Multi-process dual-line array image matching method based on laser elevation control point library

Abstract

The present invention proposes a multi-process dual-line array image matching method based on a laser elevation control point library, which belongs to the field of satellite laser ranging. The method includes two parts: a construction technology of a laser elevation control point library and a dual-line array image matching technology based on the laser elevation control point library. SQLite is used to build a laser elevation control point database. When building a database, a table storage mechanism and a naming rule are established; dual-line array image matching based on the laser elevation control point library mainly divides the rear-view image into blocks according to the construction mechanism of the laser elevation control point library and the imaging mechanism of the dual-line array image, and uses MPI multi-process to quickly traverse the laser elevation control point database to determine the laser elevation control points required for each block of the image to be processed. The method uses satellite-borne laser altimetry data as an elevation control constraint condition to participate in the regional network adjustment calculation of the image, which can effectively improve the accuracy of uncontrolled stereo mapping of optical satellite images, thereby realizing a large-scale stereo satellite mapping project without ground control points.

Description

[0001] Field of the Invention:

[0002] A multi-process dual-line array image matching method based on a laser elevation control point library belongs to the field of satellite laser ranging. Background Art:

[0003] Satellite images have the advantages of being unrestricted by national boundaries, having a strong periodicity, and a wide image range. Coupled with the continuously improving resolution, they have gradually replaced the previous aerial photogrammetry and become an important data source for current large-scale mapping production. Currently, the principles and methods of using stereo mapping satellites for global stereo mapping without ground control points are relatively mature, but they still cannot fully meet the requirements of large-scale mapping. Among them, the elevation accuracy is the main limiting factor. Laser altimetry is an active remote sensing means that can accurately obtain three-dimensional information of the ground. Satellite laser altimetry has the ability to actively obtain three-dimensional information of the global surface and targets, is not restricted by day or night, can obtain elevation control point data with centimeter-level high precision on the ground surface, and can provide services for quickly obtaining three-dimensional control points including overseas regions and stereo mapping. Using spaceborne laser altimetry data as an elevation control constraint condition to participate in the regional network adjustment calculation of images can effectively improve the accuracy of optical satellite image stereo mapping without control points, thus realizing large-scale stereo satellite mapping projects without ground control points. Summary of the Invention:

[0004] The present invention provides a dual-line array image matching technology based on a laser elevation control point library. The present invention includes two parts: the construction technology of the laser elevation control point library and the dual-line array image matching technology based on the laser elevation control point library.

[0005] (1) Construction of the Laser Elevation Control Point Library

[0006] The present invention uses SQLite to construct a laser elevation control point database. SQLite is a self-contained, serverless, zero-configuration, transactional SQL database engine, and its source code is not restricted by copyright. SQLite has the following advantages: it does not require installation or management; a complete SQLite database is stored in a single cross-platform disk file; SQLite is very small, lightweight, and does not require any external dependencies; SQLite transactions are fully compatible with ACID, allowing safe access from multiple processes or threads; SQLite supports most of the query language functions of the SQL92 (SQL2) standard; SQLite is written in ANSI-C and provides a simple and easy-to-use API; SQLite can run on UNIX (Linux, Mac OS-X, Android, iOS) and Windows (Win32, WinCE, WinRT).

[0007] The present invention establishes a sub-table storage mechanism and a naming rule when constructing a database, thereby improving the updating and searching efficiency of the database table. Firstly, the global (-180-180, -90-90) range is divided into standardized grid blocks in a 5 degree by 5 degree manner, and each standardized grid block is used as a database table. The table name is named in the form of a combination of the longitude and latitude integer value at the lower left corner of the grid block and the east longitude (E), west longitude (W), north latitude (N), and south latitude (S). For example, "E100N20" is a grid block data table of 100 degrees east longitude and 20 degrees north latitude. According to the standard 5-degree grid block, the control point range in the table can be known, and the sub-table storage improves the updating and searching efficiency of the database.

[0008] The present invention selects multiple parameters such as echo waveform oversaturation correction parameter, elevation of footprint point on reference DEM, deviation between mean position of echo Gaussian component and centroid position, air quality mark, elevation availability mark, laser pointing quality mark, PAD data indication mark, LPA status indication mark, surface reflectivity, number of Gaussian components of echo fitting waveform, standard deviation of echo Gaussian component, angle between laser pointing and nadir direction, length of echo effective signal, echo gain, etc. to screen and evaluate the quality of laser height measurement data. Only points that meet the screening conditions and have a certain quality score are selected as elevation control points, and are stored in the corresponding database table according to the latitude and longitude values ​​of the elevation control points. The technical solution for constructing a laser elevation control point library of the present invention includes the following steps:

[0009] ①Create an empty SQLite database;

[0010] ② The global area is evenly gridded at 5-degree intervals;

[0011] ③ Establish the database table name with the integer value of the longitude and latitude in the lower left corner of each grid block, such as "E100N20" for the grid block data table of 100 degrees east longitude and 20 degrees north latitude;

[0012] ④ Read GLAS data and use the laser elevation control point screening model to select high-precision control points;

[0013] ⑤ For the selected high-precision laser elevation control points, insert them into the corresponding database table for storage according to the longitude and latitude values ​​of each point.

[0014] (II) Dual-line image matching based on laser elevation control point library

[0015] Dual-line array image matching based on the laser elevation control point library is mainly based on the construction mechanism of the laser elevation control point library and the imaging mechanism of the dual-line array image. The rear-view image is divided into blocks and the laser elevation control point database is quickly traversed using MPI multi-process to determine the laser elevation control points required for each image to be processed. A dynamic pyramid layer is established around the laser elevation control point to eliminate the impact of scale changes. The laser elevation control point transfer work is performed in the pyramid layer with consistent resolution. The pyramid layer data of the search window size is established to reduce memory consumption and improve the matching success rate. The object-image relationship transformation model of the front and back views of the dual-line array image is established in blocks. The block mechanism can fully consider the texture of the object. The finer the block division, the more consistent the object category of each block can be. For example, the current block is all mountainous, and the other block is all plains. Such division can fully guarantee the accuracy of the relationship transformation model of the current block. An accurate transformation model can ensure the search window corresponding to the laser elevation control point. The accuracy of resampling is improved. The resampling search window can eliminate the distortion of the target window and the search window, and improve the success rate of matching. In order to increase the probability of laser elevation control point rotation, a circle with a radius of 25 pixels is drawn with the laser point as the center, and feature points are evenly extracted within the circle. The double-line array rotation is performed together with the laser point. Theoretically, the objects within the range of 50 pixels are basically unchanged, which can be understood as the same object. The elevation value of the same laser elevation control point can be used to ensure the success rate of laser elevation control point rotation. Based on the multi-task distribution parallel processing mechanism of MPI, the laser elevation control point rotation work can be performed multiple times to improve the matching efficiency. Through the above-mentioned various strategy mechanisms, the present invention effectively improves the success rate and efficiency of laser elevation control point rotation of double-line array images.

[0016] The dual-line array image matching solution based on the laser elevation control point library of the present invention comprises the following steps:

[0017] ① The main process performs uniform block processing of the rear view image;

[0018] ②Package each block of information as a subtask;

[0019] ③The main process pushes all task packages to each sub-process;

[0020] ④ In each subtask, search the laser elevation control point database according to the block range to determine the laser elevation control point set;

[0021] ⑤ Extract Forstner feature points within a circle with a radius of r (r is generally 25 to 50 pixels, and this algorithm takes it as 25) with each laser elevation control point as the center, and form a set of laser points and feature points;

[0022] ⑥ Use the RPC parameter file to calculate the image space relationship transformation model of the front and rear view images. The image space relationship transformation model is the image space relationship transformation model for point position conversion from the rear view image coordinate system to the front view image coordinate system;

[0023] ⑦ Perform pyramid resolution consistency processing and distortion resampling processing on each point, use the mean pyramid for resolution consistency processing, and use the image square transformation model to use the bilinear interpolation method to correct the distortion;

[0024] ⑧Perform dynamic threshold correlation coefficient texture matching according to each pyramid layer;

[0025] ⑨Use the image-space relation transformation model to transform the same-name point set;

[0026] ⑩Use the RANSAC method to eliminate gross errors and obtain the correct matching point set;

[0027] The subprocess task status is fed back to the main process;

[0028] The main process receives the return status of each subtask, and when all subtasks are completed, merges the files with the same name in each subprocess and outputs the results;

[0029] The main process ends, the overall task status is returned, and the processing flow ends.

[0030] Advantages of the invention:

[0031] (1) The laser elevation control point database is built using the lightweight, zero-configuration, single-file database SQLite, which is convenient for cross-platform porting.

[0032] (2) The laser elevation control point database creates database tables according to the latitude and longitude grids and stores data in separate tables to facilitate quick query and use of laser elevation control points.

[0033] (3) Sub-table storage facilitates deployment according to processing needs, without the need to migrate the laser elevation control point database as a whole.

[0034] (4) Using multiple parameters to screen and evaluate the quality of laser altimetry data improves the reliability of data screening and quality evaluation and ensures the accuracy of elevation control point data entered into the database.

[0035] (5) The search mechanism of the laser elevation control point library based on MPI multi-process improves the search speed.

[0036] (6) The image processing mechanism based on MPI multi-process blocks improves the conversion efficiency of laser elevation control points and dual-line array images.

[0037] (7) Dynamic pyramid, dynamic threshold, and relationship transformation model are used to transform the matching, which improves the matching success rate and processing efficiency.

[0038] (8) The strategy of combining laser points and feature points to transfer points improves the transfer success rate of laser elevation control points and ensures the use of regional network adjustment. At the same time, the introduction of laser elevation control points improves the elevation accuracy of the adjustment. Description of the drawings:

[0039] Figure 1 This is the laser elevation control point library construction process in the embodiment;

[0040] Figure 2 It is a dual-line array image matching process based on a laser elevation control point library in the embodiment;

[0041] Figure 3 is a schematic diagram of global standard degree uniform grid division in an embodiment;

[0042] Figure 4 Schematic diagram of laser elevation control point database in the embodiment;

[0043] Figure 5 : is a schematic diagram of the distribution of global laser elevation control points in the embodiment;

[0044] Figure 6 Schematic diagram of the distribution of laser elevation control points in some provinces of China in the embodiment. Specific implementation method:

[0045] (I) Construction of laser elevation control point library

[0046] The ICESat (Ice, Cloud, and land Elevation Satellite) earth observation satellite launched by the United States in 2003 is equipped with the world's first satellite-borne laser altimetry system for continuous observation of the earth, the Geo-science Laser Altimetry System GLAS. The satellite acquired a large amount of laser altimetry data on a global scale during 2003-2009. Many scholars at home and abroad have conducted research and experiments and shown that the elevation accuracy of the laser altimetry data acquired can reach 0.15 meters, which has been widely used in methods such as polar ice cap monitoring, global forest biomass estimation, and land elevation measurement. The present invention uses GLAS data as a data source to construct a global laser elevation control point database.

[0047] (1) Global standard grid division

[0048] Taking 5 degrees multiplied by 5 degrees as an example, the entire world (-180 to 180, -90 to 90) is evenly gridded, with 2592 grids. The standard grid division is as follows: Figure 3 shown.

[0049] (2) Database table naming

[0050] The database table name is named in the form of the combination of the longitude and latitude integer value of the lower left corner of the grid block and the east longitude (E), west longitude (W), north latitude (N), and south latitude (S). For example, "E100N20" is the grid block data table of 100 degrees east longitude and 20 degrees north latitude. The valid data range of longitude in the data table is [100,105) semi-closed and semi-open interval, and the valid range of latitude is [20,25) semi-closed and semi-open interval. The database table contains fields such as ObsDate, ObsTime, Source, Lon, Lat, Elev, TidalCorr, Quality, and ProduceTime. The database structure is shown in Table 1 and Figure 4 shown.

[0051] Table 1 Fields of laser elevation control point database table

[0052]

[0053] (3) Laser elevation control point screening and quality evaluation

[0054] Since the laser beam of the altimeter has a certain divergence angle and the atmosphere has a scattering effect on the beam, the footprint illuminated by the laser beam on the ground is a light spot with a certain area, and the diameter of the light spot is about tens of meters. Taking GLAS data as an example, the diameter of the footprint light spot is about 70 meters. The elevation measured by the laser altimeter data represents the average elevation of the entire light spot surface. The elevation accuracy will be affected by many factors such as terrain undulations, vegetation and other attachments, and the atmosphere within the light spot range. Not all laser footprint points can be used as elevation control points, and they must be screened and evaluated for quality according to certain rules.

[0055] The result data of GLAS official processing is released by NSIDC (National Snow and Ice Data Center). The standard products are divided into 3 levels and 15 categories. Among them, GLA14 is a land altimetry data file, which records the final data processing results of the land surface of the altimetry system, including more than 80 parameters such as the 3D coordinate value of the laser footprint, the oversaturation correction parameter of the echo waveform, the elevation of the footprint point on the reference DEM, the mean position of the echo Gaussian component and the centroid position deviation, the air quality mark, and the elevation availability mark. Some parameters can be used as the basis for the screening of GLAS laser elevation control points. The present invention first selects some parameters for preliminary screening of laser elevation control points, and the selected parameters and screening criteria are shown in Table 2. For the laser elevation control points that have been preliminarily screened, some parameters are selected for quality evaluation. Only records with a quality evaluation score greater than a certain threshold (here the threshold is set to 75 points) can be written into the database as the final laser elevation control point. The selected parameters and evaluation criteria are shown in Table 3. After screening and quality evaluation, it is guaranteed that the elevation accuracy of the selected laser elevation control points is better than 1 meter.

[0056] Table 2 Screening criteria for laser elevation control points

[0057]

[0058]

[0059] Table 3 Quality evaluation criteria for laser elevation control points

[0060]

[0061]

[0062] (4) Storage of laser elevation control points

[0063] ICESat collected about 2 billion elevation data from 2003 to 2009. After screening and quality evaluation, the present invention screened out 30.46 million laser elevation control points in the land area and entered them into the database. The global distribution of laser elevation control points is shown in the following figure. Figure 5 shown. Figure 6 This is the distribution of control points in Hebei, Shandong, Jiangsu, Anhui, Henan and Shanxi regions of China. It can be seen that the control points in the plain area are densely distributed, while the control points in the mountainous, hilly or forested areas (most areas of Shanxi Province, western Hebei Province, central Shandong Province, southern Anhui Province, western Henan Province) are relatively scarce, which meets the control point screening criteria of the present invention.

[0064] (II) Multi-process dual-line image matching based on laser control point library

[0065] (1) Laser control point library search

[0066] The dual-line array rearview image is evenly divided into blocks. The object space range of each block is calculated using the RPC parameters of the rearview image. The longitude and latitude coordinates corresponding to the four corner points (0,0), (w,0), (0,h), (w,h) of each image block are calculated, where w and h are the width and height of the image block. The longitude and latitude coordinates corresponding to the four corner points of each image block (X n , Y n ), take the largest circumscribed rectangle as the image range of the block. Then search for laser points within the range in the corresponding laser elevation control point database table according to the range to form a laser point set. Formula 1 represents the coordinates of each pixel point (x n ,y n ) and its corresponding ground longitude and latitude coordinates (X n , Y n , H n ) relationship.

[0067]

[0068]

[0069] In the formula, F 1 、F 2 、F 3 、F 4 is a general polynomial, calculated as follows:

[0070]

[0071] Where b ijk (i, j, k = 0, 1...20) are the polynomial coefficients; by using the inverse formula 3 of formula 1, the corresponding ground point longitude and latitude coordinates can be obtained. The calculation method is as follows:

[0072]

[0073] According to the range of the image block, the library table in the laser elevation control point library that overlaps with the block is determined. In the searched library table, a multi-transaction query search mechanism is used to quickly search for laser elevation control point data in multiple library tables within the range to obtain the laser elevation control point set. According to the latitude and longitude values ​​of the laser elevation control point, the image coordinates (x n ,y n ) to complete the transfer of laser height control points to the rearview image. At the same time, a circle with a radius of N pixels is established on the rearview image with each laser height control point as the center. N is generally taken as 25. An appropriate number of feature points are evenly taken within this circle to form a feature point set together with the laser height control points.

[0074] (2) Calculation of image-space transformation model

[0075] The relational transformation model is based on the transformation relationship between the rearview image coordinate system and the frontview image coordinate system. The latitude and longitude ranges of the four corner points obtained by the rearview image blocks are intersected with the latitude and longitude ranges of the four corner points of the frontview image to obtain the four corner coordinates (X n , Y n )(n=0,1,2,3), according to the above formula 1, the image coordinates of the four corner points of the overlapping area corresponding to the front view and the back view can be obtained (x' n , y' n )(n=0,1,2,3) and (x n ,y n )(n=0,1,2,3), the image-side transformation relationship model from the rear-view image coordinate system to the front-view image coordinate system can be obtained through the image-side coordinates. The relationship transformation model adopts a 6-parameter mode, as shown in Formula 4.

[0076]

[0077] a n (n=0,1,2,3,4,5) represents a 6-parameter relational transformation model, which can be solved according to the least squares fitting formula.

[0078] (3) Dynamic pyramid hierarchical relationship transformation model resampling processing

[0079] Because of the difference in the perspectives of the front and rear cameras, the texture of the object is distorted internally, and the correlation coefficient matching is not robust to distortion and scale, so the resolution consistency processing and distortion correction must be applied to the rear and front images first. This algorithm uses the establishment of a dynamic pyramid to eliminate the impact of inconsistent resolutions, and uses the relational transformation model to resample the front image to the rear image coordinate system to eliminate the impact of internal distortion. A dynamic pyramid layer is established for each point, and multi-layer pyramids are not created for the entire image. Under the premise of pyramid layers with consistent or similar resolutions, a pyramid layer target window of I (I is generally 4) layers of M*N (M and N are generally 17) is established on the rear image with each point as the center. According to the rear and front relationship transformation model, a pyramid layer of L*S (L and S are generally 61) search window size corresponding to the target window is established, and the front texture is correspondingly resampled to the search window. In this way, the coordinate system of the front data model is resampled to the coordinate system of the rear image to eliminate the impact of front and rear distortion.

[0080] (4) Dynamic Threshold Correlation Coefficient and Relation Transformation Model Matching

[0081] Adaptively select pyramid layers with the same or similar resolution to perform layer-by-layer correlation coefficient matching from low resolution to high resolution. In the pyramid matching process, the correlation coefficient matching adopts a dynamic matching threshold mechanism. The matching threshold of the pyramid layer with low resolution is lower (generally 0.5), and the matching threshold of the pyramid layer with high resolution is higher (generally 0.65).

[0082] The feature of the texture of the front and rear objects does not change much during the dual-line array image shooting process, and a matching strategy based on the correlation coefficient of grayscale is adopted. The matching strategy of the correlation coefficient has low calculation amount and fast speed compared with feature matching.

[0083] The matching principle of the correlation coefficient is:

[0084] The target window is moved in a zigzag route in the search window to calculate the correlation coefficient ρ(c,r), and the position with the largest correlation coefficient is taken as the best matching point position. The correlation coefficient calculation formula is as follows:

[0085]

[0086]

[0087]

[0088] Wherein, g and g′ are the corresponding pixel grayscale values ​​in the target window and the search window, i and j are the row and column coordinate numbers of the target window, and r and c are the row and column coordinate numbers of the corresponding search window; and is the calculation formula for the mean grayscale of the target window and the search window; define W1 as the target window in the rearview image, with a size of (2N+1)×(2N+1)(N=1, 2, 3...), and W2 as the search window of the frontview image, with a size of (2M+1)×(2M+1)(M=1, 2, 3...). Usually, M>N. In the W2 window, traverse the target window W1 in row and column order, and calculate the normalized mutual correlation coefficient ρ(c, r). The row and column coordinates of the pixel point in the frontview image corresponding to the maximum ρ(c, r) are the same-name points in the rearview image based on the rearview image coordinate system.

[0089] In the algorithm, the correlation coefficient is only used to judge similar textures. The same-name point pairs obtained according to the correlation coefficient are the same-name point pairs in the rear-view image coordinate system. It is also necessary to use the relationship transformation model from the rear-view image coordinate system to the front-view image coordinate system to transform the same-name points matched in the rear-view image coordinate system to the front-view image coordinate system as shown in Formula 4 to generate same-name point pairs. The same-name coordinate point pairs are divided into blocks using a polynomial model to eliminate gross errors and obtain correct same-name point pairs.

[0090] (5) Dual-line image matching task distribution based on MPI multi-process

[0091] Based on the MPI multi-process dual-line array image matching parallel processing mechanism, the main process is responsible for decomposing the task into multiple image block set matching tasks and distributing them, and the sub-processes are waiting for the task to be received. After the sub-process receives the task, multiple sub-processes can complete the block image matching task at the same time. After all sub-processes have completed the block matching task and fed back the execution status to the main process, the main process merges the results of each sub-process, and the matching processing task of the dual-line array image block is completed. The parallel processing mechanism of multiple sub-processes improves the efficiency of laser elevation control points in front and back view images, and this advantage is more obvious compared to single-process programs.

[0092] (6) Dual-line image matching results based on laser elevation control point library

[0093] In actual project applications, the multi-process dual-line array image matching technology based on the laser elevation control point library of the present invention is adopted. The 70000*70000 image data is evenly divided into 25 blocks and 25 processes are used to simultaneously perform the laser elevation control point transfer work. The transfer efficiency is about 30 seconds for a single image pair. At the same time, elevation information is introduced into the same-name points used for regional network adjustment, which effectively improves the elevation accuracy of the adjustment.

Claims

1. Multi-process dual-line array image matching method based on laser elevation control point library, It is characterized in that It includes the following two parts: the construction of laser elevation control point library and the dual-line array image matching based on the laser elevation control point library; The laser elevation control point library construction process includes the following steps: ①Create an empty SQLite database; ② The global range is evenly gridded at 5-degree intervals. Specifically, the global (-180 to 180, -90 to 90) range is standardized by 5 degrees by 5 degrees, and each standardized grid block is used as a database table; ③ Establish the database table name with the integer value of the longitude and latitude of the lower left corner of each grid block. Specifically, the table name is named in the form of a combination of the integer value of the longitude and latitude of the lower left corner of the grid block and the east longitude (E), west longitude (W), north latitude (N), and south latitude (S); ④ Read laser height measurement data and use the laser height control point screening model to select high-precision control points; The screening process is as follows: first, a set of parameters in the input data is selected as preliminary screening parameters, and the input laser elevation control points are preliminarily screened according to the screening criteria; for the laser elevation control points that have been preliminarily screened, another set of parameters is selected as quality evaluation parameters, and quality scores are performed according to the quality evaluation criteria. Only data with a quality evaluation score greater than a certain threshold value is finally screened as a high-precision laser elevation control point; ⑤ For the selected high-precision laser elevation control points, insert the longitude and latitude values ​​of each point into the corresponding database table for storage; The dual-line array image matching process based on the laser elevation control point library includes the following steps: ① The main process performs uniform block processing of the rear view image; ②Package each block of information as a subtask; ③The main process pushes all task packages to each sub-process; ④ In each subtask, search the laser elevation control point database according to the block range to determine the laser elevation control point set; ⑤ Extract Forstner feature points within a circle with radius r as the center of each laser elevation control point, and form a set of laser points and feature points; ⑥ Use the RPC parameter file to calculate the image space relationship transformation model of the front and rear view images. The image space relationship transformation model is the image space relationship transformation model for point position conversion from the rear view image coordinate system to the front view image coordinate system; The calculation process of the image-space relationship transformation model is as follows: the latitude and longitude ranges of the four corner points obtained by the rear-view image blocks are intersected with the latitude and longitude ranges of the four corner points of the front-view image to obtain the four corner coordinates (X n , Y n ), n = 0, 1, 2, 3, and the image square coordinates (x′) of the four corner points of the overlapping area corresponding to the front view and the back view are obtained according to the following direct calculation formula n , y′ n ) and (x n ,y n ), n=0, 1, 2, 3; The image-side coordinates can be used to obtain the transformation relationship model from the rear-view image coordinate system to the front-view image coordinate system. The transformation relationship model uses a 6-parameter mode, as shown in the following formula: x n =a 0 +a 1 x n +a 2 and n and n =a 3 +a 4 x n +a 5 and n a n Represents the relational transformation model of 6 parameters, solved according to the least squares fitting formula, n = 0, 1, 2, 3, 4, 5; ⑦ Perform pyramid resolution consistency processing and distortion resampling processing on each point, use the mean pyramid for resolution consistency processing, and use the image square transformation model to use the bilinear interpolation method to correct the distortion; ⑧Perform dynamic threshold correlation coefficient texture matching according to each pyramid layer, and the dynamic threshold is set as follows; Adaptively select pyramid layers with the same or similar resolution to perform layer-by-layer correlation coefficient matching from low resolution to high resolution. In the pyramid matching process, the correlation coefficient matching adopts a dynamic matching threshold mechanism. The matching threshold of the pyramid layer with low resolution is higher than that of the pyramid layer with high resolution. ⑨Use the image-space relation transformation model to transform the point set with the same name. The specific process is as follows: The matching point pairs obtained in the same-name point set are the same-name points in the rear-view image coordinate system. The same-name point pairs in the rear-view image coordinate system are converted to the front-view image coordinate system according to the following formula image-space transformation model to obtain the correct matching point set. x′ n =a 0 +a 1 x n +a 2 and n and' n =a 3 +a 4 x n +a 5 and n ⑩Use the RANSAC method to eliminate gross errors and obtain the correct matching point set; The subprocess task status is fed back to the main process; The main process receives the return status of each subtask, and when all subtasks are completed, merges the files with the same name in each subprocess and outputs the results; The main process ends, the overall task status is returned, and the processing flow ends.

2. The multi-process dual-line array image matching method based on the laser elevation control point library as claimed in claim 1, It is characterized in that In the process of constructing the laser elevation control point library, the threshold in step ④ is set to 75 points.

3. The multi-process dual-line array image matching method based on the laser elevation control point library as claimed in claim 1, It is characterized in that In the dual-line image matching process based on the laser elevation control point library, the matching threshold of the pyramid layer with low resolution in step ⑧ is selected as 0.5, and the matching threshold of the pyramid layer with high resolution is selected as 0.65.

Images