such as a photogrammetric integrated control program. When the processor executes the computer program, it implements the steps in the various photogrammetric integrated control method embodiments described above, for example... Figure 1 The steps S011 to S015 are shown. Alternatively, when the processor executes the computer program, it implements the functions of each module / unit in the above-described device embodiments, such as the photogrammetric integrated control module.

[0040] For example, the computer program may be divided into one or more modules / units, which are stored in the memory and executed by the processor to complete the present invention. The one or more modules / units may be a series of computer program instruction segments capable of performing a specific function, which describe the execution process of the computer program in the electronic device.

[0041] The electronic device may be a desktop computer, laptop, handheld computer, or smart tablet, etc. The electronic device may include, but is not limited to, a processor and memory. Those skilled in the art will understand that the above components are merely examples of electronic devices and do not constitute a limitation on the electronic device. It may include more or fewer components than described above, or combine certain components, or different components. For example, the electronic device may also include input / output devices, network access devices, buses, etc.

[0042] The processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor. The processor is the control center of the electronic device, connecting all parts of the electronic device via various interfaces and lines.

[0043] The memory can be used to store the computer programs and / or modules. The processor implements various functions of the electronic device by running or executing the computer programs and / or modules stored in the memory and by calling data stored in the memory. The memory may mainly include a program storage area and a data storage area. The program storage area may store the operating system, at least one application program required for a function (such as sound playback function, image playback function, etc.), etc.; the data storage area may store data created according to the use of the mobile phone (such as audio data, phonebook, etc.). In addition, the memory may include high-speed random access memory, and may also include non-volatile memory, such as hard disk, memory, plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, at least one disk storage device, flash memory device, or other volatile solid-state storage device.

[0044] Wherein, if the modules / units integrated in the electronic device are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments of the present invention can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc. It should be noted that the content included in the computer-readable medium can be appropriately added or removed according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media do not include electrical carrier signals and telecommunication signals.

[0045] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A photogrammetric integrated control method, characterized in that, include: S011, multi-source raw data of the area to be measured is synchronously acquired through an integrated data acquisition terminal. The multi-source raw data includes sequential image data, initial point cloud data and real-time attitude parameter data. The multi-source raw data is jointly preprocessed to generate preprocessed data. The edge computing node uploads the key logs and quality evaluation indicators of the preprocessing process to the blockchain network for evidence storage through a consensus mechanism. S012, based on the preprocessed data, the initial interior and exterior orientation elements of the sequence image are solved by bundle adjustment at the edge computing node; the database of historical control points in the same area stored in the blockchain network is called for verification. If the coordinate deviation value exceeds the preset threshold, the adjustment weight coefficient is dynamically adjusted and recalculated. If the coordinate deviation value is less than or equal to the preset threshold, the interior and exterior orientation elements currently solved are determined to be the final interior and exterior orientation elements. The data summary of the solution results of the final interior and exterior orientation elements is uploaded to the blockchain network. S013, combining the final interior and exterior orientation elements with the preprocessed data, constructs an association mapping model of "image pixel - point cloud coordinates" at the edge computing node, performs dense matching, and generates the initial three-dimensional point cloud; S014: Using high-precision checkpoint data measured in the field, verify the coordinate error of the initial 3D point cloud; if the error exceeds the limit, generate an accuracy correction instruction and feed it back to the dense matching step to regenerate the point cloud; after verification, store the data summary and accuracy report of the effective 3D point cloud on the blockchain. S015. The effective 3D point cloud is meshed, and the preprocessed image data is mapped as a texture map to generate a 3D visualization model; the final result data package is split and stored in multiple edge nodes, and its index address and integrity verification value are stored in the blockchain network.

2. The control method as described in claim 1, characterized in that, The multi-source raw data is subjected to joint preprocessing, including timestamp alignment, point cloud denoising, and image distortion correction, using edge computing nodes deployed near the data acquisition terminal to generate preprocessed data. The preprocessed data includes preprocessed image data and preprocessed point cloud data.

3. The control method as described in claim 2, characterized in that, The joint preprocessing includes: Timestamp alignment is based on hardware synchronization signals, which unifies the timestamps of image data, point cloud data and attitude data to the Global Positioning System. Point cloud denoising employs a statistical filtering algorithm to calculate the local density of the point cloud and remove outliers. Image distortion correction employs the Brown-Conrad model, using camera calibration parameters for geometric correction.

4. The control method as described in claim 1, characterized in that, In step S012, the bundle adjustment uses the collinearity condition equation to construct the error equation, with the image point observation value as the observation value and the exterior orientation element and object point coordinates as the unknowns. The Levenburg-Marquardt algorithm is used for iterative solution. The historical control point database contains at least three uniformly distributed control points, and the preset thresholds include the plane position deviation threshold and the elevation deviation threshold.

5. The control method as described in claim 1, characterized in that, Step S013 involves constructing the association mapping model between image pixels and point cloud coordinates, including: Based on the final interior and exterior orientation elements and the preprocessed image data, key point detection and descriptor generation operations are performed using the scale-invariant feature transformation algorithm to obtain the image feature point set. Based on the image feature point set, a random sampling consensus algorithm is used to remove mismatches and obtain optimized matching point pairs. Based on the optimized matching point pairs and the final interior and exterior orientation elements, the three-dimensional coordinate calculation operation is performed through the forward intersection algorithm to obtain a sparse three-dimensional point cloud; Based on sparse 3D point cloud and preprocessed image data, a pixel-level disparity calculation operation is performed using a semi-global matching algorithm to generate a dense 3D point cloud as the initial 3D point cloud.

6. The control method as described in claim 1, characterized in that, The verification of the coordinate error of the initial 3D point cloud in step S014 includes: Based on the initial 3D point cloud and the high-precision checkpoint data measured in the field, the point cloud and checkpoints are registered using the nearest neighbor search algorithm to obtain a set of registered point pairs. Based on the set of registration point pairs, coordinate system one operation is performed through a seven-parameter coordinate transformation model to obtain the transformed checkpoint coordinates; Based on the transformed checkpoint coordinates and the corresponding initial 3D point cloud coordinates, the coordinate residual sequence is calculated using Euclidean distance. Based on the coordinate residual sequence, error distribution analysis is performed through statistical hypothesis testing. When the error exceeds the limit, a precision correction command containing correction parameters is generated.

7. The control method as described in claim 1, characterized in that, Step S015 involves meshing and texture mapping of the effective 3D point cloud, including: Based on the effective 3D point cloud, the point cloud density field is calculated and the isosurface is extracted using the Poisson surface reconstruction algorithm to obtain the initial triangular mesh model. Based on the initial triangular mesh model, a simplified mesh model is obtained by optimizing the number of vertices through a mesh simplification algorithm based on edge folding. Based on the simplified mesh model, the mesh surface is smoothed using the Laplacian smoothing algorithm to obtain the optimized mesh model; Based on the optimized mesh model and preprocessed image data, image registration and color fusion operations are performed using a multi-view texture mapping algorithm to generate a textured 3D model. The distributed storage of the final result data package in step S015 includes: Based on the textured 3D model, the model data is segmented using a spatial block algorithm to generate multiple model data blocks. Multiple model data blocks are distributed and stored on edge nodes, and the integrity check value of each data block is calculated. The index address and integrity verification value of the model data block are stored on the blockchain network through a smart contract.

8. A photogrammetric integrated control system for implementing the photogrammetric integrated control method according to any one of claims 1-7, characterized in that, The system includes: The multi-source data acquisition module is used to synchronously acquire multi-source raw data of the area to be measured through an integrated data acquisition terminal; The data preprocessing module is used to perform joint preprocessing on multi-source raw data using edge computing nodes to generate preprocessed data; The adjustment module is used to perform initial interior and exterior orientation elements of the image sequence using bundle adjustment based on the preprocessed data, and to verify and adjust them. The point cloud generation module is used to combine the final interior and exterior orientation elements with the preprocessed data to build an association mapping model and generate an initial 3D point cloud. The point cloud verification module is used to verify the coordinate errors of the initial 3D point cloud using high-precision checkpoint data and generate accuracy correction instructions. The 3D modeling module is used to perform meshing and texture mapping on effective 3D point clouds to generate 3D visualization models.

9. An electronic device comprising a processor, a memory, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the integrated photogrammetric control method as described in any one of claims 1-7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the integrated photogrammetric control method as described in any one of claims 1-7.

Images