A high-precision 3D reconstruction method driven by real-scene 3D image control network

Abstract

The present invention relates to a high-precision 3D reconstruction method driven by a real-scene 3D image control network, and the specific steps are as follows: in the process of aerial triangulation of control points produced by base real-scene 3D, the encrypted points and their feature information are derived, and an undistorted image with accurate internal and external orientation elements is output; the output undistorted image is matched with the base real-scene 3D to obtain 3D control points; the derived added points and matched 3D control points, as well as the feature descriptions corresponding to these points, are included in unified management, and a unique spatial index is established for each point as a real-scene 3D image control network; the control points of the real-scene 3D image control network are matched to the newly acquired oblique image, and the automatic "transfer" of the control points is completed synchronously, and these successfully matched control points are used as control data to participate in the aerial triangulation of the new image free network. The present invention can realize the rapid and accurate production of real-scene 3D without ground control points and the accurate registration of multi-temporal real-scene 3D.

Description

Technical Field

[0001] The present invention relates to the field of earth information science and technology, and in particular to a high-precision three-dimensional reconstruction method driven by a real-scene three-dimensional image control network. Background Art

[0002] Real-scene 3D is an important measure of the Ministry of Natural Resources to implement the strategies of Digital China, Safe China and Digital Economy, a specific deployment for the implementation of the national new infrastructure construction, and a basic support for serving the construction of ecological civilization and economic and social development. Therefore, the construction of real-scene 3D is not only a transformation opportunity for the natural resources industry involving major spatial information technology, application system and service capacity changes, but also an inevitable requirement for the social and economic development of our country, with great and far-reaching significance. With the diversification, efficiency and convenience of image data collection methods, a large amount of photogrammetric image data generated requires efficient, automatic and intelligent processing. At present, the production of real-scene 3D from oblique images requires field image control point measurement, free network aerial triangulation, artificial puncture points of image control points, aerial triangulation of control points and 3D reconstruction; the accuracy control of real-scene 3D results is determined by the most important image control points. Most of these image control points collect road marking lines, hardened dam edge points and other easily distinguishable objects on the image, which change greatly, have high data update and maintenance costs, and have low reuse opportunities. For areas that have completed the construction of real-scene 3D, if it is necessary to update and produce the real-scene 3D according to the requirements of surveying and mapping laws or application needs, it will be necessary to go through the same arduous steps mentioned above to complete it, resulting in a large amount of human resource investment; and the newly built real-scene 3D may have matching errors because the control information used is different from the old real-scene 3D, which brings inconvenience to subsequent analysis and application. Summary of the invention

[0003] The purpose of the present invention is to solve the shortcomings of the prior art and to propose a high-precision three-dimensional reconstruction method driven by a real-scene three-dimensional image control network.

[0004] In order to achieve the above object, the present invention adopts the following technical solutions:

[0005] Design a high-precision 3D reconstruction method driven by a real-scene 3D image control network. The specific steps are as follows:

[0006] S1: Oblique photography is used to obtain oblique images. Through traditional field photo control point measurement, free network aerial triangulation, photo control point artificial point measurement, control point aerial triangulation, and model reconstruction, the real scene 3D with precise spatial position is obtained as the base real scene 3D;

[0007] S2: Based on the results of the control point aerial triangulation step in the base real scene 3D production process, derive the aerial triangulation encryption points and their feature information, and output the distortion-free image with accurate internal and external orientation elements based on the control point aerial triangulation results;

[0008] S3: Based on the base real scene 3D and distortion-free image, feature point extraction and same-name point matching are performed, and based on the matching results, feature points are read from the base real scene 3D and the 3D coordinates of each feature point are obtained as 3D control points;

[0009] S4: The encrypted points derived from the control point aerial triangulation, the 3D control points obtained from the base real scene 3D, and the feature descriptions corresponding to these points are brought into unified management, and a unique spatial index is established for each point as the real scene 3D image control network;

[0010] S5: Use the real-scene 3D image control network as basic control information, and match them to the newly acquired oblique image according to its recorded features, so as to synchronously complete the automatic "transfer" of control points. The successfully matched control points are used as control data to participate in the aerial triangulation of the new image free network;

[0011] S6: Based on the free-net aerial triangulation results, 3D reconstruction is performed to obtain a new real scene 3D with precise spatial coordinates.

[0012] Preferably, in step S2, the encrypted points derived from the real-scene aerial triangulation are taken as set 1, and the points in set 1 are all obtained by feature matching of the original oblique image, and all have feature descriptions and accurate three-dimensional coordinates.

[0013] Preferably, in step S3, the feature points of the real scene 3D matching with the undistorted image are taken as set 2, and the points in set 2 all have feature descriptions and accurate 3D coordinates.

[0014] Preferably, in step S4, the real-scene three-dimensional image control network includes all points in set 1 and set 2.

[0015] Preferably, in step S4, in the real-scene three-dimensional image control network, for a specific point, its unique spatial index, feature description and its three-dimensional coordinates are recorded.

[0016] Preferably, in step S5, the newly acquired oblique image is matched to the new oblique image according to the feature description of the points in the real-scene 3D image control network; and the points that are successfully matched are automatically "transferred" to the new oblique image.

[0017] Preferably, in step S5, in the free-net aerial triangulation link of the new image, the three-dimensional coordinates of these "turn-thorn" points are read according to their spatial indexes, and they are used as control points to participate in absolute orientation solution.

[0018] Preferably, in step S6, the new real scene 3D has a spatial position with the same accuracy as the base real scene 3D, and the two are accurately matched.

[0019] The invention proposes a high-precision 3D reconstruction method driven by a real-life 3D image control network, and its beneficial effects are as follows: the invention proposes a 3D reconstruction technology based on a real-life 3D image control network, and studies the use of oblique photography aerial triangulation results and existing real-life 3D model results to extract massive high-precision feature points, construct a real-life 3D image control network as control data, and realize rapid and accurate production of real-life 3D without ground control points and accurate registration of multi-temporal real-life 3D. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 This is a flow chart of a high-precision three-dimensional reconstruction method driven by a real-scene three-dimensional image control network proposed by the present invention. DETAILED DESCRIPTION

[0021] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments.

[0022] Reference Figure 1 , a high-precision 3D reconstruction method driven by a real-scene 3D image control network, the specific steps are as follows:

[0023] S1: Oblique photography is used to obtain oblique images. Through traditional field photo control point measurement, free network aerial triangulation, photo control point artificial point measurement, control point aerial triangulation, and model reconstruction processing, a real scene 3D with precise spatial position is obtained as the base real scene 3D.

[0024] S2: Based on the results of the control point aerial triangulation step in the base real scene 3D production process, the aerial triangulation encryption points and feature information are exported, and based on the control point aerial triangulation results, a distortion-free image with precise internal and external orientation elements is output.

[0025] Specifically, the encrypted points derived from the real-scene aerial triangulation are taken as set 1, and the points in set 1 are all obtained from the original image through feature matching, and all have feature information and accurate three-dimensional coordinates.

[0026] S3: Based on the base real scene three-dimensional, distortion-free image, feature point extraction and same-name point matching are performed, and based on the matching results, feature points are read from the base real scene three-dimensional and three-dimensional coordinates are obtained as three-dimensional control points.

[0027] Specifically, the feature points of the real scene 3D and the undistorted image are matched as set 2, and the points in set 2 all have feature information and accurate 3D coordinates.

[0028] S4: The encrypted points derived from the control point aerial triangulation, the 3D control points obtained from the base real scene 3D, and the feature descriptions corresponding to these points are brought into unified management, and a unique spatial index is established for each point as the real scene 3D image control network.

[0029] Specifically, the real-scene 3D image control network includes all the points in set 1 and set 2.

[0030] Specifically, in the real-scene three-dimensional image control network, for a specific point, its unique spatial index, feature description and three-dimensional coordinates are recorded.

[0031] S5: Use the real-life 3D image control network as the basic control information, and match them to the newly acquired oblique image according to their recorded features, so as to synchronously complete the automatic "rotation" of control points. The successfully matched points will be used as control data to participate in the aerial triangulation of the new image free network.

[0032] Specifically, for the newly acquired oblique image, it is matched to the new oblique image according to the feature description of the point in the real-scene 3D image control network; for the successfully matched points, automatic "transfer" is completed in the new oblique image.

[0033] Specifically, in the free-net aerial triangulation link, the three-dimensional coordinates of these "turn-thorn" points are read according to their spatial indexes, and they are used as control points to participate in the absolute orientation solution of aerial triangulation.

[0034] S6: Based on the free-net aerial triangulation results, 3D reconstruction is performed to obtain a new real scene 3D with precise spatial coordinates.

[0035] Specifically, the new real scene 3D has a spatial position with the same accuracy as the base real scene 3D, and the two are accurately matched.

[0036] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A high-precision 3D reconstruction method driven by a real-scene 3D image control network, characterized in that: The specific steps are as follows: S1: Oblique photography is used to obtain oblique images. Through traditional field photo control point measurement, free network aerial triangulation, photo control point artificial point measurement, control point aerial triangulation, and model reconstruction, the real scene 3D with precise spatial position is obtained as the base real scene 3D; S2: Based on the results of the control point aerial triangulation step in the base real scene 3D production process, the aerial triangulation encryption points and their feature information are derived, and based on the control point aerial triangulation results, an undistorted image with accurate internal and external orientation elements is output; S3: Based on the base real scene three-dimensional and distortion-free image, feature point extraction and same-name point matching are performed, and based on the matching results, feature points are read from the base real scene three-dimensional and their three-dimensional coordinates are obtained as three-dimensional control points; S4: The encrypted points derived from the control point aerial triangulation, the 3D control points obtained from the base real scene 3D, and the feature descriptions corresponding to these points are brought into unified management, and a unique spatial index is established for each point as the real scene 3D image control network; S5: Use the real-scene 3D image control network as basic control information, match them to the newly acquired oblique image according to their recorded features, and automatically transfer the control points simultaneously. These successfully matched control points will be used as control data to participate in the aerial triangulation of the new image free network; S6: Based on the free-net aerial triangulation results, 3D reconstruction is performed to obtain a new real scene 3D with precise spatial coordinates.

2. The high-precision 3D reconstruction method driven by a real-scene 3D image control network according to claim 1, characterized in that: In step S2, the encrypted points derived from the aerial triangulation of the control points in the process of making the base real scene three-dimensional are taken as set 1. The points in set 1 are all obtained by feature matching of the original oblique image, and all have feature descriptions and accurate three-dimensional coordinates.

3. The high-precision 3D reconstruction method driven by a real-scene 3D image control network according to claim 2, characterized in that: In step S3, the feature points of the base real scene 3D matching with the undistorted image are taken as set 2, and the points in set 2 all have feature descriptions and accurate 3D coordinates.

4. The high-precision 3D reconstruction method driven by a real-scene 3D image control network according to claim 3, characterized in that: In step S4, the real-scene 3D image control network includes all the points in set 1 and set 2.

5. The high-precision 3D reconstruction method driven by a real-scene 3D image control network according to claim 1, characterized in that: In step S4, in the real-scene 3D image control network, for a specific point, its unique spatial index, feature description and 3D coordinates are recorded.

6. The high-precision 3D reconstruction method driven by a real-scene 3D image control network according to claim 1, characterized in that: In step S5, the newly acquired oblique image is matched to the new oblique image according to the feature description of the points in the real-scene 3D image control network; and the points that are successfully matched are automatically transferred to the new oblique image.

7. The high-precision 3D reconstruction method driven by a real-scene 3D image control network according to claim 1, characterized in that: In step S5, in the free-net aerial triangulation link of the new image, the three-dimensional coordinates of these puncture points are read according to their spatial indexes, and they are used as control points to participate in the absolute orientation solution of the aerial triangulation.

8. The high-precision 3D reconstruction method driven by a real-scene 3D image control network according to claim 1, characterized in that: In step S6, the new real scene 3D has a spatial position with the same accuracy as the base real scene 3D, and the two are accurately matched.

Images