Local-to-global deep space orbiter image stereo image pair optimization method

Abstract

The invention relates to a local-to-global deep space orbiter image stereo image pair optimization method, which comprises the following steps of: acquiring deep space orbiter images, performing image preprocessing, determining an initial stereo image pair by judging an overlapping area between the images, and performing image pair preliminary screening to obtain a preliminarily screened stereo image pair set; obtaining information of a region of interest, carrying out rasterization and gridding processing on the region of interest, calculating the local quality cost of each stereo image pair for each grid, and obtaining a local optimal image pair and a local alternative image pair set corresponding to the current grid based on the local quality cost; calculating the global quality cost of each stereo image pair based on the local optimal image pair and the local alternative image pair set corresponding to each grid; and combining the local quality cost and the global quality cost of the stereo image pair to obtain a final optimized stereo image pair. Compared with the prior art, the method can automatically and effectively select the stereo image pair meeting the requirement of photogrammetry mapping.

Description

Technical Field

[0001] The present invention relates to the field of planetary mapping technology, and in particular to a method for optimizing local to global deep space orbiter image stereo pairs. Background Art

[0002] Planetary mapping is a basic task necessary for planetary scientific research and exploration engineering missions. It is necessary to find data that meets the mapping conditions from a large amount of redundant orbiter image data. Due to the large amount of orbiter image data and the varying quality, manual screening is time-consuming and laborious. Therefore, constructing an automated optimization method for deep space orbiter stereo image pairs plays a vital role in efficient and accurate planetary mapping.

[0003] Some studies have proposed methods for selecting the best optical satellite images of the Earth. However, these methods are not completely applicable to deep space orbiter images. There are three main reasons for this. First, the stereo pairs of high-resolution images of deep space orbiters are mostly constructed through images from different orbits, so the screening requirements for stereo pairs are more stringent; second, the overlapping parts of stereo pairs are mostly irregular polygons, and it is more difficult to find stereo pairs with low redundancy and complete coverage of the area of ​​interest; finally, the lighting conditions have a greater impact on deep space orbiter images, and the lighting differences between the two images are not considered when screening the pairs.

[0004] In summary, there is currently a lack of a method for optimizing stereo pairs of deep space orbiter images for scenarios such as planetary mapping. Summary of the invention

[0005] The purpose of the present invention is to overcome the defects of the above-mentioned prior art and to provide a local to global deep space orbiter image stereo pair optimization method, so as to solve or partially solve the problem of inaccurate selection of deep space orbiter image stereo pairs in application scenarios such as planetary mapping, and the problem of easily falling into local optimality.

[0006] The purpose of the present invention can be achieved by the following technical solutions:

[0007] The present invention provides a method for optimizing local to global deep space orbiter image stereo pairs, which is characterized by comprising the following steps:

[0008] Acquire deep space orbiter images and perform image preprocessing, determine the initial stereo image pairs by judging the overlapping area between images, and perform preliminary screening of the image pairs to obtain a set of preliminarily screened stereo image pairs;

[0009] Obtaining information of the region of interest and performing rasterization and gridding processing on the region of interest, calculating the local quality cost of each stereo image pair for each grid, and obtaining a local optimal image pair and a local candidate image pair set corresponding to the current grid based on the local quality cost;

[0010] Based on the local optimal image pair and the local candidate image pair set corresponding to each grid, the global quality cost of each stereo image pair is calculated;

[0011] The local quality cost and the global quality cost of the stereo image pair are combined to obtain a final preferred stereo image pair.

[0012] As a preferred technical solution, the process of acquiring deep space orbiter images and performing image preprocessing to obtain a stereo image pair set through initial screening of image pairs includes:

[0013] Obtain the initial deep space orbiter image set, remove images with radiation anomalies, and correct the image Shapefile according to the corner point coordinates to achieve preprocessing;

[0014] For all pre-processed deep space orbiter images, the overlapping area between images is calculated based on the geographical relationship between the images, and the initial deep space orbiter stereo image pair set is obtained based on the preset overlap threshold;

[0015] Calculate the solar azimuth difference of the initial deep space orbiter stereo image pair, the difference between the row and column sampling distance ratios of the left and right images, and the image pair intersection angle, and select the stereo image pair that meets the preset requirements.

[0016] As a preferred technical solution,

[0017] The local quality cost is calculated by combining the overlap rate of the stereo image pair and the grid, the intersection angle of the image pair and the number of matching points of the stereo image pair;

[0018] The global quality cost of each stereo pair is calculated by combining the repeated coverage ratio, effective area ratio and consistency of indicators within the neighborhood.

[0019] As a preferred technical solution, the overlap ratio of the stereo image pair and the grid is calculated using the following formula:

[0020]

[0021] Among them, c or (i,g) is the overlap cost of stereo pair i and grid g, r ov (i,g) represents the overlap ratio between stereo image pair i and grid g, r ov (g) min With r ov (g) max They represent the minimum and maximum overlap ratios of all stereo pairs and the grid g, respectively. ov (i,g) represents the overlapping area between the stereo pair i and the grid g, S gr (g) represents the area of ​​grid g.

[0022] As a preferred technical solution, the image pair intersection angle is calculated using the following formula:

[0023]

[0024] Among them, c ia (i) represents the intersection angle cost of stereo pair i, A ia_i is the intersection angle of stereo image pair i, a ia_min With A ia_max They represent the minimum and maximum intersection angles in the stereo pair set respectively.

[0025] As a preferred technical solution, the number of matching points of the stereo image pair is calculated using the following formula:

[0026]

[0027] Among them, c mp (i) is the cost of the number of matching points of stereo pair i, N mp_i Represents the number of matching points of stereo pair i, N mp_min and N mp_max Respectively represent the minimum and maximum number of matching points in the stereo pair set.

[0028] As a preferred technical solution, the repeated coverage ratio is calculated using the following formula:

[0029]

[0030] Among them, c rep (I g ) represents the selected image pair I in the grid g g The cost of repeated coverage ratio, N lo_can (I g ) represents the image pair I g The number of times it appears in the local candidate image pair, N lo_opt (I g ) represents the image pair I g The number of times it appears in the local optimal pair.

[0031] As a preferred technical solution, the effective area ratio is calculated using the following formula:

[0032]

[0033] Among them, c roi (I g ) is the selected image pair I in the grid g g The global effective area ratio cost, S ov_roi (I g ) represents the image pair I g The area of ​​the part that overlaps with the entire region of interest, SR Represents the area of ​​the entire region of interest.

[0034] As a preferred technical solution, the consistency of the indicators within the neighborhood is calculated using the following formula:

[0035]

[0036] Among them, c con (I g ,N g ) represents the selected image pair I in the grid g g With adjacent grid N g The cost of indicator consistency, n and N g They represent one of the neighboring grids adjacent to grid g and all the eight neighboring grids respectively. T(·) is the indicator function, which means that when the image pairs selected in the two grids are different, the function value is 1, otherwise it is 0. φ(·) is the indicator function, which means that when there are no multiple connection points between the image pairs selected in the two grids, the function value is 1, otherwise it is 0.

[0037] As a preferred technical solution, it also includes:

[0038] Based on the final optimized stereo image pair, planetary mapping is achieved.

[0039] Compared with the prior art, the present invention has at least one of the following beneficial effects:

[0040] (1) The preferred stereo image pairs are more suitable for downstream scenarios such as planetary mapping: The present invention divides the deep space orbiter image screening into two parts, local optimization and global optimization, which have a progressive relationship, and effectively selects a set of stereo image pairs that meet the mapping requirements. In the case of complex distribution of stereo image pairs, the mapping success rate and accuracy of the selected stereo image pairs are guaranteed, which effectively replaces manual screening and improves mapping efficiency and automation.

[0041] (2) Ensure image pair quality and reduce redundancy: The present invention combines quality constraints with global constraints to improve the optimization model, ensure image pair quality and effectively reduce image pair redundancy.

[0042] (3) Improving mapping accuracy: The present invention provides image pair index consistency constraints, increases the association between selected image pairs, and is conducive to improving mapping accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 A schematic diagram of a flow chart of a method for optimizing local to global stereoscopic deep space orbiter images in an embodiment;

[0044] Figure 2 This is a schematic diagram of Shapefile before and after correction in the embodiment;

[0045] Figure 3A schematic diagram of generating a Shapefile for a stereo image pair in an embodiment;

[0046] Figure 4 Schematic diagram of the distribution of the initial stereo image pairs of (a) the Moon and (b) Mars within the ROI in the embodiment;

[0047] Figure 5 Schematic diagram of the distribution of (a) the moon and (b) Mars preferred stereo image pairs within the ROI in the embodiment;

[0048] Figure 6 Schematic diagram of three-dimensional terrain constructed by the preferred image pair of (a) the moon and (b) Mars in the embodiment;

[0049] Figure 7 Schematic diagram of an electronic device in an embodiment. DETAILED DESCRIPTION

[0050] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0051] Example 1

[0052] In response to the problems existing in the aforementioned prior art, the present embodiment provides a local-to-global method for optimizing stereo pairs of deep space orbiter images, which performs initial screening of image pairs based on orbiter image metadata, rasterizes and grids the region of interest, and implements local optimization based on stereo pair quality constraints; combines global distribution constraints with consistency constraints on adjacent image pair indicators, constructs and solves a global optimization model, and implements automated optimization of high-quality and less-redundant stereo image pairs.

[0053] See also Figure 1 This method mainly includes image preprocessing, initial image pair acquisition and initial screening of image pairs based on metadata, local optimal screening based on quality constraints, and global optimal screening combined with distribution constraints and adjacent index consistency constraints, and finally achieves high-quality and low-redundancy deep space orbiter image stereo pair optimization. This method can automatically and effectively select stereo image pairs that meet the requirements of photogrammetry and mapping. This method includes the following steps:

[0054] Step S1, acquiring deep space orbiter images and performing image preprocessing, determining an initial stereo image pair by judging the overlapping area between images, performing a preliminary screening of the image pairs based on metadata, and obtaining a set of preliminarily screened stereo image pairs.

[0055] In order to avoid the influence of abnormal images on the screening, images with abnormal radiation are removed by visually judging the image thumbnails in the preprocessing stage, because these abnormal images cannot provide effective texture information. The retained images may have file anomalies, which are generally manifested as irregular polygons in the shapefile image outline, such as Figure 2 (a) is shown. In addition to recording the image corner coordinates, the wrong Shapefile also records the non-corner coordinates. Find the correct image corner coordinates from these coordinates, arrange them in order and replace the corner coordinates in the wrong Shapefile to achieve image Shapefile correction. The correction result is shown in Figure 2 As shown in (b), red and blue represent two sets of example data.

[0056] The preprocessed images all correspond to a correct Shapefile. The overlapping area between the images is calculated based on the geographic relationship between the Shapefiles, and the initial stereo image pair set Set is obtained according to the given overlap threshold. raw The image metadata is used to calculate the solar azimuth difference of the left and right images of the stereo pair, the difference between the row and column sampling distance ratios of the left and right images, and the image pair intersection angle. The stereo pairs whose indicators meet the given threshold are retained, which is the initial screened image pair set Set pre .

[0057] In stereo image pair screening, the two images that constitute the image pair are regarded as a whole. In order to simplify the screening process, the Shapefile of the overlapping part of the two images is obtained according to the geographical location of the two images, such as Figure 3 The overlapping part of the two images is used to represent the stereo image pair for subsequent screening.

[0058] Step S2, obtaining the information of the region of interest and performing rasterization and gridding processing on the region of interest, calculating the local quality cost of each stereo image pair for each grid, and obtaining the local optimal image pair and the local candidate image pair set corresponding to the current grid based on the local quality cost.

[0059] After determining the region of interest (ROI), the ROI bounding rectangle is rasterized and the area is gridded according to a given width. The local optimal screening of stereo pairs is to select the optimal and alternative image pairs in each grid. In the local optimal screening, the key parameters in the orbiter image metadata are extracted, and the quantized parameters are used to define the quality of the stereo pair. At the same time, the overlap ratio of the stereo pair and the current grid is combined to determine the stereo pair quality cost of the current local grid:

[0060]

[0061] Among them, C loc(i,g) represents the quality cost of the i-th stereo image pair in the g-th grid; C is the cost value corresponding to each quality constraint indicator, including the cost value c of the overlap rate between the image pair and the grid or , the cost of the image to the intersection angle c ia The cost value c corresponding to the number of matching points of the image pair mp ; W = {w or ,w ia ,w mp} is the weight corresponding to each cost value. Since the above indicators are derived from different measurements, each indicator is normalized when defined, which ensures the consistency of the evaluation process. The smaller the value, the better the quality. The three local quality costs of the stereo pair are defined as follows:

[0062] (1) Overlap ratio between stereo image pair and grid.

[0063] The footprints of deep space orbiter images are mostly polygons with large aspect ratios, and the heading is not parallel to the longitude or latitude, resulting in partial overlap between the image and some grids divided by the ROI. This overlap ratio determines the coverage of the image in the grid, so this indicator has a great impact on the coverage of stereo mapping within the ROI and can be used as one of the indicators to represent image quality. The overlap ratio cost is defined as:

[0064]

[0065] Among them, c or (i,g) is the overlap ratio cost value between stereo image pair i and grid g; r ov (i,g) represents the overlap ratio between stereo image pair i and grid g; r ov (g) min With r ov (g) max Respectively represent the minimum and maximum overlap ratios of all stereo pairs and the grid g; S ov (i,g) represents the overlapping area of ​​stereo image pair i and grid g; S gr (g) represents the area of ​​grid g. The smaller the overlap cost, the better the quality of the image pair in the current grid.

[0066] (2) Image intersection angle.

[0067] The intersection angle is an important indicator in stereo image pairs and has a great impact on the elevation accuracy of the mapping results. In order to avoid the situation where poor positioning accuracy is caused by weak intersection, selecting a suitable intersection angle is a key step in image pair screening. The intersection angle cost is defined as:

[0068]

[0069] Among them, c ia(i) represents the intersection angle cost of stereo image pair i; A ia_i is the intersection angle of stereo image pair i; A ia_min With A ia_max They represent the minimum and maximum intersection angles in the stereo pair set. The smaller the intersection angle cost value, the better the quality of the stereo pair.

[0070] (3) Number of matching points in stereo image pairs.

[0071] The result of feature matching can effectively reflect the quality of the overlapping part of the stereo image pair. A large number of matching points indicates that the terrain features of the overlapping area are obvious, the overlapping range is acceptable, and sufficient connection points can be provided for bundle adjustment, which has a non-negligible impact on the mapping results. The cost of the number of matching points is defined as:

[0072]

[0073] Among them, c pp (i) is the cost value of the number of matching points of stereo image pair i; N mmp_i N represents the number of matching points of stereo pair i; mp_min and N mp_max They represent the minimum and maximum number of matching points in the stereo pair set. The smaller the cost value, the better the quality of the stereo pair.

[0074] In determining the stereo pair quality cost C of the local grid loc After (i, g), the cost values ​​of the local grid stereo image pair set are sorted, and the image pair with the smallest cost value is selected as the local optimal image pair of the current grid. The other image pairs are selected in order from the smallest to the largest cost value as candidate image pairs. All local grids are traversed to obtain the local optimal image pair set Set in the entire ROI. loc and the local candidate image pair set Set can ,It should be explained that each local grid has an optimal image pair selection and an ordered set of ,alternative image pairs.

[0075] Step S3, calculating the global quality cost of each stereo image pair based on the local optimal image pair and the local candidate image pair set corresponding to each grid.

[0076] Global optimal screening achieves optimization by combining multiple constraints on spatial distribution, including:

[0077] (1) Percentage of duplicate coverage.

[0078] In the local grid, the ratio of the number of times the current stereo pair appears in the local candidate pair set results to the total number of times it appears in the local optimal and candidate results is the repeated coverage ratio of the stereo pair, and its cost is defined as:

[0079]

[0080] Among them, c rep (I g ) represents the selected image pair I in the grid g g The cost of repeated coverage ratio; N lo_can (I g ) represents the image pair I g The number of times it appears in the local alternative results; N lo_opt (I g ) represents the image pair I g The number of times it appears in the local optimal result. The smaller the cost value, the better the image is. g The better the quality and distribution.

[0081] (2) Percentage of effective area.

[0082] The global effective area ratio of the stereo image pair is determined according to the overlapping area of ​​the stereo image pair and the entire ROI, and its cost value is defined as:

[0083]

[0084] Among them, c roi (I g ) The selected image pair I in the grid g g The global effective area ratio cost; S ov_roi (I g ) represents the image pair I g The area of ​​the part overlapping with the entire ROI; S R Represents the area of ​​the entire ROI. The smaller the cost value, the closer the image is to I g The better the effective proportion within the ROI, the greater the possibility of being selected as the preferred image pair.

[0085] (3) Consistency of indicators within the neighborhood.

[0086] According to the correlation index between the current stereo pair and the selected stereo pair in the adjacent grid, the consistency of the index in the neighborhood is determined. More specifically, there should be multiple connection points between the ideal stereo pair and the adjacent stereo pair to ensure the correlation between the stereo pairs, which is conducive to improving the accuracy of the final mapping product. The cost value of this index is defined as:

[0087]

[0088] Among them, c con (I g ,N g ) represents the selected image pair I in the grid g g With adjacent grid N g The cost of indicator consistency; n and Ng They represent one of the neighboring grids and all the eight neighboring grids of the grid g respectively; T(·) is an indicator function, which means that when the image pairs selected in the two grids are different, the function value is 1, otherwise it is 0; φ(·) is also an indicator function, which means that when there are no multiple connection points between the image pairs selected in the two grids, the function value is 1; otherwise it is 0. The smaller the cost value, the more it means that there are multiple connection points between the adjacent image pairs, and the stronger the correlation between the image pairs.

[0089] Combining the above three constraints, the model for constructing global optimal screening is as follows:

[0090]

[0091] Among them, C glo (I) represents the global cost value of the selected image pair set; G represents all grids; w d With w c Represent the weights of distribution constraints and consistency constraints respectively. The smaller the global cost value, the better the overall results of image quality and distribution.

[0092] Step S4: combining the local quality cost and the global quality cost of the stereo image pair to obtain a final optimal stereo image pair.

[0093] Combining the above-mentioned local quality constraints with the global multiple distribution constraints, we can obtain the local-to-global stereo pair optimization model:

[0094]

[0095] The optimization model can be expressed as a Markov random field, using the greedy algorithm to initialize the result, and the belief propagation algorithm to solve the optimal result, which is the final preferred stereo pair Set opt .

[0096] The following experiments will illustrate the effectiveness of this method:

[0097] (1) Experimental data.

[0098] In order to verify the effectiveness of this method, the Lunar Reconnaissance Orbiter Narrow Angle Camera (LRO NAC) images and the Mars Reconnaissance Orbiter Context Camera (MRO CTX) images were selected as experimental data. The experimental area of ​​the moon is 35×45km centered at (20.5°N, 30.8°E), and the experimental area of ​​Mars is 180×270km centered at (23.7°N, 18.7°W). After the initial screening of image pairs, the number of initial images in the lunar ROI is 225, and the number of stereo image pairs is 1033; the number of initial images in the Mars ROI is 343, and the number of stereo image pairs is 5992.

[0099] (2) Experimental results

[0100] In order to verify the performance of the proposed method, this embodiment performs a qualitative evaluation by comparing the distribution diagrams of stereo pairs in the ROI before and after screening; the performance of the proposed method is quantitatively evaluated in detail using five different indicators: the number of images (NI), the number of stereo pairs (NS), the average intersection angle of stereo pairs (AIA), the redundancy rate of stereo pairs (RR), and the coverage rate of stereo pairs in the ROI (CR). The calculation formulas of AIA and RR are as follows:

[0101]

[0102] Where I represents the set of selected image pairs; A ia_i represents the intersection angle of stereo image pair i; N I Represents the number of all image pairs; S i_ROI represents the area of ​​stereo image pair i within ROI; S R is the area of ​​ROI. In order to verify the correctness and availability of the selected image pair set, the three-dimensional terrain of ROI is constructed by photogrammetry and mapping, and the terrain product released by NASA is selected as the reference terrain. The reliability of the proposed method is verified by statistically analyzing the elevation deviation between the three-dimensional terrains. Due to the low resolution of the Mars reference terrain, the laser altimetry data within the Mars ROI is selected as the reference terrain elevation data for comparison and verification.

[0103] The two sets of data were subjected to local optimal screening and global optimal screening in turn according to the method proposed in this patent, and the following results were obtained:

[0104] For the lunar ROI, the number of images after screening is 27, the number of stereo pairs is 24, the average intersection angle is 23.32°, the redundancy rate of stereo pairs is 113%, and the coverage rate of stereo pairs in ROI is 100%;

[0105] For the Mars ROI, the number of images after screening is 65, the number of stereo pairs is 50, the average intersection angle is 17.97°, the redundancy rate of stereo pairs is 146%, and the coverage rate of stereo pairs in ROI is 100%.

[0106] Schematic diagram of the distribution of stereo image pairs in ROI before and after the selection of the moon and Mars Figure 4 and Figure 5 As shown, Figure 4 is the initial distribution of stereo pairs, Figure 5 This is the distribution of stereo image pairs after optimization.

[0107] The quantitative indicators of the stereo image pairs before and after statistical screening are shown in Table 1:

[0108] Table 1 Quantitative indicators of stereo image pairs before and after screening

[0109]

[0110] The three-dimensional topographic results of the experimental area are as follows Figure 6 As shown in Table 2, the elevation deviation statistics of the three-dimensional terrain produced by the preferred image pair and the reference terrain (the Mars reference elevation is the laser altimetry data) are shown.

[0111] Table 2 Statistics of elevation deviation of three-dimensional terrain

[0112]

[0113] The above experimental results show that the stereo image pairs selected by this method have the characteristics of reasonable distribution, low redundancy, and good image quality. In addition, the three-dimensional terrain produced by the selected image pairs has a good effect, and the elevation deviation from the reference terrain is within an acceptable range. Therefore, this method can effectively improve the efficiency and automation of planetary mapping while ensuring the accuracy of mapping.

[0114] Example 2

[0115] This embodiment provides an electronic device, including: one or more processors and a memory, wherein the memory stores one or more programs, and the one or more programs include instructions for executing the local to global deep space orbiter image stereo pair optimization method as described in Example 1.

[0116] like Figure 7 As mentioned above, at the hardware level, the electronic device includes a processor, an internal bus, a network interface, a memory, and a non-volatile memory, and may also include other hardware required for the business. The processor reads the corresponding computer program from the non-volatile memory into the memory and then runs it to achieve the above Figure 1The local to global deep space orbiter image stereo pair optimization method. Of course, in addition to the software implementation, the present invention does not exclude other implementations, such as logic devices or a combination of software and hardware, etc., that is, the execution subject of the following processing flow is not limited to each logic unit, but can also be hardware or logic devices.

[0117] Compared with the existing technology, the present invention divides the deep space orbiter image screening into two parts, local optimization and global optimization, which have a progressive relationship, and effectively selects a set of stereo image pairs that meet the mapping requirements; combines quality constraints and global constraints to improve the optimization model, ensure the quality of image pairs and effectively reduce image pair redundancy; proposes image pair index consistency constraints, increases the correlation between selected image pairs, and is conducive to improving mapping accuracy. The advantage of the present invention is that it can effectively handle the situation of complex distribution of stereo image pairs, ensure the mapping success rate and accuracy of the selected stereo image pairs, effectively replace manual screening, and improve mapping efficiency and automation.

[0118] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can easily think of various equivalent modifications or replacements within the technical scope disclosed by the present invention, and these modifications or replacements should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention shall be based on the protection scope of the claims.

Claims

1. A method for optimizing local to global deep space orbiter image stereo pairs, characterized in that: The steps include: Acquire deep space orbiter images and perform image preprocessing, determine the initial stereo image pairs by judging the overlapping area between images, and perform preliminary screening of the image pairs to obtain a set of preliminarily screened stereo image pairs; Obtaining information of the region of interest and performing rasterization and gridding processing on the region of interest, calculating the local quality cost of each stereo image pair for each grid, and obtaining a local optimal image pair and a local candidate image pair set corresponding to the current grid based on the local quality cost; Based on the local optimal image pair and the local candidate image pair set corresponding to each grid, the global quality cost of each stereo image pair is calculated; The local quality cost and the global quality cost of the stereo image pair are combined to obtain a final preferred stereo image pair.

2. The method for optimizing local to global deep space orbiter image stereo pairs according to claim 1, characterized in that: The process of acquiring deep space orbiter images and preprocessing them to obtain a stereo image pair set through initial screening of image pairs includes: Obtain the initial deep space orbiter image set, remove images with radiation anomalies, and correct the image Shapefile according to the corner point coordinates to achieve preprocessing; For all pre-processed deep space orbiter images, the overlapping area between images is calculated based on the geographical relationship between the images, and the initial deep space orbiter stereo image pair set is obtained based on the preset overlap threshold; Calculate the solar azimuth difference of the initial deep space orbiter stereo image pair, the difference between the row and column sampling distance ratios of the left and right images, and the image pair intersection angle, and select the stereo image pair that meets the preset requirements.

3. The method for optimizing local to global deep space orbiter image stereo pairs according to claim 1, characterized in that: The local quality cost is calculated by combining the overlap rate of the stereo image pair and the grid, the intersection angle of the image pair and the number of matching points of the stereo image pair; The global quality cost of each stereo pair is calculated by combining the repeated coverage ratio, effective area ratio and consistency of indicators within the neighborhood.

4. The method for optimizing local to global deep space orbiter image stereo pairs according to claim 3, characterized in that: The overlap ratio of the stereo image pair and the grid is calculated using the following formula: Among them, c or (i,g) is the overlap cost of stereo pair i and grid g, r ov (i,g) represents the overlap ratio between stereo image pair i and grid g, r ov (g) min With r ov (g) ma They represent the minimum and maximum overlap ratios of all stereo pairs and the grid g, respectively. ov (i,g) represents the overlapping area between the stereo pair i and the grid g, S gr (g) represents the area of ​​grid g.

5. The method for optimizing local to global deep space orbiter image stereo pairs according to claim 3, characterized in that: The image pair intersection angle is calculated using the following formula: Among them, c ia (i) represents the intersection angle cost of stereo pair i, A ia_i is the intersection angle of stereo image pair i, A ia_min With A ia_max They represent the minimum and maximum intersection angles in the stereo pair set respectively.

6. The method for optimizing local to global deep space orbiter image stereo pairs according to claim 3, characterized in that: The number of matching points of the stereo image pair is calculated using the following formula: Among them, c mp (i) is the cost of the number of matching points of stereo pair i, N mp_i Represents the number of matching points of stereo pair i, N mp_min and N mp_max Respectively represent the minimum and maximum number of matching points in the stereo pair set.

7. The method for optimizing local to global deep space orbiter image stereo pairs according to claim 3, characterized in that: The duplicate coverage ratio is calculated using the following formula: Among them, c rep (I g ) represents the selected image pair I in the grid g g The cost of repeated coverage ratio, N lo_can (I g ) represents the image pair I g The number of times it appears in the local candidate image pair, N lo_opt (I g ) represents the image pair I g The number of times it appears in the local optimal pair.

8. The method for optimizing local to global deep space orbiter image stereo pairs according to claim 3, characterized in that: The effective area ratio is calculated using the following formula: Among them, c roi (I g ) is the selected image pair I in the grid g g The global effective area ratio cost, S ov_roi (I g ) represents the image pair I g The area of ​​the part that overlaps with the entire region of interest, S R Represents the area of ​​the entire region of interest.

9. The local-to-global deep space orbiter image stereo pair optimization method according to claim 3, characterized in that: The consistency of the indicators within the neighborhood is calculated using the following formula: Among them, c con (I g ,N g ) represents the selected image pair I in the grid g g With adjacent grid N g The cost of indicator consistency, n and N g They represent one of the neighboring grids adjacent to grid g and all the eight neighboring grids respectively. T(·) is the indicator function, which means that when the image pairs selected in the two grids are different, the function value is 1, otherwise it is 0. φ(·) is the indicator function, which means that when there are no multiple connection points between the image pairs selected in the two grids, the function value is 1, otherwise it is 0.

10. The local-to-global deep space orbiter image stereo pair optimization method according to claim 1, characterized in that: Also includes: Based on the final optimized stereo image pair, planetary mapping is achieved.

Images