Wide area network cooperative mode large-scale image aerial triangulation method

Abstract

The invention discloses a wide area network cooperative mode large-scale image aerial triangulation method. The method comprises the following steps of: firstly, decomposing a test area into a plurality of sub-areas with certain overlap according to a certain rule, and distributing the sub-areas to a remote computing cluster; then, independent aerial triangulation processing is carried out according to the subareas, and aerial triangulation data is simplified and uploaded to a cloud or a central server to be merged; and finally, carrying out overall block adjustment calculation on the combined measurement areas, and decomposing an adjustment result to the sub-areas. The distributed aerial triangulation method has the advantages that distributed aerial triangulation can be carried out by maximally utilizing a remote computing cluster, the data transmission quantity of a wide area network is reduced, and the operation efficiency and the overall precision consistency of a measurement area are improved.

Description

Technical Field

[0001] The invention belongs to the field of photogrammetry and remote sensing, and specifically relates to a method for large-scale aerial triangulation of images in a wide area network collaborative mode, and more specifically, it is a method for large-scale aerial triangulation of images based on a wide area network collaborative mode. Background Art

[0002] In the field of aerospace photogrammetry, high-precision geometric positioning of optical sensor image data is usually achieved through aerial triangulation (Aerial Triangulation). Aerial triangulation refers to the use of a certain number of ground control points and observation data on the film based on the geometric relationship between the photographed image and the ground objects, and the use of other data to assist in processing, while obtaining the orientation parameters of the film and the three-dimensional coordinates of the points to be measured (also called encrypted points). Aerial triangulation can maintain consistent relative accuracy and absolute positioning accuracy for all image data in the survey area, and is the core link in the geometric processing of aerospace images.

[0003] In response to the continuous increase in the amount of image data, the operation mode of aerial triangulation has gradually developed from single-machine independent processing to multi-machine cluster processing, that is, using a number of computing nodes in a local area network to form a computing cluster for parallel processing, thereby improving processing efficiency. However, with the increasing demand for the production and update of large-scale geographic information products (such as covering a country or even a continent), the tasks of large-scale satellite image aerial triangulation are increasing. The amount of data for such tasks is huge and requires rapid completion within a certain period of time. The existing cluster aerial triangulation operation mode is also difficult to meet the efficiency requirements (the existing cluster aerial triangulation operation mode relies on operators to manually puncture connection points at the junction for adjustment, and also export RPC to puncture some control points in the junction area under stereo conditions for inspection, which is a huge workload and extremely inefficient in large-scale satellite image aerial triangulation operations). In order to accelerate the processing of the cluster aerial triangulation operation mode, a common practice is to decompose the complete survey area into multiple sub-areas, then perform aerial triangulation operations in different areas, and finally connect the regional network through offline survey area junctions. However, the entire process of zoning and edge connection is basically achieved through manual assistance. With the help of external software, data is manually divided. It is applicable to smaller survey areas and has low efficiency and is difficult to automate. Summary of the invention

[0004] The purpose of the present invention is to provide a method for aerial triangulation of large-scale images in a wide area network collaborative mode, which is a method for aerial triangulation of large-scale satellite images based on a wide area network and integrating multiple computing cluster resources distributed in different places. The method performs aerial triangulation in the process of "survey area decomposition - sub-area off-site processing - sub-area merging and overall adjustment - adjustment result distribution", makes full use of off-site computing clusters, quickly completes the aerial triangulation operation of the task survey area, and the partitioning and edge connection are fully automated, applicable to a larger survey area with higher efficiency; solves the problem that the existing aerial triangulation operation mode based on cluster computing cannot effectively utilize off-site multi-cluster computing resources, and solves the problem that the existing technology of partitioning and edge connection is basically realized by manual assistance, and the data is manually divided by external software, which is applicable to a smaller survey area with low efficiency and difficult to automate.

[0005] In order to achieve the above object, the technical solution of the present invention is: a large-scale image aerial triangulation method in a wide area network collaborative mode, characterized in that it includes the following steps:

[0006] Step 1, measurement area decomposition;

[0007] Automatically divide the complete survey area into multiple sub-areas according to a certain grid or task sub-area range;

[0008] Step 2, sub-area aerial triangulation;

[0009] For each sub-area, after applying for the original images and geometric control data, we will independently carry out the aerial triangulation operation according to the aerial triangulation operation specifications to ensure that the results of the sub-area meet the requirements;

[0010] Step 3: Upload and summarize the simplified data of sub-area aerotriangulation;

[0011] Upload and aggregate the simplified aerial triangulation data of each sub-area that can be used for adjustment calculation to the central server or cloud;

[0012] Step 4: Merge the survey areas;

[0013] Merge all the simplified aerial triangulation data of the sub-areas into a complete survey area and automatically connect the survey areas;

[0014] Step 5, joint adjustment;

[0015] Perform joint adjustment on the merged complete survey area to obtain accurate positioning parameters of all images and upload them to the central server or cloud;

[0016] Step 6, updating sub-area image positioning parameters;

[0017] Take the positioning parameters of the sub-area image from the precise positioning parameters of all images uploaded in step 5 and save them.

[0018] In the above technical solution, in step 1, the automatic decomposition of the measurement area includes the following steps:

[0019] First, the geographic range of each image in the survey area is calculated based on the initial positioning parameters of the image or directly extracted from the image metadata information;

[0020] Secondly, the union of the geographic ranges of all images is counted to obtain the geographic range of the entire survey area;

[0021] Then, the entire survey area is automatically decomposed into multiple sub-areas according to certain rules. The automatic division of the survey area includes:

[0022] (1) Automatically divide the survey area into multiple sub-areas according to a certain grid size;

[0023] (2) Automatic division by administrative region: Automatic division by administrative region corresponding to the task areas of multiple remote units involved in the operation;

[0024] (3) Automatically divide by clustering; cluster the point set formed by the center points of the stereo image pair composed of multiple images, and the number of clusters is set according to the actual situation; after clustering, the convex hull of each cluster is calculated;

[0025] (4) Automatically divide by geographical attributes;

[0026] You can select one of the above automatic measurement area division methods according to your needs, or select several of them and combine them.

[0027] In the above technical solution, in order to ensure that subsequent sub-areas can be merged, adjacent sub-areas should ensure that 1 to 2 image pairs overlap along their boundary lines, or more image pairs can be overlapped along the boundary lines according to actual conditions.

[0028] In the above technical solution, in step 2, the operating unit of each sub-area applies for the original images and control data within the sub-area, and independently performs aerial triangulation operations according to relevant operating specifications. The operating unit of each sub-area only needs to apply for all the original image data and control data within the sub-area, and does not need to apply for the original image data and control data of other sub-areas.

[0029] In the above technical solution, in step 3, after completing the aerial triangulation task in the sub-area, the aerial triangulation results need to be uploaded to the central server or cloud through the wide area network; during the transmission process, each sub-area only needs to upload simplified aerial triangulation data for adjustment and solution, with a small amount of data and fast transmission; among them, the simplified aerial triangulation data includes control point files, sub-area image point files, aerial triangulation project files containing image lists and stereo model list information, and coordinate system parameter files, but does not include original image files.

[0030] In the above technical solution, in step 4, it is necessary to merge and summarize the sub-areas. By merging the uploaded simplified aerial triangulation data of all sub-areas, the aerial triangulation results of each sub-area are summarized into a complete survey area; sub-area merging mainly includes project merging and connection point merging:

[0031] (1) Project merging, including merging of image lists and merging of stereo models, merging the project files of multiple sub-areas into a complete project file of the survey area, and deleting duplicate images and stereo models during merging;

[0032] (2) Merging of connection points. Since the connection point numbers of different sub-areas may have duplicate names, in order to ensure the uniqueness of the connection point numbers after the survey area is merged, the connection points need to be deduplicated and numbered uniformly. The specific method is to add a large number before the sub-area connection point number. The large number is the sub-area number, and the digit of the added large number is the digit before the maximum digit of all connection point numbers, so as to retain all the connection points in different sub-areas of the repeated image. Through the connection points of different survey areas existing in the repeated image, the connection relationship between different survey areas is established, so as to realize the automatic edge connection of the survey area without the need for re-matching or manual transfer of connection points.

[0033] In the above technical solution, in step 5, the regional network adjustment solution is performed on the merged complete survey area, and the precise positioning parameters of the image that meets the accuracy requirements must be obtained in accordance with the operating specifications, and uploaded to the central server or cloud for download by each sub-area.

[0034] It should be noted that the present invention is highly practical and is applicable to aerial triangulation of various types of images (such as satellite images, aerial images, and drone images). The positioning parameters of different types of images are in different forms. Satellite images are generally RPC parameters, while aerial and drone images are internal and external orientation elements. However, this method can be applied to different types of positioning parameters, which can be uploaded to the central server as a unified result, and each sub-area can be downloaded separately, and compatibility can be achieved for various forms of positioning parameters; it overcomes the problem that the existing technology has a small scope of application and is only applicable to a certain type of image (aerial image or drone image), and cannot be applied to aerial triangulation of satellite images.

[0035] In the above technical solution, in step 6, each sub-area obtains the precise positioning parameters after the overall solution from the cloud or central server, updates the sub-area image positioning parameters, and obtains the sub-area aerial triangulation results. Each sub-area uses the image name contained in its own measurement area as a unique identifier, takes the positioning parameters of the sub-area image from the precise positioning parameters of all images uploaded in step 5, and saves and updates them. When downloading, each sub-area only needs to download the positioning parameters corresponding to the images contained in its own measurement area.

[0036] The large-scale images in the wide area network collaborative mode large-scale image aerial triangulation method described in the present invention refer to satellite images at the continental level. The present invention adopts a method of automatically dividing the survey area and the edges, and each sub-area is aerially triangulated, and the simplified aerial triangulation parameters are returned through the wide area network to realize the reuse of computing resources in various places, and realize large-scale (such as covering a country or even a continent) satellite image aerial triangulation, which is applicable to large data volumes and high measurement efficiency; it overcomes the existing technology of using artificial puncture point methods or partitions to perform aerial triangulation operations, and the method that the partition and edge connection are basically realized by manual assistance is only applicable to the aerial triangulation of aerial images in survey areas with small data volumes (such as a province or several cities), and cannot decompose and merge large survey areas, cannot effectively utilize multi-cluster computing resources in different locations, cannot realize large-scale satellite image aerial triangulation with large data volumes, and has low measurement efficiency.

[0037] The present invention has the following advantages:

[0038] (1) Fully utilizing the computing resources of multiple clusters in different locations solves the problem that it is difficult to complete tasks on time with computing resources in only one location when tasks need to be processed quickly, thereby improving computing efficiency and emergency production capabilities. It overcomes the problems that the existing technology can process small data scale (mainly performing aerial triangulation processing of partitions in one area), cannot utilize computing resources of multiple clusters in different locations (the existing technology relies on raw image data and needs to transmit image data, which has a large amount of image data, slow transmission speed, and high cost), has low computing efficiency, and cannot complete emergency situations.

[0039] (2) The partitioning production method is adopted to reduce the requirements for the computing performance of a single cluster, solve the problem of insufficient computing resources leading to difficulty in processing large amounts of data, and improve the ability and efficiency of processing large amounts of data; the automatic partitioning method of the present invention is not only applicable to the case of a small survey area (such as covering a certain area in an urban area), but also applicable to a large survey area (such as covering a province / country or even a continent), with a large scope of application and high efficiency; it overcomes the problem that the prior art is only applicable to the case of a small survey area (such as covering a certain area in an urban area), and the data is manually divided by using external software, which has a small scope of application and low efficiency;

[0040] (3) Make full use of the fact that the adjustment solution does not rely on the original image data to ensure that the amount of data uploaded and downloaded from the cloud is small and the transmission is fast, thus avoiding the time waste of transmitting a large amount of image data;

[0041] (4) The present invention adopts a parallel aerial triangulation method of partition data processing for continental satellite image data processing, which can make full use of the computing resources of multiple clusters in different places, and at the same time utilize the characteristic that the adjustment solution does not rely on the original image data. In the transmission process of uploading the aerial triangulation result to the central server or the cloud through the wide area network, each sub-survey area only needs to upload the simplified aerial triangulation data (including control point files, sub-area image point files, aerial triangulation engineering files containing image lists and stereo model list information, and coordinate system parameter files) for adjustment solution, ensuring that the data volume of the cloud upload and download of data is small and the transmission is fast; at the same time, the present invention realizes large-scale data processing to make the integrity of the adjustment result better and reduce the systematic error; overcomes the problem that the existing technology image aerial triangulation needs to rely on the original image data and control data of the survey area and the large data volume, and at the same time overcomes the problem that the existing technology cannot realize large-scale satellite image aerial triangulation, and can only separate different regions one by one and then perform aerial triangulation measurement, cannot realize the overall processing of large-scale data, and has large systematic errors (such as, separate the aerial triangulation measurement processing of each province, which may cause the existence of corresponding systematic errors in different provinces, thereby causing the systematic error of large-scale data processing to increase);

[0042] (5) The present invention is generally applicable to the overall process of satellite aerial triangulation, has high measurement efficiency, and processes a large amount of data; it overcomes the problem that the prior art is only applicable to the measurement of control data when there is control data in the partitioned aerial triangulation, has low measurement efficiency, and processes a small amount of data;

[0043] (6) The present invention can realize uniform automatic partitioning by setting a clustering method with parameters (such as the number of clusters), or automatically partitioning by the scope of administrative districts, or automatically partitioning by a regular grid of longitude and latitude, or automatically partitioning by geographical properties when auxiliary data is added; these partitioning methods do not require direct human participation, only auxiliary work such as manually adjusting options and setting some parameters is required, and there is no need to mobilize other software; the present invention automatically completes the edge connection during the merging and adjustment of the areas through a program algorithm, and the partitioning and edge connection are efficient and accurate; it overcomes the problem that the partitioning and edge connection of the prior art are basically all dependent on manual implementation, it is difficult to realize automation, and the partitioning and edge connection efficiency is low and the accuracy is not high. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 It is the overall technical flow chart of the present invention;

[0045] Figure 2 This is a schematic diagram of the overall image distribution of the survey area according to an embodiment of the present invention;

[0046] Figure 3 This is a schematic diagram of the results of the K-means clustering method for dividing the measurement area of ​​the present invention;

[0047] Figure 4This is a schematic diagram of overlapping images along the dividing line when dividing the survey area in the present invention;

[0048] Figure 5 It is the image distribution map of each sub-area of ​​the present invention;

[0049] Figure 6 It is the distribution map of control points and check points in each sub-area of ​​the present invention;

[0050] Figure 7 It is the connection point distribution diagram of each sub-area of ​​the present invention;

[0051] Figure 8 The present invention simplifies the composition diagram of the aerotriangulation data;

[0052] Fig. 9 This is a schematic diagram of the unified numbering method for connection points of the present invention. DETAILED DESCRIPTION

[0053] The following is a detailed description of the implementation of the present invention in conjunction with the accompanying drawings, but they do not constitute a limitation of the present invention and are only given as examples. At the same time, the advantages of the present invention are made clearer and easier to understand through the description.

[0054] The present invention discloses a large-scale image aerial triangulation method based on a wide area network collaborative mode. The core idea is to decompose the large-scale image aerial triangulation task into subtasks, and distribute them to multiple computing clusters located in different places through a wide area network, so as to improve processing efficiency. First, the survey area is automatically decomposed into multiple sub-areas with a certain overlap according to certain rules, and they are assigned to remote computing clusters; then, independent aerial triangulation processing is performed on each sub-area, and simplified aerial triangulation data is uploaded to the cloud or a central server for merging; finally, the merged survey area is subjected to overall regional network adjustment and solution, and the adjustment results are decomposed into sub-areas. This method is no longer limited to a single computer or a single computing cluster, and can maximize the use of remote computing clusters for distributed aerial triangulation, solving the problem that it is difficult to complete tasks on time with computing resources in one place when tasks need to be processed quickly, thereby improving computing efficiency and emergency production capabilities. In addition, the present invention also solves the problems of insufficient consistency and low transmission efficiency among the aerial triangulation results of each data processing node, proposes a collaborative aerial triangulation method of off-site processing of sub-tasks - merging of sub-tasks and adjustment solution - distribution of adjustment results, and the concept of simplifying aerial triangulation data, thereby reducing the amount of wide area network data transmission, and improving computing efficiency and the overall accuracy consistency of the survey area.

[0055] The present invention is mainly realized through the following technical solutions: Based on the idea of ​​wide area network collaborative mode, the aerial triangulation process is decomposed from the original data stage, and multiple clusters located in different places are used to give full play to their respective computing resources to complete the aerial triangulation tasks of each sub-area, ensuring that the results of each sub-area meet the requirements, and then the simplified aerial triangulation data of each sub-area is uploaded to the central server or cloud, and the sub-area merging and joint adjustment are completed to ensure the overall consistency of the aerial triangulation results. The overall technical process is as follows Figure 1 shown.

[0056] The core process includes the following steps:

[0057] Step 1: Decompose the survey area. Automatically divide the complete survey area into multiple sub-areas according to a certain grid or task sub-area range;

[0058] Step 2: Sub-area aerial triangulation. For each sub-area, after applying for the original images and control data, carry out the aerial triangulation operation independently according to the aerial triangulation operation specifications to ensure that the results of the sub-area meet the requirements;

[0059] Step 3: Upload and summarize the simplified aerial triangulation data of each sub-area. Upload and summarize the simplified aerial triangulation data of each sub-area that can be used for adjustment calculation to the central server or cloud;

[0060] Step 4: Merge the survey area. Merge all the simplified aerotriangulation data of the sub-areas into a complete survey area and automatically connect the survey areas.

[0061] Step 5: Overall adjustment: Perform joint adjustment on the merged complete survey area to obtain the precise positioning parameters of all images and upload them to the central server or cloud.

[0062] Step 6, update the sub-area image positioning parameters. Obtain the positioning parameters of the sub-area image from the precise positioning parameters of all images uploaded in step 5 and save them. The present invention first automatically partitions the image and merges the sub-area aerial triangulation without relying on the original data. By realizing the automatic partitioning of the original image and using the corresponding algorithm, the sub-area aerial triangulation data is extracted and the merged area aerial triangulation is calculated. At the same time, it is dispersed to various places to use remote resources for collaborative processing. It is suitable for satellite images with large data volume and coverage (capable of obtaining a large amount of data at one time) and suitable for global or large-scale real-time monitoring and analysis (such as climate change research, global forest cover monitoring, marine protection, etc.), and improves processing efficiency; it solves the problem that the prior art is only suitable for aerial triangulation of aerial images with small data volume and coverage, suitable for scenes requiring high precision and local detailed information (such as urban three-dimensional modeling, building deformation monitoring, and archaeological site identification), and cannot realize aerial triangulation such as the division, calculation and transmission of large-scale satellite image data, and cannot realize aerial triangulation of large-scale satellite images with too large data volume (for example, satellite image data with a data level of TB).

[0063] In step 1, the steps of survey area decomposition are as follows:

[0064] First, the geographic range of each image in the survey area is calculated based on the initial positioning parameters of the image or directly extracted from the image metadata information;

[0065] Secondly, we count the union of all the image geographic ranges and get Figure 2 The geographical extent of the entire survey area is shown;

[0066] Then, the entire survey area is divided into multiple sub-areas according to certain rules. The division methods include:

[0067] (1) Divide by regular grid. Divide the survey area into multiple sub-areas according to a certain grid size;

[0068] (2) Division by administrative region: Division is based on the administrative regions corresponding to the mission areas of multiple remote units involved in the operation;

[0069] (3) Division by clustering. Cluster the point set formed by the center points of the stereo image pair composed of multiple images (such as K-means clustering). The number of clusters can be set according to the actual situation. After clustering, the convex hull (i.e., the minimum circumscribed convex polygon) of each cluster is calculated. The division effect is as follows: Figure 3 As shown, points of different colors represent different clusters, and the points enclosed by each black dotted frame are the image pairs contained in the sub-area;

[0070] (4) Division by geographical attributes: For example, division by geographical attributes such as high altitude and low altitude areas, urban areas and forest areas.

[0071] The above division methods can be selected one by one or a combination of several according to needs.

[0072] Finally, in order to ensure that the subsequent sub-areas can be merged, adjacent sub-areas should ensure that 1 to 2 image pairs or images overlap along their boundary lines. Figure 4 As shown in the figure, it is a schematic diagram of the rectangular grid division results and the overlapping of sub-areas, where the small quadrilateral vector is the geographical range of the image pair, the blue is the overlapping image pair, the black is the non-overlapping image, and the red vector is the divided sub-area range.

[0073] In step 2, after assigning the corresponding work unit to each sub-area, apply for the original image and geometric control data within the sub-area, perform the aerial triangulation work independently according to the relevant work specifications, and ensure that the aerial triangulation results meet the accuracy requirements. In addition, each sub-area should measure the control points within the sub-area according to the work specifications. The image distribution, control point and check point distribution, and connection point distribution of each sub-area are as follows: Figure 5 , Figure 6 and Figure 7 As shown (above Figure 1 The entire process is shown; Figure 2 The image distribution of the entire survey area is shown; Figure 3 and Figure 4 It embodies two automatic division methods of survey area: division by clustering and division by longitude and latitude; Figure 5 The image distribution of the sub-survey area is shown; Figure 6 , Figure 7 It is shown that the connection points and control points of each sub-survey area in the present invention are evenly distributed, indicating that the sub-survey area aerial triangulation effect of the present invention is good and the accuracy is high; Figure 2 , Figure 5 , Figure 6 , Figure 7 The blue edges in the figure all represent the boundary lines between a province and surrounding provinces in the embodiment of the present invention; Figure 2 , Figure 4 , Figure 5 , Figure 6 The red, blue, and black small quadrilaterals in the figure represent the geographical range of the image pair; Figure 6 The triangles in the figure represent the checkpoint distribution, and the circles represent the control point distribution; Figure 7 The small quadrilaterals in the figure represent the distribution of connection points. Red represents one-degree overlap, rose-red represents two-degree overlap, and green represents three-degree overlap or above).

[0074] In step 3, after completing the sub-area triangulation task, the sub-area triangulation results need to be uploaded to the central server or cloud through the wide area network for subsequent aggregation and merging. In order to improve transmission efficiency, the amount of data should be reduced as much as possible. Therefore, in the actual transmission process, each sub-area only needs to upload simplified triangulation data for adjustment and solution, avoiding the transmission of huge data such as original image data. Figure 8 As shown, the simplified aerial triangulation data includes a control point file, a sub-area image point file, an aerial triangulation engineering file containing image list and stereo model list information, and a coordinate system parameter file. The wide area network of the present invention refers to a computer network covering a wide geographical area, which can connect local area networks, computers, terminals and other devices distributed in different locations to achieve resource sharing and communication. The present invention uploads the sub-area aerial triangulation results to a central server or cloud through a wide area network to achieve the extraction of sub-area aerial triangulation data and the calculation of combined area aerial triangulation, and realizes large-scale satellite image aerial triangulation. At the same time, it is dispersed to various places to use remote resources for collaborative processing and make full use of remote computing clusters to improve processing efficiency; it overcomes the problem that the existing technology realizes the connection of regional networks by connecting the edges of offline survey areas (the regional network of photogrammetry refers to a specific area divided from multiple flight strips or image areas with overlapping images in order to perform overall adjustment and improve measurement accuracy during the photogrammetry process), and the application range is small and the efficiency is low.

[0075] This step only transmits the necessary data for survey area merging and joint adjustment, that is, simplified aerial triangulation data, including image list and positioning parameter information, stereo model list, image point observation data, and control point coordinates; it does not include the original image file, so it can greatly reduce data and improve transmission speed, making the partitioned aerial triangulation results suitable for transmission in a wide area; the transmission of partitioned aerial triangulation results in a wide area can realize the sharing of aerial triangulation results, and it is also convenient for subsequent merging operations (first transmit the simplified aerial triangulation data together and then perform the merging operation), while realizing large-scale satellite image aerial triangulation, reducing the cost of transmitting data in processing large-scale satellite image data, and improving data Transmission efficiency and cost savings; it solves the problem that the existing technology relies on original image data and control information, and the data volume of aerial triangulation results is too large to be transmitted in a wide area; at the same time, it overcomes the problem that the aerial triangulation processing flow of the existing technology is only applicable to aerial images with small data volume and coverage, suitable for scenes requiring high precision and local detailed information (such as urban 3D modeling, building deformation monitoring, and archaeological site identification), but cannot be applied to satellite images with large data volume and coverage (capable of obtaining a large amount of data at one time) and suitable for global or large-scale real-time monitoring and analysis (such as climate change research, global forest cover monitoring, marine protection, etc.).

[0076] In step 4, merge and summarize the sub-areas. By merging the uploaded simplified aerial triangulation data of all sub-areas, the aerial triangulation results of each sub-area are summarized into a complete survey area. Sub-area merging includes:

[0077] (1) Project merging. This includes merging image lists and stereo pairs, merging the project files of multiple sub-areas into a complete survey area project file. Since each adjacent sub-area has a certain overlap range, it is necessary to delete duplicate images and stereo pairs when merging.

[0078] (2) Merge and unify the connection points. The connection points on the overlapping images of the sub-areas need to be merged together. Since the connection point numbers of different sub-areas may have duplicate names, in order to ensure the uniqueness of the connection point numbers after the survey area is merged, the connection points need to be deduplicated and unified. The specific method is to add a large number before the sub-area connection point number. The large number is the sub-area number. The digit of the added large number is the digit before the maximum digit of all the connection point numbers, ensuring that the connection point numbers of different sub-areas are in different, non-overlapping intervals. Fig. 9 As shown, the new number of the connection point is composed of the sub-area number and the partition connection point number. For example, the connection point 1000345 of sub-area 1 is named 11000345 after merging, and the connection point 1000345 of sub-area 2 is named 21000345 after merging, which ensures the uniqueness of the name and can identify which sub-area the connection point comes from; the connection points are deduplicated and uniformly numbered to retain all the connection points in different sub-areas of the repeated images, and the connection relationship between different survey areas is established through the connection points of different survey areas on the repeated images, so as to realize the automatic edge connection of the survey area. Since the automatic division of the survey area in step 1 repeatedly divides some images into different sub-areas (several images overlap when the survey area is automatically divided in step 1), the connection points are merged and uniformly numbered when the survey areas are merged to avoid the name conflict, so as to retain all the connection points in different sub-areas of the repeated images, so as to realize the automatic edge connection of the survey area without matching again or manually transferring the connection points. For example, image a is repeatedly assigned to survey area 1 and survey area 2, and there will be connection points when aerial triangulation is performed in the survey areas; image a may have connection points such as 3003, 3004, 3005 in survey area 1, and may have connection points such as 6111, 6112, 6113 in survey area 2. When the sub-areas are merged, they are renumbered and all the connection points of image a in survey area 1 and survey area 2 are retained. Therefore, after the merged area, there will be connection points such as 13003, 13004, 13005, 26111, 26112, 26113. Since there are points of two survey areas (survey area 1 and survey area 2) on one image a, this image a can establish the connection relationship between the two survey areas (survey area 1 and survey area 2), thereby realizing automatic edge connection.

[0079] In step 5, the combined complete survey area is adjusted and solved as a whole to obtain the precise positioning parameters of all images, which are then uploaded to the central server or cloud for download by each sub-area. When performing joint adjustment and solving of the complete survey area, it is also necessary to meet the operating specifications to obtain the precise positioning parameters of the images that meet the accuracy requirements.

[0080] It should be noted that the positioning parameters of different types of images are in different forms. Satellite images are generally RPC parameters, while aerial and drone images are internal and external orientation elements. However, this method can be applied to different types of positioning parameters, which are uniformly uploaded to the central server as results, and each sub-area can be downloaded separately, and it can be compatible with various positioning parameter forms.

[0081] In step 6, each sub-area obtains the precise positioning parameters after the overall solution from the cloud or central server, and updates the sub-area image positioning parameters. Each sub-area uses the image name contained as a unique identifier, and obtains the positioning parameters of the sub-area image from the precise positioning parameters of all images uploaded in step 5, and saves and updates them.

[0082] Example

[0083] The present invention is now described in detail by taking the application of the present invention to a certain province in the aerial triangulation of satellite images as an example, which also has a guiding role in the application of the present invention to other large-scale aerial triangulation of images.

[0084] This example selects 138 stereoscopic scenes of a province's Resource III satellite image data (each stereoscopic scene contains three images: front view, front view and bottom view). The survey area is about 160,000 square kilometers, with a total of 404 images.

[0085] The technical solution provided by the present invention is: first, sub-area decomposition is performed, and then sub-area aerial triangulation is performed. After the aerial triangulation solution of each sub-area is completed, the aerial triangulation data is uploaded, and then the sub-area projects are merged and the adjustment solution of the overall project is performed in the cloud to generate adjustment results, and then each sub-area imports the corresponding updated adjustment results to complete the update of the aerial triangulation results.

[0086] The process of collaborative aerotriangulation is as follows: Figure 1 As shown, the following steps are included:

[0087] Step 1: First, calculate the geographic range of each image in the survey area based on the initial positioning parameters, and merge them to obtain the survey area range, such as Figure 2 The original data is divided into four measurement areas and the data volume of each sub-area is ensured to be similar. At the same time, there are 1 to 2 overlapping stereoscopic scenes at the edges between the sub-areas.

[0088] Since the stereoscopic scene data of target survey area 138 is distributed in the east-west direction, according to the actual situation, due to the irregular distribution of the survey area, K-means clustering method is used to divide it, so as to ensure that the image data volume is evenly distributed as much as possible and there are 1 to 2 overlapping images at the edges between sub-areas. It is actually divided into 4 sub-areas, including 44, 45, 34 and 41 stereoscopic scenes respectively (this embodiment adopts the method of the present invention, and the partitioning and edge connection of the survey area are automatically realized by the set program without manual intervention; the program for automatic partitioning of the survey area in this embodiment is: automatic partitioning program by clustering method; the automatic edge connection program of the survey area is: when the survey areas are merged, the connection points are merged and numbered uniformly to avoid point name conflicts, so as to retain all connection points of repeated images in different sub-areas).

[0089] Step 2: Perform independent triangulation based on the sub-areas divided in step 1 to ensure that the results of the sub-areas meet the requirements (such as Figure 3 As shown, Figure 3 This is the effect of automatic partitioning using the clustering method of the present invention in this embodiment).

[0090] For each sub-area, first match the connection points to ensure that the connection points are evenly distributed, and then measure the control points. The control points are distributed as follows: Figure 6 As shown in the figure, the final adjustment is performed. During the adjustment process, the image point residual error limit is 2 pixels, and the plane and elevation error limits of the control points are 12 meters and 10 meters respectively. The control points of the four sub-areas are distributed as follows: Figure 6 As shown in Figure 2, the number of control points are 283, 225, 168 and 199 respectively; the connection points are distributed as follows: Figure 7 As shown, the quantities are 374918, 369589, 263662 and 245881 respectively.

[0091] Step 3, upload the simplified aerial triangulation data of each sub-area, including image point files, engineering files, control point files and other data to the cloud. The sizes of the aerial triangulation simplified data files of the four sub-areas are 30.3MB, 34.0MB, 21.4MB and 23.7MB respectively (the data volume of the aerial triangulation simplified data files in this embodiment is small and the transmission is fast; compared with other existing methods, the image data transmitted by the method of the present invention in this embodiment is many orders of magnitude smaller, and the image data transmitted by other existing methods is generally in the order of GB or above).

[0092] Step 4: Merge the aerial triangulation results of each sub-area in the cloud, remove duplicate images and image points, summarize the aerial triangulation results of the four sub-areas into a complete survey area, and automatically connect the survey areas. The number of connection points and control points in the merged survey area is 1254050 and 875 respectively. The project merging includes the merging of image lists and stereo models. The project files of the four sub-areas are merged into a complete project file of the survey area, and duplicate images and stereo models are deleted during the merging. When merging connection points, the connection points of different sub-areas are uniformly renumbered.

[0093] Step 5: Perform regional block adjustment on the merged complete survey area. The mean error of the adjusted image points is 0.34 pixels, the mean error of the control point plane and elevation is 2.84 meters and 1.59 meters, and the mean error of the check point plane and elevation is 2.74 meters and 1.90 meters; the size of the generated exterior orientation element file is 76KB, which is uploaded to the central server or cloud for download by each sub-area.

[0094] Step 6, updates geometric positioning parameters according to sub-area distribution. Each sub-area processing unit opens the engineering of the sub-area, imports the adjustment results, and automatically downloads the newly generated geometric positioning parameter file from the central server or the cloud in sequence according to the image name of the sub-area, updates and saves (when the present invention is applied to other large-scale images (such as countries or continents) satellite image aerial triangulation, the method of the present invention can be used to automatically partition into multiple provinces or multiple countries, and when ensuring the edge connection, each province or country (i.e. each sub-survey area) processes the data of its own province or country (i.e., performs sub-area aerial triangulation), and utilizes the computing resources (i.e., the computing resources of each province) to complete the measurement task on time, improves the computing efficiency and the ability of emergency production), then merges each province or country (i.e., sub-survey area) and performs overall adjustment, and finally distributes the adjustment results; when the present invention is applied to other types of image (such as satellite image, aerial image, drone image) aerial triangulation, the measuring method is the same).

[0095] Conclusion: This embodiment adopts the method of the present invention to carry out aerial triangulation of satellite images of a certain province, makes full use of off-site computing clusters, quickly completes the aerial triangulation operations of the mission survey area, and can automatically perform partitioning and edge joining with high efficiency and accuracy. The amount of data uploaded and downloaded from the cloud is small, and the transmission speed is fast.

[0096] The large-scale image aerial triangulation method based on the wide area network collaborative mode proposed in the present invention aims to integrate multiple computing clusters distributed in different places through the wide area network, decompose the aerial triangulation of the overall survey area into multiple sub-areas through partitioning, and perform independent aerial triangulation operations in different places. After the operation is completed, the simplified aerial triangulation data of the sub-area (excluding image data, only including image point observations, control points, image lists and positioning parameter information) is submitted through the wide area network, and then the overall regional network adjustment solution is summarized, so as to maximize the realization of the task requirements of rapid processing of large-scale satellite images; solve the problem that the existing technology of partitioning and edge connection is basically realized by manual assistance, which is difficult to automate, and the overall accuracy of aerial triangulation is also affected to a certain extent.

[0097] The present invention mainly solves the problem that the traditional cluster parallel processing method cannot effectively utilize the off-site computing cluster resources, and proposes a large-scale aerial triangulation method with a task flow of "survey area decomposition - off-site processing of sub-areas - sub-area merging and overall adjustment - distribution of adjustment results". The original data is divided at the task level, and each off-site sub-area only needs to process the data of the local survey area of ​​the sub-area. Finally, only the simplified aerial triangulation data without the original image file is uploaded to the cloud. The data volume is small, and fast upload can be achieved to meet the needs of emergency production.

[0098] The specific embodiments described herein are merely examples of the spirit of the present invention. Those skilled in the art may make various modifications or additions to the specific embodiments described or replace them in similar ways, but they will not deviate from the spirit of the present invention or exceed the scope defined by the appended claims.

[0099] Other parts not described belong to the prior art.

Claims

1. A large-scale image aerial triangulation method in a wide area network collaborative mode, characterized by: The following steps are involved: Step 1, measurement area decomposition; Automatically divide the complete survey area into multiple sub-areas according to a certain grid or task sub-area range; Step 2, sub-area aerial triangulation; For each sub-area, after applying for the original images and geometric control data, we will independently carry out the aerial triangulation operation according to the aerial triangulation operation specifications to ensure that the results of the sub-area meet the requirements; Step 3: Upload and summarize the simplified data of sub-area aerotriangulation; Upload and aggregate the simplified aerial triangulation data of each sub-area that can be used for adjustment calculation to the central server or cloud; Step 4: Merge the survey areas; Merge all the simplified aerial triangulation data of the sub-areas into a complete survey area and automatically connect the survey areas; Step 5, joint adjustment; Perform joint adjustment on the merged complete survey area to obtain accurate positioning parameters of all images and upload them to the central server or cloud; Step 6, updating sub-area image positioning parameters; Take the positioning parameters of the sub-area image from the precise positioning parameters of all images uploaded in step 5 and save them.

2. The wide area network collaborative mode large-scale image aerial triangulation method according to claim 1, characterized in that: In step 1, the automatic decomposition of the survey area includes the following steps: First, the geographic range of each image in the survey area is calculated based on the initial positioning parameters of the image or directly extracted from the image metadata information; Secondly, the union of the geographic ranges of all images is counted to obtain the geographic range of the entire survey area; Then, the entire survey area is automatically decomposed into multiple sub-areas according to certain rules. The automatic division of the survey area includes: (1) Automatically divide the survey area into multiple sub-areas according to a certain grid size; (2) Automatic division by administrative region: Automatic division by administrative region corresponding to the task areas of multiple remote units involved in the operation; (3) Automatically divide by clustering; cluster the point set formed by the center points of the stereo image pair composed of multiple images, and the number of clusters is set according to the actual situation; after clustering, the convex hull of each cluster is calculated; (4) Automatically divide by geographical attributes; You can select one of the above automatic measurement area division methods according to your needs, or select several of them and combine them.

3. The method for large-scale aerial triangulation of images in a wide area network collaborative mode according to claim 1 or 2, characterized in that: Overlap image pairs along the boundary lines of adjacent sub-areas.

4. The method for large-scale aerial triangulation of images in a wide area network collaborative mode according to claim 3, characterized in that: In step 2, the operating units of each sub-area apply for the original images and control data within the divided sub-area, and independently carry out aerial triangulation operations in accordance with relevant operating specifications.

5. The method for large-scale aerial triangulation of images in a wide area network collaborative mode according to claim 4, characterized in that: In step 3, after completing the aerial triangulation task in the sub-area, the aerial triangulation results are uploaded to the central server or cloud through the wide area network; during the transmission process, each sub-area uploads simplified aerial triangulation data for adjustment solution; among them, the simplified aerial triangulation data includes control point files, sub-area image point files, aerial triangulation project files containing image list and stereo model list information, and coordinate system parameter files.

6. The wide area network collaborative mode large-scale image aerial triangulation method according to claim 5, characterized in that: In step 4, the aerial triangulation results of each sub-area are summarized into a complete survey area by merging the uploaded aerial triangulation simplified data of all sub-areas; Sub-area mergers include: (1) Project merging, including merging of image lists and merging of stereo models, merging the project files of multiple sub-areas into a complete project file of the survey area, and deleting duplicate images and stereo models during merging; (2) Connecting point merging: Add a large number before the sub-area connecting point number. The large number is the sub-area number. The digit of the large number added is the digit before the maximum digit of all connecting point numbers. This is to retain all connecting points in different sub-areas of the repeated image. Through the connecting points of different survey areas existing in the repeated image, the connection relationship between different survey areas is established, so as to realize the automatic edge connection of the survey areas without the need for re-matching or manual transfer of connecting points.

Images