A method and system for crater identification for landing visual navigation

Abstract

A crater identification method and device for landing visual navigation, comprising the following steps: establishing a crater pair database: selecting a proper number of craters to establish a database, removing small craters and overlapping craters according to selection principles to obtain a crater pair table; screening and matching the crater pair database; and re-projecting and verifying the screened and matched crater pairs and estimating the current pose. The database of the present application only stores the serial numbers of the crater pairs and corresponding invariants, greatly reducing the database capacity and improving the search efficiency. The present application proposes a dynamic threshold to solve the problem that the projective invariant is sensitive to noise, improving the matching rate and average processing speed. Meanwhile, the iterative pyramid algorithm is proposed to quickly screen the redundancy, estimate the pose on the basis of preliminary identification, and verify the identification result after re-projecting and matching. The present application has the characteristics of high identification rate, robustness to noise, and accurate and fast pose estimation, reaching the leading performance level of initial identification in landing navigation.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of planetary landing navigation, in particular to a crater identification method and system for landing visual navigation. BACKGROUND

[0002] Planetary landing navigation technology has become an important part of deep space exploration. The relative navigation method based on terrain is an optical autonomous navigation method, which enhances the navigation capability by providing measurement information relative to the known celestial surface. The navigation method based on craters has the advantages of low power consumption, saving memory, high calculation efficiency, rich available terrain features, and can estimate absolute pose, etc., and has a wide research and application prospect. Unlike tracking recognition with initial attitude information, the crater identification method without providing any initial attitude is also called "Lost In Space (LIS)" identification, which is more difficult and has greater research significance.

[0003] In the optical navigation method, the method of directly using the crater pattern for image matching is only applicable to tracking recognition and is not applicable to LIS (Lost In Space) identification. In addition, the method of using inertial devices for navigation has a large cumulative error and can only measure the relative pose, which can be corrected by the optical navigation method. The method of estimating the absolute pose using craters generally uses the projective invariant of the elliptic quadratic curve pair for crater identification, but the projective invariant is disturbed by the noise of the elliptic parameter detection. Although the LIS (Lost In Space) identification rate can be improved by increasing the projective invariant, this will cause the database to be more massive, to a certain extent, it will increase the redundancy of the crater triangle search, and limit the further improvement of the LIS (Lost In Space) identification rate.

[0004] LIS (Lost In Space) identification is the process of matching the craters in the picture with the craters in the pre-established crater database. Currently, there are still the following difficulties: the identification rate is sensitive to the error of the elliptical crater detection; the large matching library causes slow search, low efficiency, and more redundant matching will result in a lower identification rate; the unstable state of the lander during landing often causes a large tilt angle, which causes the actual crater projected onto the image plane to be greatly deformed, increasing the difficulty of identification. In view of the above problems, an effective crater LIS (Lost In Space) identification method has not been proposed so far. SUMMARY

[0005] The present application provides a crater identification method for landing visual navigation. The method of the present application is a crater LIS (Lost In Space) identification method based on elliptical crater pairs, establishes a small-capacity and high-efficiency crater pair database, proposes an iterative pyramid algorithm to obtain a robust identification result, designs a dynamic threshold to quickly reduce redundancy, and improves the algorithm efficiency and identification rate.

[0006] The technical scheme of the present application is as follows:

[0007] The first aspect of the present application provides a crater identification method for landing visual navigation, the method comprising:

[0008] Step S100, establishing a crater pair database: selecting a proper number of craters to establish a database, removing small craters and overlapping craters according to the selection principle to obtain a crater pair table;

[0009] Step S200, screening and matching the crater pair database;

[0010] Step S300, re-projecting and verifying the screened and matched crater pairs and estimating the current pose.

[0011] Further, the step S200 comprises:

[0012] Step S210, pre-processing the image;

[0013] Step S220, screening the craters based on a dynamic threshold;

[0014] Step S230, constructing a triple based on the pyramid base triple search algorithm for the screened craters;

[0015] Step S240, performing iterative pyramid matching on the constructed triple based on the pyramid LIS identification algorithm.

[0016] Further, the step S230 comprises:

[0017] Step S231, setting a state identifier for each crater in the crater table, initializing it before matching and identification, and setting the state of each crater to 0;

[0018] Step S232, scanning the set , setting the state of all craters contained in the crater pair in the set to 1, and recording the serial number j of the other crater constituting the crater pair;

[0019] Step S233, scanning the set , if the state of the corresponding crater is not 0, setting the state of the crater to 2, and recording the serial number j of the other crater constituting the crater pair; ;

[0020] Step S234, scanning the state table, if the state of the crater is 2, judging whether the recorded belongs to the set , if yes, storing it into the set .

[0021] Furthermore, step S240 includes:

[0022] Step S241: Read the current sequence image and generate a set of triples to be matched, sorted by redundancy from smallest to largest, according to the priority selection mechanism.

[0023] Step S242: Traverse the set to be matched. If the number of redundant triplets in the current set exceeds the upper limit, the triplet matching algorithm will take too long, so the image sequence matching is directly set to fail. For the current crater to be matched... The binary search database forms a set of meteorite crater pairs. ;

[0024] Step S243: Input the triplet matching algorithm to obtain the set. ;

[0025] Step S244: Traverse the remaining impact craters. The binary search method yields the set. The set is obtained by inputting the pyramid algorithm. If the number of triangles is reduced to 1, the pyramid algorithm returns the result of this triplet identification and the last impact crater. The algorithm terminates when the intersection of the given points is found.

[0026] Step S245, if set The number of elements is 0, return to step S242;

[0027] Step S246, if the meteorite crater This is the last impact crater in the current image. Return to step S242.

[0028] Step S247: If the number of matched triples reaches the upper limit. The algorithm ends.

[0029] Furthermore, in step S220, a dynamic threshold is used, denoted as... The dynamic threshold factor is defined as the ratio of the change in an invariant to the true value of the invariant, denoted as . Therefore, the dynamic threshold calculated for each pair of impact craters ;

[0030] Invariants calculated for each pair of meteorite craters and through dynamic threshold factor Calculated search range The lower and upper limits of the search for two sets of invariants are obtained by the following formula, denoted as Binary search yields all satisfying... Possible meteorite crater pair collection,

[0031]

[0032] Furthermore, in step S2300, based on the camera's perspective projection model, the world coordinate system of the crater is transformed into the pixel coordinates of the image. The crater identification algorithm estimates the correspondence between the crater observed on the image plane and the crater in the world coordinate system. The LIS identification algorithm outputs the identification results of multiple craters and the crater that constitutes the last pyramid. The identification set is based on the 3D coordinates of multiple crater centers in the world coordinate system and the 2D coordinates in the image coordinate system, as well as the camera intrinsic parameter matrix. The transformation matrix from the world coordinate system to the camera coordinate system is estimated, and the craters are traversed. The pose is obtained by solving the set;

[0033] For each possible impact crater First, determine if the crater could be within the same field of view as the previously identified craters. If so, solve for the rotation matrix and translation vector. Then, transform all craters to the image coordinate system to obtain a reprojected image. Compare the corresponding distances of the craters in the reprojected image and the ground truth image. If the Gaussian angular distance is less than 1.57, it indicates that the current pose is correct, meaning the current crater is within the correct range. For correct identification.

[0034] A second aspect of the present invention provides a crater identification system for landing visual navigation, comprising:

[0035] The database creation module is used to create a crater pair database: select an appropriate number of craters to create the database, and based on the selection criteria, remove small craters and overlapping craters to obtain a crater pair table;

[0036] The filtering module is used to filter and match the meteorite crater data in the database;

[0037] Verification module: Used to perform reprojection verification on the selected and matched crater pairs.

[0038] Another aspect of the present invention provides an electronic device, the electronic device comprising:

[0039] At least one processor; and,

[0040] A memory communicatively connected to the at least one processor; wherein,

[0041] The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the method.

[0042] The present invention also provides a non-transitory computer-readable storage medium that stores computer instructions for causing the computer to perform the method described thereon.

[0043] Compared with the prior art, the present invention has the following beneficial technical effects:

[0044] This invention addresses the problem of massive databases by storing only the crater pair indices and corresponding invariants, and establishing only the projective invariants of crater pairs, thus greatly reducing database size and improving search efficiency.

[0045] This invention addresses the problem of projective invariants being greatly affected by noise by proposing a dynamic threshold to reduce screening redundancy and solve the problem of projective invariants being sensitive to noise, thereby improving the matching rate and average processing speed.

[0046] This invention proposes an iterative pyramid algorithm to quickly filter redundancies, estimate pose based on preliminary identification, verify identification results after reprojection matching, and proposes an iterative pyramid algorithm to quickly obtain unique matching results, providing a solution for crater identification based on landing images without an initial pose.

[0047] This invention features high recognition rate, robustness to noise, and accurate and fast attitude estimation. It is applicable to situations without any initial attitude and achieves leading performance in landing navigation initialization recognition. Attached Figure Description

[0048] Figure 1 This is a schematic diagram of the meteorite crater identification process of the present invention;

[0049] Figure 2 This is a schematic diagram illustrating the selection of meteorite craters to form a database according to the present invention;

[0050] Figure 3 This is a schematic diagram of the LIS (Lost Space) identification process of the present invention;

[0051] Figure 4 This is a schematic diagram illustrating the construction of a triplet set based on a state table according to the present invention;

[0052] Figure 5 To add a fourth meteorite crater to the present invention A schematic diagram illustrating the process of eliminating redundancy in a pyramid structure;

[0053] Figure 6 This is a schematic diagram of pose estimation according to the present invention;

[0054] Figure 7 This diagram illustrates the false detection of overlapping meteorite craters in this invention. Detailed Implementation

[0055] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments and the accompanying drawings. It should be understood that these descriptions are merely exemplary and not intended to limit the scope of the invention. Furthermore, descriptions of well-known structures and techniques are omitted in the following description to avoid unnecessarily obscuring the concept of the invention.

[0056] The present invention will now be described in further detail with reference to specific embodiments.

[0057] Specifically, such as Figure 1 As shown, the present invention provides a crater identification method for landing visual navigation, comprising: selecting craters to establish a small-capacity, high-efficiency crater database; using an iterative pyramid algorithm to match the craters in the database; reprojecting to verify the identification results and estimating the current pose.

[0058] Considering the unstable landing conditions, the camera will not always be shooting vertically downwards. Increased tilt angles of the lander often lead to increased ellipticity of the crater in the image plane. Therefore, this invention uses elliptical features for crater identification. Let... Let these be the coordinates of the center of the ellipse. The major and minor semi-axes of the ellipse, The angle of inclination of the ellipse. The equation of the ellipse is expressed by equation (1), which represents the coordinates of points on an elliptic curve:

[0059] (1)

[0060] The parameters of A, B, C, D, F, and G are described in (2).

[0061] (2)

[0062] Let elliptic matrix Then equation (1) can be written in matrix form:

[0063] (3)

[0064] The key to crater identification lies in finding the invariant relationships of craters within projective transformations, and invariant theory proves the existence of crater projection invariants. Because spacecraft are relatively close to the planetary surface during landing, it can be assumed that all craters are locally coplanar. In this case, pairs of elliptical craters can form quadratic curve invariants. As shown in equation (4):

[0065] (4)

[0066] Where Tr represents the trace of a matrix. , The normalized elliptic parameter matrix is ​​shown in equation (5).

[0067] (5)

[0068] Specifically, the method of the present invention includes the following steps:

[0069] Step S100: Establish the meteorite crater database.

[0070] A suitable number of impact craters should be selected to build a database. Too many craters will result in a large database, leading to increased redundancy in crater filtering and increased computational load; conversely, too few craters will lead to a decreased matching rate. Appropriate selection criteria should be designed to remove small and overlapping craters, resulting in a crater pair table. Removing small craters requires first selecting based on the centerline (…). The map is divided into two regions, left and right. Craters with a radius less than 9 pixels are removed from the left region, and craters with a radius less than 7 pixels are removed from the right region. For example... Figure 2 As shown in Table 1, the crater table is helpful for designing the state table in the identification algorithm process and speeds up the algorithm efficiency.

[0071] Let d be the maximum field of view diameter visible at the highest landing altitude. For all craters less than d apart... Calculate the invariants according to formula (3) The constructed database is as follows Figure 2 As shown in Table 2, the database of this invention has advantages such as small capacity, fast search, and suitability for iterative recognition algorithms. Compared to storing at least 7 projective invariants formed by crater triangles, this database only stores 2 invariants formed by crater pairs. Whether from the perspective of the number of stored records or the number of invariants in a single record, the database capacity is greatly reduced, thereby increasing the search speed.

[0072] Step S200: Filter and match the meteorite crater data against the database;

[0073] The crater selection and matching algorithm includes image preprocessing, crater pair selection based on dynamic thresholds, crater triplet construction, and iterative pyramid matching. The image preprocessing principles correspond to the crater selection principles: removing overlapping craters, sorting the remaining craters in the current image sequence from largest to smallest size, and identifying craters according to this sorting. After preprocessing, three craters are selected. By pairwise calculation of the invariants and dynamic thresholds for each of the three crater pairs, a database is searched to obtain a set of three crater pairs. .pass The set of three impact crater pairs yields all consecutive triples, forming a set of triples. Finally, a fourth impact crater L was added, along with... Constructing a pyramid, the same method is used to search the database to obtain the collection. , combined The set of triples can remove some redundancy; each additional L can remove [redundancy]. Part of the redundancy is resolved when only one triplet remains, at which point the identification ends and the last set of L is retained.

[0074] Specifically, the steps for filtering and matching the meteorite crater against the database include:

[0075] Step S210: Preprocess the image;

[0076] Before implementing a crater recognition algorithm, the craters in the image need to be preprocessed. Preprocessing principles include removing overlapping craters and prioritizing larger craters. This is because overlapping craters are prone to false detection (e.g., ...). Figure 7 As shown in the image, all overlapping craters are first removed. Secondly, because the invariant values ​​calculated for large craters have a smaller range of variation (<1000), large craters are prioritized for identification. Figure 7 As shown, when elliptical craters 1 and 2 overlap, it is generally easier to detect only 1. When elliptical craters 2 and 3 overlap, it is generally easy to mistakenly detect 4.

[0077] Specifically, during the landing process, the craters on the image plane will become larger and larger, so the scale difference is large; large crater is a relative concept. In the current image sequence, all craters are sorted from largest to smallest according to their area, and the largest crater is identified first.

[0078] Step S220: Filter crater pairs based on dynamic thresholds;

[0079] This invention uses a dynamic threshold, denoted as The dynamic threshold factor is defined as the ratio of the change in an invariant to its true value, denoted as . Therefore, the dynamic threshold calculated for each pair of impact craters .

[0080] Invariants calculated for each pair of meteorite craters and through dynamic threshold factor Calculated search range From formula (6), we can obtain the lower and upper limits of the search for two sets of invariants, denoted as... Binary search yields all results satisfying the condition. A possible set of impact crater pairs.

[0081] (6)

[0082] Step S230: Construct triplets for the selected crater pairs based on the pyramid base triplet search algorithm;

[0083] There are many impact crater pairs that meet the invariant threshold range. We filter all possible impact crater pairs based on a pyramid structure. First, we need to find all redundant sets of the pyramid's base triplet, such as... Figure 4 As shown.

[0084] Let the sets of pit pairs that have been searched be respectively The number of pit pairs it contains is The first step in identification is to find three pit pairs. And satisfy These craters are connected end-to-end, meaning that every pair of craters shares one and only one common impact crater. (The following is a list of conditions that must be met:) These can form the base triples of a pyramid, and their set is denoted as . Generally speaking, in Searching using traversal and combination methods in three sets requires... This search involves numerous comparison operations, which consumes a considerable amount of time. To avoid this, a crater table can be used (…). Figure 3 Table 1 uses a simple and fast search method based on status flags, which finds matching conditions by setting and judging status flags. ),like Figure 4 As shown.

[0085] Specifically, the steps for constructing triples for the selected crater pairs based on the pyramid-based triplet search algorithm include:

[0086] 1) Set a status flag for each crater in the crater table and initialize it before matching and identification by setting the status of each crater to 0.

[0087] 2) Scan Set the state of all crater pairs in the set to 1, and record the index j of the other crater that forms a crater pair with it.

[0088] 3) Scanning If the state of the corresponding crater is not 0, then set the state of the crater to 2 and record the number of the other crater that forms a crater pair with it. ;

[0089] 4) Scan the status table. If the crater status is 2, then determine the recorded status. Does it belong to a set? If so, then store in In the collection.

[0090] Step S240: Based on the pyramid-based LIS (Lost Space) identification algorithm, iterative pyramid matching is performed on the crater pairs constructed from the triples.

[0091] The purpose of the pyramid algorithm is to find the triplet set Finding the correct identifier requires providing filtering conditions to remove redundant triples in the set. The pyramid algorithm adds a crater to the base triples to form a crater pyramid, thus adding three more crater pairs to filter redundancy.

[0092] The fourth impact crater and the basic triplet can each form a 3-crater pair. The invariants and thresholds are calculated using the same method, and the database is searched to obtain the set of impact crater pairs. ,like Figure 5 As shown. The correct impact crater. They must be in the set respectively In the middle, use this as the filtering condition to remove sets. Redundant triples in the pyramid. A single pyramid often cannot yield a unique selection result, so multiple iterations are needed to select multiple impact craters. The pyramid is constructed. Meteorite craters can be obtained by sieving and reducing the triplet. The identification results are obtained, but there are often multiple solutions. Therefore, when only one triplet is obtained, the unique triplet result and the last impact crater are returned. The set of identifiers.

[0093] Generally, running the iterative pyramid algorithm on only a single set of triplets may result in redundancy being filtered to zero due to non-crater-related factors, or it may fail to obtain a unique identification result even after traversing all possible Ls. Therefore, it is necessary to traverse other sets of triplets to be identified and perform multiple iterations of the pyramid algorithm to find a unique identifier, combining a crater-based filtering algorithm with a pyramid bottom-level triplet search algorithm. The steps of the pyramid-based LIS (Lost Space in Space) identification algorithm include:

[0094] 1) Read the current sequence image and generate a set of triples to be matched, sorted by redundancy from smallest to largest, according to the priority selection mechanism;

[0095] 2) Traverse the set to be matched. If the number of redundant triplets in the current set exceeds the upper limit (20,000), the triplet matching algorithm will take too long, so the image sequence will be set to fail matching. For the current meteorite crater to be matched... The binary search database forms a set of meteorite crater pairs. ;

[0096] 3) Input triplet matching algorithm to obtain set ;

[0097] 4) Traverse the remaining impact craters The binary search method yields the set. The set is obtained by inputting the pyramid algorithm. If the number of triangles is reduced to 1, the pyramid algorithm returns the result of this triplet identification and the last impact crater. The algorithm terminates when the intersection of the given points is found.

[0098] 5) If set The number of elements is 0, return to step 2);

[0099] 6) If the impact crater This is the last impact crater in the current image; return to step 2).

[0100] 7) If the number of matched triples reaches the upper limit. upper limit The algorithm ends.

[0101] Step S300: Perform reprojection verification on the selected and matched crater pairs and estimate the current pose.

[0102] Set the world coordinate system to The center of the map is the origin of the world coordinate system, and the lander's body coordinate system is... Simultaneously, the camera is fixed on the lander and takes vertically downward photos; the camera coordinate system is... The camera coordinate system is transformed to the image plane through perspective projection transformation. According to the perspective projection model of the camera, the world coordinate system of the crater is transformed into the pixel coordinates of the image as shown in (7), where K is the camera intrinsic parameter matrix, R is the rotation matrix, t is the translation vector, and R and t constitute the pose T. = Here, (u, v) are the coordinates in the world coordinate system, and (u, v) are the pixel coordinates on the image plane. is the scale factor on the u-axis. The scale factor on the v-axis. The x-coordinate of the image center. y is the vertical coordinate of the image center, and s is the scale factor.

[0103] (7)

[0104] Crater identification can estimate the correspondence between craters observed on the image plane and craters in the world coordinate system. The LIS (Lost Space) identification algorithm outputs three craters. The identification results and the impact crater that forms the last pyramid The identification set identified a total of four impact craters. Given the 3D coordinates of the centers of these four craters in the world coordinate system and their 2D coordinates in the image coordinate system, as well as the camera intrinsic parameter matrix, the problem of estimating the transformation matrix from the world coordinate system to the camera coordinate system can be transformed into a PNP problem. Furthermore, because the impact craters... There are multiple solutions, so it is necessary to iterate through them. The pose is obtained by solving the set. For example... Figure 6 As shown.

[0105] For each possible impact crater First, determine whether it matches the already identified impact craters. Is it possible to find the craters within the same field of view? If so, solve for the rotation matrix and translation vector, and then transform all the craters to the image coordinate system according to formula (7) to obtain the reprojected image. Compare the corresponding distances of the craters in the reprojected image and the real image. If the Gaussian angle distance is less than 1.57, it indicates that the current pose is correct, that is, the current crater... For correct identification.

[0106] A second aspect of the present invention provides a crater identification system for landing visual navigation, characterized in that it includes: a database establishment module for establishing a crater pair database: selecting an appropriate number of craters to establish the database, and removing small craters and overlapping craters according to the selection principle to obtain a crater pair table;

[0107] The filtering module is used to filter and match the crater pairs in the database; the verification module is used to perform reprojection verification on the filtered and matched crater pairs.

[0108] The present invention also provides an electronic device, the electronic device comprising: at least one processor; and,

[0109] A memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the method.

[0110] The present invention also provides a non-transitory computer-readable storage medium that stores computer instructions for causing the computer to perform the method described thereon.

[0111] In summary, this invention provides a method and system for crater identification in landing visual navigation, comprising the following steps: establishing a crater pair database: selecting an appropriate number of craters to establish a database; removing small and overlapping craters according to selection principles to obtain a crater pair table; filtering and matching the crater pair database; and reprojecting and verifying the filtered and matched crater pairs and estimating their current pose. The database of this invention only stores the crater pair sequence number and corresponding invariants, greatly reducing database size and improving search efficiency. This invention proposes a dynamic threshold to address the sensitivity of projective invariants to noise, improving the matching rate and average processing speed. Simultaneously, an iterative pyramid algorithm is proposed to quickly filter redundancy, estimate pose based on preliminary identification, and verify the identification results after reprojection matching. This invention features high recognition rate, robustness to noise, and accurate and fast attitude estimation, achieving a leading performance level in landing navigation initialization identification.

[0112] It should be understood that the specific embodiments described above are merely illustrative or explanatory of the principles of the invention and do not constitute a limitation thereof. Therefore, any modifications, equivalent substitutions, improvements, etc., made without departing from the spirit and scope of the invention should be included within the protection scope of the invention. Furthermore, the appended claims are intended to cover all variations and modifications falling within the scope and boundaries of the appended claims, or equivalent forms of such scope and boundaries.

Claims

1. A method for identifying meteorite craters for landing visual navigation, characterized in that, The method includes: Step S100, establish a crater pair database: select an appropriate number of craters to establish a database, and according to the selection principle, remove small craters and overlapping craters to obtain a crater pair table. Step S200: Filter and match the meteorite crater data against the database; Step S200 includes: Step S210: Preprocess the image; Step S220: Filter crater pairs based on dynamic thresholds; In step S220, a dynamic threshold is used, denoted as... The dynamic threshold factor is defined as the ratio of the change in an invariant to the true value of the invariant, denoted as . Therefore, the dynamic threshold calculated for each pair of craters ; Invariants calculated for each pair of meteorite craters and through dynamic threshold factor Calculated search range The lower and upper limits of the search for two sets of invariants are obtained by the following formula, denoted as Binary search yields all satisfying... Possible meteorite crater pair collection, ； ； Step S230: Construct triplets for the selected crater pairs based on the pyramid base triplet search algorithm; Step S240: Based on the pyramid-based LIS recognition algorithm, iterative pyramid matching is performed on the crater pairs constructed from the triplet. Step S240 includes: Step S241: Read the current sequence image and generate a set of triples to be matched, sorted by redundancy from smallest to largest, according to the priority selection mechanism. Step S242: Traverse the set to be matched. If the number of redundant triplets in the current set exceeds the upper limit, the triplet matching algorithm will take too long. Therefore, the matching of this sequence of images is directly set to fail. For the current crater to be matched... The binary search database forms a set of meteorite crater pairs. ; Step S243: Input triplet matching algorithm to obtain set ; Step S244: Traverse the remaining impact craters. The binary search method yields the set. The set is obtained by inputting the pyramid algorithm. If the number of triangles is reduced to 1, the pyramid algorithm returns the result of this triplet identification and the last impact crater. The algorithm terminates when the intersection of the given points is found. Step S245, if set The number of elements is 0, return to step S242; Step S246, if the meteorite crater This is the last impact crater in the current image. Return to step S242. Step S247: If the number of matched triples reaches the upper limit. The algorithm ends; Step S300: Perform reprojection verification on the selected and matched crater pairs and estimate the current pose.

2. The crater identification method for landing visual navigation according to claim 1, characterized in that, Step S230 includes: Step S231: Set a status identifier for each crater in the crater table and initialize it before matching and identification by setting the status of each crater to 0. Step S232, scan the set Set the state of all crater pairs in the set to 1, and record the index j of the other crater that forms a crater pair with it. Step S233, scan the set If the state of the corresponding crater is not 0, then set the state of the crater to 2 and record the number of the other crater that forms a crater pair with it. ; Step S234: Scan the state table. If the crater state is 2, then determine the recorded ( Does it belong to the set? If so, then store in In the collection.

3. The crater identification method for landing visual navigation according to claim 1, characterized in that, In step S300, based on the camera's perspective projection model, the world coordinate system of the crater is transformed to the pixel coordinates of the image. The crater identification algorithm estimates the correspondence between the crater observed on the image plane and the crater in the world coordinate system. The LIS identification algorithm outputs the identification results of multiple craters and the crater that constitutes the last pyramid. The identification set is based on the 3D coordinates of multiple crater centers in the world coordinate system and the 2D coordinates in the image coordinate system, as well as the camera intrinsic parameter matrix. The transformation matrix from the world coordinate system to the camera coordinate system is estimated, and the craters are traversed. The pose is obtained by solving the set; For each possible impact crater First, determine if the crater could be within the same field of view as the previously identified craters. If so, solve for the rotation matrix and translation vector. Then, transform all craters to the image coordinate system to obtain a reprojected image. Compare the corresponding distances of the craters in the reprojected image and the ground truth image. If the Gaussian angular distance is less than 1.57, it indicates that the current pose is correct, meaning the current crater is within the correct range. For correct identification.

4. A crater identification system for landing visual navigation, the system performing the crater identification method for landing visual navigation as described in claim 1, characterized in that, include: The database creation module is used to create a crater pair database: select an appropriate number of craters to create the database, and based on the selection criteria, remove small craters and overlapping craters to obtain a crater pair table; The filtering module is used to filter and match the meteorite crater database, including: image preprocessing; The system filters impact crater pairs based on dynamic thresholds; constructs triples for the selected impact crater pairs based on a pyramid-based triplet search algorithm; and performs iterative pyramid matching on the impact crater pairs after triplet construction using a pyramid-based LIS recognition algorithm. Verification module: Used to perform reprojection verification on the selected and matched crater pairs.

5. An electronic device, characterized in that, The electronic device includes: At least one processor; and, A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the method of any one of claims 1-3.

6. A non-transitory computer-readable storage medium storing computer instructions for causing the computer to perform the method of any one of claims 1-3.

Images