A multi-source remote sensing data joint positioning method and system for urban areas

Abstract

The application provides a kind of urban area-oriented multi-source remote sensing data joint positioning method and system, comprising: obtaining several optical stereo images and corresponding connection point data, constructing optical stereo regional network, extracting laser elevation control point data and digital surface model data within the scope of optical stereo regional network;Calculate the ground coverage range of several optical stereo images, map the ground coverage range to the grid unit using the grid lookup table, obtain the accurate image coordinates of the laser points combined with the screened laser elevation control point data, and output the laser elevation control point set;Based on the rational function model and affine transformation, calculate the laser elevation control point set, and construct a multi-source remote sensing data joint adjustment function model;Solve the multi-source remote sensing data joint adjustment function model, form the observation error equation, and solve the observation error equation by heuristic search, output the adjustment model parameter result and the final connection point ground coordinate.

Description

Technical Field

[0001] This invention relates to the field of remote sensing image processing technology, and in particular to a method and system for joint positioning of multi-source remote sensing data for urban areas. Background Technology

[0002] The joint adjustment technique of high-resolution optical imagery and laser satellite altimetry data is an important research direction in the field of remote sensing and mapping. It significantly improves the accuracy of geospatial data by integrating the advantages of two different sensors. High-resolution optical satellite imagery (such as the domestically produced "Gaofen" series) provides rich two-dimensional texture and spectral information, while spaceborne laser altimetry systems (such as the laser altimeter onboard Gaofen-7) can directly acquire elevation data with sub-meter or even centimeter-level accuracy. The fusion processing of these two data sources is of groundbreaking significance for solving problems such as the difficulty in obtaining control points and insufficient elevation accuracy in traditional photogrammetry.

[0003] With the rapid development of aerospace technology, spaceborne laser altimetry technology has made significant progress. The Gaofen-7 satellite, China's first civilian sub-meter-level high-resolution optical transmission stereo mapping satellite, utilizes its dual-line array stereo camera and laser altimetry system to provide a novel solution for 1:10,000 scale stereo mapping. This multi-sensor joint adjustment technology not only serves the production of basic surveying and mapping products but also shows broad application prospects in fields such as urban planning, disaster monitoring, forestry surveys, and polar research.

[0004] However, most current research focuses on high-precision fusion mapping of large areas, and the elevation changes within these areas are relatively gradual. There is little research on smaller areas like cities, where elevation changes are more frequent and dramatic.

[0005] Therefore, it is necessary to propose new joint positioning methods based on the characteristics of urban areas and data from multiple remote sensing sources. Summary of the Invention

[0006] This invention provides a multi-source remote sensing data joint positioning method and system for urban areas, which solves the problem of difficulty in constructing elevation control benchmarks in areas with large surface elevation changes, such as cities, and achieves high-precision positioning in areas with drastic elevation changes.

[0007] In a first aspect, the present invention provides a method for joint localization of multi-source remote sensing data for urban areas, comprising:

[0008] Acquire several optical stereo images and corresponding connection point data, construct an optical stereo regional network, and extract laser elevation control point data and digital surface model data within the optical stereo regional network.

[0009] The laser elevation control point data is coarsely screened using digital surface model data, and then finely screened using attribute data of the laser elevation control point data to obtain the filtered laser elevation control point data.

[0010] The ground coverage of several optical stereo images is calculated. A grid lookup table is used to map the ground coverage to grid cells. The precise image-side coordinates of the laser points are obtained by combining the filtered laser elevation control point data. After connecting with the tie point data, the ground coordinates of the adjusted laser elevation control points are obtained by adjusting the control area network. The confidence ellipsoid screening mechanism is used to remove gross differences between the ground coordinates of the adjusted laser elevation control points and the original ground coordinates. The laser elevation control point set is then output.

[0011] Based on the rational function model and affine transformation to calculate the laser elevation control point set, a joint adjustment function model of multi-source remote sensing data is constructed.

[0012] The joint adjustment function model of multi-source remote sensing data is solved to form the observation error equation. The observation error equation is solved by heuristic search, and the adjustment model parameter results and the ground coordinates of the final tie point are output.

[0013] According to the present invention, a multi-source remote sensing data joint positioning method for urban areas is provided, which performs coarse screening of laser elevation control point data using digital surface model data, and fine screening using attribute data of the laser elevation control point data to obtain filtered laser elevation control point data, including:

[0014] Extract the elevation value of the location corresponding to the digital surface model in each laser elevation control point data, subtract the elevation value of the location corresponding to the digital surface model from the elevation value of the laser elevation control point data itself to obtain the elevation difference, and remove the elevation points whose elevation difference is greater than the preset difference threshold to obtain the coarsely screened laser elevation control points.

[0015] The laser elevation control points were screened using the surface slope, cloud cover, and observation time data. Elevation points outside the preset threshold ranges for surface slope, cloud cover, and observation time were removed to obtain the filtered laser elevation control point data.

[0016] According to the present invention, a multi-source remote sensing data joint positioning method for urban areas calculates the ground coverage of several optical stereo images, including:

[0017] Set an initial ground elevation value for any corner point on any scene image, where the initial ground elevation value is equal to the spatial rectangular coordinates corresponding to the regularized translation parameter of the elevation in the rational function model.

[0018] Based on the image-side coordinates of any corner point and the corresponding rational function model of each image, the spatial rectangular coordinates of the connection point of any corner point are calculated by spatial forward intersection assisted by the digital surface model. Based on the planar coordinate components in the spatial rectangular coordinates of the connection point, the elevation value is interpolated from the corresponding position of the digital elevation model.

[0019] Calculate the coordinate difference between the elevation value and the spatial rectangular coordinates of the connection point. If the coordinate difference is less than the given coordinate difference threshold, the calculation of the spatial rectangular coordinates of any corner point ends. Otherwise, repeat the calculation of the coordinate difference until the coordinate difference is less than the given coordinate difference threshold.

[0020] Repeat all the aforementioned steps until the spatial rectangular coordinates of the four corner points on any image are calculated.

[0021] Repeat all the aforementioned steps until the spatial rectangular coordinates of the four corner points on all scene images are calculated, and the ground coverage of several optical stereo images is obtained.

[0022] According to the present invention, a multi-source remote sensing data joint positioning method for urban areas is provided. This method uses a grid lookup table to map the ground coverage area to grid cells, combines filtered laser elevation control point data to obtain the precise image-side coordinates of laser points, and then performs uncontrolled area network adjustment after concatenation with tie point data to obtain the adjusted ground coordinates of the laser elevation control points. The method includes:

[0023] The coordinates of two diagonal points are determined based on the ground coverage area, and the bounding box is determined by the coordinates of the two diagonal points.

[0024] The grid size is determined based on the preset proportion of the laser footprint;

[0025] The number of grid points is calculated based on the coordinates of the two diagonal points and the grid size, and all grid point values ​​are initialized to -1.

[0026] Coverage grid points are obtained by statistically analyzing the coverage area of ​​a single image, and unique identifiers are stored in the coverage grid points.

[0027] Repeat the steps of determining the grid size until the coverage grid points are statistically obtained based on the coverage area of ​​a single image, until all image coverage grid points are determined;

[0028] Calculate the grid coverage of any laser elevation control point based on its planar coordinates and the size of its laser footprint.

[0029] Obtain the unique image identifier corresponding to the grid covering any laser elevation control point, and calculate the image-side coordinates of any laser elevation control point using the rational function model corresponding to the unique image identifier.

[0030] Using the image-side coordinates of any laser elevation control point as the center, a window of preset size is set, and the least squares matching method is used to match point by point within the preset size window to obtain the set of accurate image-side coordinates of the laser elevation control point, which constitutes the set of corresponding connection points.

[0031] By incorporating the set of homing points into the adjustment model, an uncontrolled area network adjustment is performed to obtain the ground coordinates of the laser elevation control points after adjustment.

[0032] According to the present invention, a multi-source remote sensing data joint positioning method for urban areas is provided. A confidence ellipsoid screening mechanism is used to remove gross discrepancies between the ground coordinates of adjusted laser elevation control points and the original ground coordinates, outputting a set of laser elevation control points, including:

[0033] Remove connection points with the same name that are only located in one image;

[0034] The difference between the ground coordinates of the laser elevation control points after adjustment (after removing the tie points) and the ground coordinates of the laser elevation control points before adjustment is calculated to obtain the set of coordinate difference values.

[0035] Calculate the mean and variance of the differences in the three-dimensional directions of the ground coordinates at the connection point based on the set of coordinate differences, and construct a confidence ellipsoid;

[0036] Outlier points located outside the confidence ellipsoid are removed, and the remaining laser elevation control points are used as the laser elevation control point set.

[0037] According to the present invention, a multi-source remote sensing data joint positioning method for urban areas is provided, which calculates a set of laser elevation control points based on a rational function model and affine transformation, and constructs a multi-source remote sensing data joint adjustment function model, including:

[0038] A rational function model is constructed from the set of laser elevation control points;

[0039] An affine transformation model is constructed based on the image-side coordinates of the connection points, and then the affine transformation model is added to the rational function model.

[0040] Determine the initial values ​​of the ground coordinates of the connection points and the initial values ​​of the parameters of the affine transformation model. The initial values ​​of the object coordinates of each connection point are obtained by forward intersection calculation from the initial rational function model of the image where each connection point is located.

[0041] The joint adjustment model is linearized to obtain a linearized joint adjustment model;

[0042] Error equations are constructed based on the linearized joint adjustment model. The error equations include the elevation control point error equation and the tie point error equation, which are composed of the partial derivative coefficient matrix, constant vector and weight matrix of the adjustment parameters to be solved.

[0043] By determining any constant from a set of uncorrelated observations and their corresponding variances, we can obtain the weights corresponding to any observation.

[0044] By combining the weights and error equations corresponding to arbitrary observations, a joint adjustment function model for multi-source remote sensing data is obtained.

[0045] According to the present invention, a multi-source remote sensing data joint positioning method for urban areas is provided, which solves the multi-source remote sensing data joint adjustment function model to form an observation error equation, performs a heuristic search to solve the observation error equation, and outputs the adjustment model parameter results and the final tie point ground coordinates, including:

[0046] For each scene image to be adjusted, a regular grid is divided on the corresponding image plane at a preset equal interval;

[0047] For the center point of each grid, the object point is obtained by forward intersection on any local elevation datum plane in the object space using the initial rational function model of the corresponding image. The image point and the object point constitute a set of virtual control points.

[0048] The virtual control points are treated as real control points with a preset accuracy. The observation error equations are constructed according to the regional network adjustment model based on the rational function model. The observation error equations constructed from all the image points of the connecting points and the image points of the virtual control points are converted into matrices.

[0049] The scaling factor for the number of connection points and virtual control points in each image scene is calculated, and the weights of all virtual control point image observations are multiplied by the scaling factor.

[0050] The coefficient matrix of the normal equation is stored using a sparse matrix, which stores the number of non-zero elements in any row, the column coordinates of each non-zero element, and the value of each non-zero element, thus forming a sparse matrix storage format.

[0051] Based on the sparse matrix storage format, the conjugate gradient method is used to heuristically solve the normal equations of the joint adjustment function model of multi-source remote sensing data, and obtain the corrections of the adjustment model parameters and the coordinates of ground tie points.

[0052] Calculate the back projection error of each connection point, and use the preset multiple error of the back projection error of all image points as the evaluation standard threshold. When the difference between the residual of the connection point and the mean residual is greater than the evaluation standard threshold, the connection point is determined to be a gross error point.

[0053] After each adjustment, the weight of each observation in the next iteration is calculated according to the weight function, and the weight is included in the adjustment calculation. The first weight function is used before the preset number of iterations, and the second weight function is used after the preset number of iterations is reached.

[0054] After multiple iterations, the adjustment model parameters and the final ground coordinates of the connection points are obtained.

[0055] Secondly, the present invention also provides a multi-source remote sensing data joint positioning system for urban areas, comprising:

[0056] The extraction module is used to acquire several optical stereo images and corresponding connection point data, construct an optical stereo area network, and extract laser elevation control point data and digital surface model data within the optical stereo area network.

[0057] The filtering module is used to perform coarse filtering of laser elevation control point data from digital surface model data, and fine filtering by combining the attribute data of laser elevation control point data to obtain filtered laser elevation control point data.

[0058] The calculation module is used to calculate the ground coverage of several optical stereo images. It uses a grid lookup table to map the ground coverage to grid cells, and combines the filtered laser elevation control point data to obtain the precise image-side coordinates of the laser points. After connecting with the tie point data, it performs adjustment without control area network to obtain the ground coordinates of the adjusted laser elevation control points. It uses a confidence ellipsoid screening mechanism to remove gross differences between the ground coordinates of the adjusted laser elevation control points and the original ground coordinates, and outputs the laser elevation control point set.

[0059] The module is used to calculate the laser elevation control point set based on the rational function model and affine transformation, and to construct a joint adjustment function model of multi-source remote sensing data.

[0060] The solution module is used to solve the joint adjustment function model of multi-source remote sensing data, form the observation error equation, perform heuristic search to solve the observation error equation, and output the adjustment model parameter results and the ground coordinates of the final tie points.

[0061] Thirdly, the present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the multi-source remote sensing data joint positioning method for urban areas as described above.

[0062] Fourthly, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the multi-source remote sensing data joint positioning method for urban areas as described above.

[0063] The multi-source remote sensing data joint positioning method and system for urban areas provided by this invention have the following beneficial effects:

[0064] (1) By using a reference digital surface model based on laser altimetry data for coarse screening and laser data attribute for fine screening, abnormal points such as cloud reflection can be quickly eliminated, and high-precision laser points can be retained as control points. This enables the rapid extraction of high-precision laser elevation control points in the implementation area and improves the efficiency of data preprocessing.

[0065] (2) In view of the problem that the laser elevation control points are difficult to be effectively utilized due to the large topographic relief and complex occlusion relationship in urban areas, a lookup table based on the ground grid is used to achieve rapid association between laser elevation control points and images using least squares matching; the ground coordinates of the laser points are obtained through free network adjustment and the difference between them and the original coordinates is further calculated. A confidence ellipsoid is constructed to statistically eliminate gross errors, ensuring reliable and high-precision elevation control point extraction in urban areas.

[0066] (3) An adaptive weighting strategy is proposed to quantify the accuracy proportion of elevation control points and connection points, so as to realize the accurate and reasonable construction of the multi-source remote sensing data adjustment stochastic model;

[0067] (4) Sparse matrix compression is used to reduce memory usage during computation and storage, giving the multi-source remote sensing data positioning model a universality of operating equipment. Combined with gradient heuristic search, it avoids getting stuck in local optima during multi-source remote sensing data settlement, while significantly improving the settlement speed of the model. Furthermore, the weighted iteration mechanism dynamically eliminates gross errors, ensuring the accuracy and timeliness of the joint positioning results of multi-source remote sensing data, and providing technical support for the construction and settlement of a high-precision spatial reference benchmark framework for urban areas. Attached Figure Description

[0068] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0069] Figure 1 This is one of the flowcharts illustrating the multi-source remote sensing data joint positioning method for urban areas provided by the present invention;

[0070] Figure 2 This is the second flowchart illustrating the multi-source remote sensing data joint positioning method for urban areas provided by the present invention.

[0071] Figure 3 This is a schematic diagram of constructing a grid lookup table provided by the present invention;

[0072] Figure 4 This is a schematic diagram of using a confidence ellipse to remove gross error laser elevation control points provided by the present invention;

[0073] Figure 5 This is a schematic diagram of the virtual control point generation method provided by the present invention;

[0074] Figure 6 This is a schematic diagram of the structure of the multi-source remote sensing data joint positioning system for urban areas provided by the present invention;

[0075] Figure 7 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0076] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0077] Figure 1 This is one of the flowcharts illustrating the multi-source remote sensing data joint positioning method for urban areas provided in this embodiment of the invention, such as... Figure 1 As shown, it includes:

[0078] Step 100: Acquire several optical stereo images and corresponding connection point data, construct an optical stereo regional network, and extract laser elevation control point data and digital surface model data within the optical stereo regional network.

[0079] Step 200: Perform coarse screening of laser elevation control point data using digital surface model data, and fine screening using attribute data of laser elevation control point data to obtain filtered laser elevation control point data;

[0080] Step 300: Calculate the ground coverage of several optical stereo images, map the ground coverage to grid cells using a grid lookup table, obtain the precise image-side coordinates of the laser points by combining the filtered laser elevation control point data, perform uncontrolled area network adjustment after connecting with the tie point data to obtain the ground coordinates of the adjusted laser elevation control points, use a confidence ellipsoid filtering mechanism to remove gross differences between the ground coordinates of the adjusted laser elevation control points and the original ground coordinates, and output the laser elevation control point set;

[0081] Step 400: Calculate the laser elevation control point set based on the rational function model and affine transformation, and construct a joint adjustment function model of multi-source remote sensing data;

[0082] Step 500: Solve the joint adjustment function model of multi-source remote sensing data to form the observation error equation. Perform a heuristic search to solve the observation error equation and output the adjustment model parameter results and the ground coordinates of the final tie point.

[0083] Specifically, such as Figure 2 As shown, the technical approach of this invention includes two parts: the construction of a joint adjustment function model of multi-source remote sensing data, and the solution of the joint adjustment function model of multi-source remote sensing data, thereby realizing high-precision positioning of multi-source remote sensing data.

[0084] The first step is data preparation. In this embodiment of the invention, several optical stereo images and their corresponding connection point data are prepared to construct an optical stereo area network. This includes a set of laser elevation control point data that can cover the corresponding range of the stereo area network, and a set of digital surface model (DSM) data that can cover the corresponding range of the stereo area network.

[0085] The second step is the extraction of laser elevation control point data. Based on the reference DSM (Digital Sounding Model), the acquired laser altimetry data is coarsely screened to remove laser points with significant elevation errors. Then, combined with the attribute information of the laser data, a finer screening is performed, retaining laser points with gentle terrain, low cloud cover, and high elevation accuracy as ground elevation control points. Through these two steps, usable laser point data are automatically obtained.

[0086] The third step is the extraction of the laser elevation control point set. First, using Rational Function Model (RFM) regularized elevation as the initial value, and combining corner coordinates and DSM elevation information, the ground coverage boundary of each image is determined. Second, a fast retrieval mechanism based on a regular grid is designed. By establishing a dynamic grid index covering the entire area, the coverage area of ​​each image is mapped to the corresponding grid cell, realizing the spatial association between laser points and candidate images. Based on this, the precise image coordinates of the laser points are obtained using a combination of RFM forward calculation and least squares matching. These coordinates are then jointly adjusted with automatically matched connection points to output the ground coordinates of the adjusted laser elevation control points, and the difference between these and the original ground coordinates is calculated. Then, a confidence ellipsoid screening mechanism is introduced to calculate the mean and variance of the difference, eliminating outliers that deviate from the confidence interval, ultimately forming a highly reliable laser elevation control point set.

[0087] The fourth step is to construct a joint adjustment function model for multi-source remote sensing data. Based on the Rational Polynomial Coefficients (RPC) imaging model in the optical image auxiliary data and the matched tie point data, the adjustment observations and parameters to be adjusted are identified, and a joint adjustment function model for multi-source remote sensing data is constructed. The adjustment model is then linearized and normalized. Then, based on the analysis of multi-source data observation errors, the relationship between the adjustment accuracy of the joint regional network of multi-source data and prior knowledge such as the proportion of various observations including tie points, geometric intersection conditions, and initial value accuracy is explored. An adaptive weighting model corresponding to the adjustment function model is established to determine the corresponding weights of tie point observations and elevation control points.

[0088] The fifth step is the adjustment model solution for the multi-source remote sensing data regional network. First, virtual control points are generated to ensure the positioning accuracy of the joint model under conditions without ground control. Second, an efficient and robust solution to the equations is achieved using a gradient-heuristic optimal solution search method based on sparse matrices. Third, bundle-level gross error removal is performed, and after each adjustment, the weight matrix is ​​updated based on the tie point residuals and adjustment parameter corrections to participate in the next adjustment solution. Finally, the high-precision adjustment model parameters and tie point ground coordinates of the multi-source data regional network are obtained.

[0089] In one embodiment, step 100 specifically includes:

[0090] Acquire regional images to construct a regional network. Use appropriate matching and outlier removal algorithms to obtain reliable connection points on each image. Obtain DEM data with sufficient accuracy and coverage of the entire region. Acquire a sufficient number of evenly distributed laser satellite altimetry data points for the implementation area.

[0091] In one embodiment, step 200 specifically includes:

[0092] Based on each laser altimeter data point, the elevation value of the corresponding location in the DSM is extracted. And the elevation value of the laser altimetry data itself. The calculation is relatively poor. Elevation points with elevation differences greater than a threshold are discarded as gross errors.

[0093]

[0094] in This indicates the absolute elevation accuracy of the reference DEM. An elevation difference greater than or equal to three times is considered accurate. The laser points are used to remove outliers, thus completing the initial screening based on the reference DEM.

[0095] The embodiment employs an empirical threshold method, further filtering based on attribute data provided by laser altimetry. This is achieved by analyzing the surface slope... Cloud cover Observation time These three attributes are used to set thresholds to achieve further filtering based on attribute data. Specifically, the refined filtering conditions based on attribute parameters are set as follows:

[0096]

[0097]

[0098]

[0099] in, The slope threshold is determined by experiments analyzing the relationship between elevation accuracy and slope. This is the cloud cover threshold, typically ranging from [0,2], meaning only data with cloud cover less than or equal to 20% is retained. This is the day / night threshold, usually set to 1, meaning only nighttime observation data is retained. However, considering that this would result in a loss of nearly 50% of the data, this condition will not be used when laser data is limited, i.e., it will be set to 0.

[0100] In one embodiment, step 300 specifically includes:

[0101] First, based on the initial RFM model of each image in the regional network, and combined with the DSM model of the implementation area, the corner coordinates of each image are calculated inversely to obtain the ground coverage of the image. The form of the RFM model is shown below:

[0102]

[0103]

[0104]

[0105] in, and To connect the image-side coordinates of the points, , , Let the ground coordinates of the connection point be... , , , Representing the ground coordinates of the connecting points respectively Let be a rational polynomial with respect to the independent variable. , For the general form of rational polynomials, , The coefficients of the rational polynomial. , , Let be the power of the ground coordinates in the rational polynomial, typically at most 3. , , Represents the ground coordinates of the connection points containing powers. Represent the ground coordinates of the connection points containing powers. Any power in the equation.

[0106] The specific steps include:

[0107] (1) Set the initial ground elevation value of a certain corner point on the image. It is usually equal to the regularized translation parameter of the RFM model elevation. The corresponding spatial rectangular coordinates;

[0108] (2) Based on the image-side coordinates of the corner points And the corresponding RFM, the spatial rectangular coordinates of the connection point of the corner can be calculated by DSM-assisted spatial forward intersection. Based on planar coordinate components An elevation value can be interpolated from the corresponding location on the DEM. ;

[0109] (3) Calculate the elevation value Rectangular coordinates of the connection point The difference between If the difference is less than a given threshold This threshold is typically defined as a difference of 0.5% between the two values ​​(any value can be used as a benchmark). If the difference is within this range, the spatial rectangular coordinates of the connection point are considered to be calculated. Otherwise, the above steps are repeated until the difference is less than the threshold. ;

[0110] (4) Repeat steps (1) to (3) until the spatial rectangular coordinates of the four corner points of the scene image are calculated. All calculations have been completed;

[0111] (5) Repeat steps (1) to (4) until the spatial rectangular coordinates of the four corner points of all scene images are calculated, and the ground coverage of each scene image is obtained.

[0112] Note that when there are many images, a single laser control point may be located in multiple images. To calculate its image point coordinates across all images, this embodiment of the invention uses a lookup table-like data structure for retrieval. For example... Figure 3 As shown, the specific steps are as follows:

[0113] (1) Determine a bounding box based on the overall ground coverage of the image. Its range can be determined using the coordinates of two diagonal points. express;

[0114] (2) Set the grid size To ensure that the laser elevation control points are projected onto the image as accurately as possible, The size can be set to 1 / 5 of the size of the laser footprint;

[0115] (3) Calculate the number of grid points This constructs a rule grid covering the entire implementation area and initializes the value of all grid points to -1.

[0116] (4) For a single image, count the number of grid points covered based on its coverage area and assign a unique identifier to each point. Stored at this grid point;

[0117] (5) Repeat steps (2) to (4) until all grid points covered by the images are determined.

[0118] Based on the generated grid points, the center coordinates and footprint size of the laser elevation control points, the image in which they are located can be determined, and their estimated image-space coordinates can be calculated according to the corresponding RFM model. Furthermore, the accurate image-space coordinates of the laser elevation control points are calculated using point-to-point matching. This is then incorporated into the connection points. After performing an adjustment of the uncontrolled area network, the adjusted object-space coordinates of the laser elevation control points can be obtained. The specific steps are as follows:

[0119] (1) For any laser elevation control point, based on its plane coordinates And the size of the footprint, calculate the grid it covers;

[0120] (2) Based on the unique image identifier corresponding to the overlay grid The image-side coordinates of the laser elevation control points are calculated using the corresponding RFM model.

[0121] (3) Using its image point coordinates as the center, set A search window of a specific size is used. A least-squares matching method is employed to match points point-by-point within the search window, obtaining a precise set of image-square coordinates for the laser elevation control points. And form a set of connected points with the same name. ;

[0122] (4) Incorporate these connection points into the adjustment model and perform uncontrolled area network adjustment to obtain the ground coordinates of the laser elevation control points after adjustment. .

[0123] Considering the significant variations in surface elevation and slope in urban areas, and the potential for obstruction in densely built-up areas, not all obtained laser elevation control points can be used as elevation control points in the final adjustment calculation. This embodiment of the invention first calculates the difference between the two sets of coordinates of the laser elevation control points, and then further sets a confidence ellipsoid to remove gross errors. The specific steps are as follows:

[0124] (1) Based on the laser elevation control points, the set of corresponding connection points is formed. First, remove points that are only located on one scene;

[0125] (2) Calculate the ground coordinates of the remaining laser elevation control points after adjustment. Compared with the ground coordinates before adjustment The difference is obtained by comparing the coordinates to obtain the set of coordinate differences. ;

[0126] (3) Further calculate its value based on this set of differences. Mean of directional difference and variance Based on this, a biased confidence ellipsoid expression is constructed:

[0127]

[0128] in, 、 、 Represents the ground coordinates of the connection point. The scale parameter of the ellipsoid is used to adjust the confidence interval;

[0129] (4) Due to } follows a normal distribution, therefore the statistics constructed from it It follows a chi-square distribution, and this invention believes that... The confidence interval for} is typically 95%, at which point ;

[0130] (5) Obviously, in areas with gentle slopes, the coordinates of laser elevation control points are less likely to fall outside the confidence ellipse, while in areas with steeper slopes, they will fall outside the confidence ellipse. Therefore, if the coordinates of the laser elevation control points are less likely to fall outside the confidence ellipse in the residual space, then:

[0131]

[0132] Images outside the defined area are considered coarse errors and are discarded. The remaining laser elevation control points constitute the final set of elevation control points, such as... Figure 4 The diagram shows the use of confidence ellipses to eliminate gross laser elevation control points.

[0133] In one embodiment, step 400 includes:

[0134] Because RFM, based on a general imaging model, possesses excellent interpolation characteristics, high positioning accuracy can be achieved with a suitable amount of control information. Typically, to ensure positioning accuracy, current optical stereo satellites employ linear array pushbroom imaging systems. After rigorous on-orbit geometric calibration and sensor correction, the geometric errors of their single-scene image products are primarily low-order linear errors. Therefore, this embodiment of the invention uses the RFM model as the basic model for multi-source remote sensing data positioning and selects an image-side supplementary model. and The specific construction steps for the affine transformation model are as follows:

[0135] (1) Construct an RFM model based on the observations of the connection points obtained by matching optical stereo images;

[0136] (2) Based on the RFM model, an affine transformation model is constructed and added to the RFM model for subsequent correction of possible distortions in the image. The form of the affine transformation model is as follows:

[0137]

[0138]

[0139] in, and To connect the image-side coordinates of the points, , For the affine transformation model parameters, , Affine transformation model representing row and column directions;

[0140] (3) In order to solve the positioning parameters, the ground coordinates of the connection point and the parameters of the affine transformation model need to be assigned appropriate initial values. and The initial values ​​of the object coordinates of each connection point can be obtained by forward intersection calculation from the initial RFM of the image in which it is located.

[0141] (4) Linearize the joint adjustment model as shown in the following equation:

[0142]

[0143] in, , Represents the residual vector of the observed values ​​in the error equation. , For a term after the RPC model is expanded, 、 、 Represents the ground coordinates of the connection point. Rational terms in the x-direction abbreviation, Rational term in the y-direction abbreviation, To expand the ground coordinates of the connection point corresponding to the last item, The ground three-dimensional coordinates of the target point, The sign of the partial derivative;

[0144] (5) Based on the above function model, the error equation can be constructed as follows:

[0145]

[0146] Subscript and The equations represent the elevation control points and the connection points, respectively. Represent the equation of the connection point. Represent the equation of the elevation control point; and The parameters to be solved in the adjustment are represented by the image-side additional parameter vector and the object-side coordinate correction vector of the tie point in the RFM image to be adjusted, respectively. , These are the partial derivative coefficient matrices for the corresponding unknowns. and These are the corresponding constant vectors and weight matrices, respectively. for The partial derivative coefficient matrix of the connection points, for The partial derivative coefficient matrix of elevation control points, for The partial derivative coefficient matrix of the connection points, for The partial derivative coefficient matrix of elevation control points, and Let these represent the constant vector and weight matrix of the connection point, respectively. and Let represent the constant vector and weight matrix of the elevation control point, respectively.

[0147] In joint adjustment of multi-source satellite remote sensing data, introducing an accurate weighting model to relatively measure the accuracy of observations from different sources can achieve the effect of fully integrating the advantageous geometric information of multi-source satellite remote sensing data and suppressing the influence of low-precision observations.

[0148] According to photogrammetry theory, a set of uncorrelated observations is assumed. , Let be the number of observations, and their variances be respectively , ,..., If any constant is selected Then any observation value corresponding weights Defined as:

[0149]

[0150] Due to various objective factors present during the imaging process, heterogeneous satellite remote sensing data exhibit inconsistencies in elevation and planar accuracy. Simply assigning the same weight matrix to multiple data sources fails to differentiate the importance of observations with varying accuracies during joint adjustment. Therefore, it is necessary to comprehensively consider factors such as the proportion of different data types, initial value accuracy, and intersection conditions to construct an adaptive weighting model suitable for joint regional network adjustment of multi-source satellite remote sensing data.

[0151] In one embodiment, step 500 includes:

[0152] In the model solution process, the adjustment of multi-source remote sensing data regional networks often lacks control point constraints, resulting in a high degree of freedom in the adjustment model. Directly solving for the parameters to be adjusted as free unknowns can lead to ill-conditioned normal equation matrices. To address this issue, this embodiment of the invention proposes to optimize the model by adding virtual control points and incorporating the constructed virtual observation equations into the multi-source remote sensing data regional network adjustment model. Specifically:

[0153] (1) For each scene of image to be adjusted, divide its image plane into a regular grid at a certain interval, such as Figure 5 As shown;

[0154] (2) Using the initial RFM of the image, obtain the object point by forward intersection on any local elevation datum plane in the object space for the center point of each grid. The image point and the object point constitute a set of virtual control points;

[0155] (3) Treat the virtual control points as real control points with a certain accuracy, and construct the corresponding error equations according to the RFM-based regional network adjustment model. The observation error equations constructed from all connected point image points and virtual control point image points are written in matrix form, as shown in the following equations:

[0156]

[0157] Subscript , and These represent virtual control points, connection points, and elevation control points, respectively. and The parameters to be solved in the adjustment are represented by the image-side additional parameter vector and the object-side coordinate correction vector of the tie point in the RFM image to be adjusted, respectively. , These are the partial derivative coefficient matrices for the corresponding unknowns. and These are the corresponding constant vectors and weight matrices, respectively. Represent the equation of the virtual control point. for The virtual control point partial derivative coefficient matrix, and These represent the constant vector and weight matrix of the virtual control point, respectively;

[0158] (4) To avoid an imbalance in the number of virtual control points and tie points, which could lead to excessively different weights, the ratio factor of the number of tie points and virtual control points is calculated separately for each scene during adjustment. ,in, , These represent the number of connection points and virtual control points on the image, respectively. Then, the weights of the image point observations of all virtual control points are multiplied by this scaling factor.

[0159] Typically, multi-source remote sensing data regional networks use automatic matching methods to obtain tie points, resulting in a huge number of unknown parameters and thus a massive normal equation coefficient matrix for the adjustment model. To ensure the settlement efficiency and stability of the adjustment model on different devices, this invention stores the normal equation coefficient matrix in a sparse matrix format. For example, the row only stores the number of non-zero elements in the row, the column coordinates of each non-zero element, and its corresponding value, avoiding arbitrary operations on zero elements.

[0160] Building upon this foundation, the present invention first detects and eliminates any gross errors that may exist in the observations of the connection points. Secondly, it uses a heuristic solution algorithm to quickly and accurately solve the adjustment model of the multi-source remote sensing data regional network. Then, in each iteration of the solution process, a corresponding weight function is constructed based on the correction values ​​to continuously update the weights, thereby achieving accurate and robust solution calculation for the adjustment of the multi-source remote sensing data regional network. The specific steps are as follows:

[0161] (1) Based on the storage format of sparse matrices, the conjugate gradient method is used to heuristically solve the normal equations of the adjustment model of multi-source remote sensing data to obtain the adjustment model parameters. Coordinates of the connection point with the ground number of corrections;

[0162] (2) Calculate the back projection error of each connection point. The standard error is three times the mean error of all image point back projection errors. When the difference between the residual of a connecting point and the mean residual exceeds the standard error, that point is considered a potential gross error. The calculation formula is as follows:

[0163]

[0164] in, It is the first Image-side residuals at each connection point It is the mean of the residuals of all connection points on the image. It is the number of connection points on the image;

[0165] (3) After each adjustment, the weight of each observation in the next iteration is calculated according to the weight function and included in the adjustment calculation. Generally, two different functions are used. In the first three iterations, the weight function can take the following form:

[0166]

[0167] After the third iteration, the weight function is as follows:

[0168]

[0169] In the formula, Represents the residual vector of observed values. As a weighting factor, denoted as the standard deviation of the observed values.

[0170] (4) By solving the above multi-source remote sensing data regional network adjustment model, accurate adjustment model parameters and ground coordinates of connection points can be obtained, and a corresponding spatial reference benchmark can be constructed, thereby realizing high-precision positioning of multi-source remote sensing data in urban areas.

[0171] The following describes the multi-source remote sensing data joint positioning system for urban areas provided by the present invention. The multi-source remote sensing data joint positioning system for urban areas described below can be referred to in correspondence with the multi-source remote sensing data joint positioning method for urban areas described above.

[0172] Figure 6 This is a schematic diagram of the structure of a multi-source remote sensing data joint positioning system for urban areas provided in an embodiment of the present invention, as shown below. Figure 6 As shown, it includes: extraction module 61, filtering module 62, calculation module 63, construction module 64, and solution module 65, wherein:

[0173] Extraction module 61 is used to acquire several optical stereo images and corresponding tie point data, construct an optical stereo region network, and extract laser elevation control point data and digital surface model data within the optical stereo region network; Filtering module 62 is used to perform coarse filtering of the laser elevation control point data using the digital surface model data, and fine filtering using the attribute data of the laser elevation control point data to obtain the filtered laser elevation control point data; Calculation module 63 is used to calculate the ground coverage area of ​​several optical stereo images, using a grid lookup table to map the ground coverage area to grid cells, and combining the filtered laser elevation control point data to obtain the precise image-side coordinates of the laser points. After connecting with the tie point data, the ground coordinates of the adjusted laser elevation control points are obtained through uncontrolled area network adjustment. A confidence ellipsoid screening mechanism is used to remove gross discrepancies between the ground coordinates of the adjusted laser elevation control points and the original ground coordinates, and the laser elevation control point set is output. The construction module 64 is used to calculate the laser elevation control point set based on the rational function model and affine transformation, and construct a joint adjustment function model of multi-source remote sensing data. The solution module 65 is used to solve the joint adjustment function model of multi-source remote sensing data, form the observation error equation, perform a heuristic search to solve the observation error equation, and output the adjustment model parameter results and the final tie point ground coordinates.

[0174] Figure 7 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 7As shown, the electronic device may include a processor 710, a communication interface 720, a memory 730, and a communication bus 740. The processor 710, communication interface 720, and memory 730 communicate with each other via the communication bus 740. The processor 710 can call logical instructions in the memory 730 to execute a multi-source remote sensing data joint positioning method for urban areas. This method includes: acquiring several optical stereo images and corresponding tie point data; constructing an optical stereo area network; extracting laser elevation control point data and digital surface model data within the optical stereo area network; performing coarse screening of the laser elevation control point data using the digital surface model data; performing fine screening using the attribute data of the laser elevation control point data; calculating the ground coverage area of ​​several optical stereo images; mapping the ground coverage area to grid cells using a grid lookup table; and combining the screening... After selecting the laser elevation control point data, the precise image-side coordinates of the laser points are obtained. These coordinates are then concatenated with the tie-point data and adjusted using a controlless area network to obtain the ground coordinates of the adjusted laser elevation control points. A confidence ellipsoid screening mechanism is used to remove gross discrepancies between the adjusted ground coordinates and the original ground coordinates, outputting the laser elevation control point set. Based on a rational function model and affine transformation, the laser elevation control point set is calculated, and a joint adjustment function model of multi-source remote sensing data is constructed. This model is then solved to generate observation error equations. Heuristic search is used to solve these equations, outputting the adjustment model parameters and the final tie-point ground coordinates.

[0175] Furthermore, the logical instructions in the aforementioned memory 730 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, essentially, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0176] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for jointly positioning multi-source remote sensing data for urban areas, characterized in that, The application relates to a method for constructing a multi-source remote sensing data joint adjustment function model. The method comprises the following steps: acquiring a plurality of optical stereo images and corresponding connection point data, constructing an optical stereo regional network, extracting laser height control point data and digital surface model data within the range of the optical stereo regional network; roughly screening the laser height control point data from the digital surface model data, and combining attribute data of the laser height control point data to perform fine screening, so as to obtain screened laser height control point data; calculating the ground coverage range of the plurality of optical stereo images, mapping the ground coverage range to a grid unit by using a grid lookup table, combining the screened laser height control point data to obtain accurate image coordinates of laser points, connecting the laser points with the connection point data, performing a control-free regional network adjustment to obtain ground coordinates of the adjusted laser height control points, and removing coarse error points of the ground coordinates of the adjusted laser height control points from original ground coordinates by using a confidence ellipsoid screening mechanism, and outputting a laser height control point set; calculating the laser height control point set based on a rational function model and affine transformation, and constructing a multi-source remote sensing data joint adjustment function model; solving the multi-source remote sensing data joint adjustment function model, forming an observation error equation, and solving the observation error equation by using a heuristic search, and outputting adjustment model parameter results and final connection point ground coordinates; calculating the ground coverage range of the plurality of optical stereo images, comprising the following steps: setting an initial ground height value of any corner point on any image, wherein the initial ground height value is equal to a space rectangular coordinate corresponding to a regularization translation parameter of a rational function model height; calculating the space rectangular coordinate of the any corner point by using a digital surface model assisted space forward intersection according to image coordinates of the any corner point and corresponding rational function models of each image, and interpolating an elevation value from a digital elevation model according to a plane coordinate component in the space rectangular coordinate of the connection point; calculating a coordinate difference value between the elevation value and the space rectangular coordinate of the connection point, if the coordinate difference value is less than a given coordinate difference threshold value, the space rectangular coordinate calculation of the any corner point is ended, otherwise, the coordinate difference value is repeatedly calculated until the coordinate difference value is less than the given coordinate difference threshold value; repeating all the preceding steps until the space rectangular coordinate calculation of four corner points on any image is completed; repeating all the preceding steps until the space rectangular coordinate calculation of four corner points on all images is completed, and the ground coverage range of the plurality of optical stereo images is obtained; mapping the ground coverage range to a grid unit by using a grid lookup table, combining the screened laser height control point data to obtain accurate image coordinates of laser points, connecting the laser points with the connection point data, performing a control-free regional network adjustment to obtain ground coordinates of the adjusted laser height control points, comprising the following steps: determining two pairs of corner point coordinates according to the ground coverage range, and determining a bounding box according to the two pairs of corner point coordinates; determining the grid size according to a preset proportion of the laser footprint; calculating the number of grid points based on the two pairs of corner point coordinates and the grid size, and initializing all the grid point values to-1; counting the covered grid points according to the coverage range of an image, and storing a unique identifier to the covered grid points; repeating the step of determining the grid size according to the coverage range of an image to count the covered grid points until all the image coverage grid points are determined. Calculating a grid covered by any laser height control point based on the planar coordinates of the laser height control point and the size of the laser footprint; Obtaining the unique identifier of the image corresponding to the grid covered by any laser height control point, and calculating the image coordinate of any laser height control point by using the rational function model corresponding to the unique identifier of the image; Setting a window of a preset size with the image coordinate of any laser height control point as the center, and performing point-by-point matching in the window by using the least square matching method to obtain a set of accurate image coordinates of the laser height control point, which forms a set of homonymous connection points; Including the set of homonymous connection points in a adjustment model to perform area network adjustment without control to obtain the ground coordinates of the laser height control points after adjustment.

2. The urban area-oriented multi-source remote sensing data joint positioning method according to claim 1, characterized in that, Performing coarse screening on the laser height control point data by using the digital surface model data, and performing fine screening on the laser height control point data by combining the attribute data of the laser height control point data to obtain screened laser height control point data, including: Extracting the elevation value of the corresponding position of the digital surface model in each laser height control point data, and obtaining the elevation difference by subtracting the elevation value of the corresponding position of the digital surface model from the elevation value of the laser height control point data itself, and removing the elevation points with an elevation difference greater than a preset difference threshold to obtain the coarse screened laser height control points; Screening the coarse screened laser height control points by using the ground slope, cloud cover and observation time of the laser height control point data, and removing the elevation points outside the preset ground slope threshold range, cloud cover threshold range and observation time threshold range, respectively, to obtain the screened laser height control point data.

3. The urban area-oriented multi-source remote sensing data joint positioning method according to claim 1, characterized in that, Using a confidence ellipsoid screening mechanism to remove the gross error points of the ground coordinates of the laser height control points after adjustment and the original ground coordinates, and outputting the laser height control point set, including: Removing the connection points in the set of homonymous connection points that are located in only one image; Calculating the difference between the ground coordinates of the laser height control points after adjustment and the ground coordinates of the laser height control points before adjustment to obtain a set of coordinate difference values; Calculating the mean and variance of the difference in the three-dimensional direction of the ground coordinates of the connection points based on the set of coordinate difference values to construct a confidence ellipsoid; Removing the gross error points located outside the confidence ellipsoid, and taking the remaining laser height control points as the laser height control point set.

4. The urban area-oriented multi-source remote sensing data joint positioning method according to claim 1, characterized in that, Calculating the laser height control point set based on the rational function model and affine transformation, and constructing a joint adjustment function model of multi-source remote sensing data, including: Constructing a rational function model from the laser height control point set; Constructing an affine transformation model based on the image coordinates of the connection points, and adding the affine transformation model to the rational function model; Determining the initial values of the ground coordinates of the connection points and the initial values of the parameters of the affine transformation model, and calculating the initial object coordinates of each connection point by forward intersection based on the initial rational function model of the image in which each connection point is located; Performing linearization processing on the joint adjustment model to obtain a linearized joint adjustment model; Constructing an error equation based on the linearized joint adjustment model, and the error equation includes the error equations of the height control points and the connection points composed of the partial derivative coefficient matrix of the adjustment parameters to be solved, the constant vector and the weight matrix; Determining any constant based on a set of unrelated observation values and corresponding variances to obtain the weight value corresponding to any observation value; Combine the weight value corresponding to any observation value and the error equation to obtain the multi-source remote sensing data joint adjustment function model.

5. The urban area-oriented multi-source remote sensing data joint positioning method according to claim 1, characterized in that, Solve the multi-source remote sensing data joint adjustment function model to form an observation error equation, perform heuristic search on the observation error equation, and output adjustment model parameter results and final connection point ground coordinates, including: Divide the corresponding image plane into regular grids according to a preset equal interval division rule for each scene image to be adjusted; For the center point of each grid, use the initial rational function model of the corresponding image to obtain an object point on any local height reference surface through forward intersection, and form a group of virtual control points from the image points and the object points; Treat the virtual control points as real control points with a preset accuracy, construct an observation error equation according to a regional network adjustment model based on the rational function model, convert the observation error equation set constructed by all connection point image points and virtual control point image points into a matrix, multiply the weight values of all virtual control point image points by a proportional factor, and store the equation coefficient matrix in a sparse matrix storage method; Based on the storage format of the sparse matrix, use the conjugate gradient method to perform heuristic solving on the normal equation of the multi-source remote sensing data joint adjustment function model to obtain adjustment model parameters and correction numbers of the ground connection point coordinates; Calculate the back projection error of each connection point, and use a preset multiple error of all image back projection errors as a judgment standard threshold value; when the difference between the residual error of the connection point and the average residual error is greater than the judgment standard threshold value, the connection point is determined to be a gross error point; After each adjustment, calculate the weight value of each observation value in the next iteration according to the weight function, and include the weight value in the adjustment calculation, wherein the first weight function is used before a preset number of iterations, and the second weight function is used after the preset number of iterations. After multiple iterations, the adjustment model parameter results and the final connection point ground coordinates are obtained. The extraction module is configured to acquire a plurality of optical stereo images and corresponding connection point data, construct an optical stereo regional network, and extract laser height control point data and digital surface model data within the optical stereo regional network. The screening module is configured to coarsely screen the laser height control point data based on the digital surface model data, finely screen the laser height control point data based on attribute data of the laser height control point data, and obtain screened laser height control point data.

6. A multi-source remote sensing data joint positioning system for urban areas, based on the multi-source remote sensing data joint positioning method for urban areas according to any one of claims 1 to 5, characterized in that, The calculation module is configured to calculate ground coverage ranges of the plurality of optical stereo images, map the ground coverage ranges to grid units using a grid lookup table, obtain accurate image coordinates of laser points based on the screened laser height control point data, obtain ground coordinates of the adjusted laser height control points after connection point data connection and adjustment of a control-free regional network, remove gross error points of the ground coordinates of the adjusted laser height control points and the original ground coordinates using a confidence ellipsoid screening mechanism, and output a laser height control point set. The construction module is configured to calculate the laser height control point set based on a rational function model and an affine transformation, and construct a multi-source remote sensing data joint adjustment function model. ​ ​ ​ A solving module is configured to solve a joint adjustment function model of multi-source remote sensing data, form an observation error equation, perform heuristic search on the observation error equation, and output adjustment model parameter results and final connection point ground coordinates.

7. An electronic device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, The processor implements the urban area-oriented multi-source remote sensing data joint positioning method according to any one of claims 1 to 5 when executing the program.

8. A non-transitory computer-readable storage medium having stored thereon a computer program, characterized in that, The computer program implements the urban area-oriented multi-source remote sensing data joint positioning method according to any one of claims 1 to 5 when executed by the processor.

Images