A triangle-based star pattern registration method and system

Abstract

The application discloses a kind of star map registration method and system based on triangle, which comprises: star map is preprocessed, and star map foreground layer is obtained;Star selection algorithm is used to star selection to star map foreground layer in sub-region, and star target point is obtained;Based on triangle distance constraint, same name star point registration is carried out to star target point, and same name star point registration matrix is obtained;Based on same name star point registration matrix, star map motion background compensation is carried out using affine transformation, and star map registration is completed.The system comprises: preprocessing module, selection module, registration module, compensation module.The scheme greatly improves the accuracy of star map registration on the basis of reducing the complexity of calculation, and is suitable for satellite image processing field.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of satellite image processing, in particular to a triangle-based star map registration method and system. BACKGROUND

[0002] Star map registration is a process of precise matching of star points in the same starry sky background area for images obtained by multiple sensors at different time periods and different imaging angles, which have geometric deformation. At present, star map registration is the most critical technology for the key link of background star suppression by difference method. The matching accuracy determines the suppression effect of background stars, and further determines the detection effect of optical image spatial debris. Therefore, high-precision star map registration is the basis for ensuring that weak and small spatial debris are effectively detected.

[0003] At present, existing same-name star point registration methods include ICP algorithm and Delauney triangulation method. The ICP algorithm has large amount of calculation and is suitable for registration of a small number of same-name star points. Although the Delauney triangulation method can reduce the number of reference triangles and thus reduce the amount of calculation, for a small number of star points, this algorithm greatly reduces the success rate of same-name star point registration. SUMMARY

[0004] To solve the above technical problems, the purpose of the present application is to provide a triangle-based star map registration method and system, which can improve registration accuracy on the basis of small amount of calculation.

[0005] The first technical solution adopted by the present application is: a triangle-based star map registration method, comprising the following steps:

[0006] Pretreating the star map to obtain a star map foreground layer;

[0007] Selecting stars in the star map foreground layer by using a regional star selection algorithm to obtain star target points;

[0008] Registering same-name star points based on the star target points by using a triangle distance constraint to obtain a same-name star point registration matrix;

[0009] Based on the same-name star point registration matrix, performing star map motion background compensation by using an affine transformation to complete star map registration.

[0010] Further, the step of selecting stars by using a regional star selection algorithm to obtain star target points specifically comprises:

[0011] Scanning the selected region in the regional star map foreground layer in order, and marking the pixel point serial number according to the gray value;

[0012] Reorder all the marked serial numbers, fill in the missing serial numbers in the middle, form a set of continuous marked serial numbers, each of which represents a connected domain;

[0013] Calculate the number of pixels and the maximum gray value of each connected domain;

[0014] Select the determined star according to the pixel number standard and the maximum gray value standard of each connected domain;

[0015] Select the star target point from the determined star.

[0016] This process effectively avoids the distribution of selected stars being too concentrated, which helps to improve the star map registration accuracy.

[0017] Further, the step of scanning the selected region in the region-based star map foreground layer according to the gray value and marking the pixel point serial number is specifically:

[0018] From top to bottom, from left to right, scan the selected region in the region-based star map foreground layer;

[0019] If the pixel gray value is 0, continue to scan backward, if the pixel gray value is 1, mark the pixel.

[0020] Further, the step of scanning the selected region in the region-based star map foreground layer according to the gray value and marking the pixel point serial number, specifically includes:

[0021] If only one of the upper point and the left point is marked, the pixel point is given the same mark as the same mark;

[0022] If the upper point and the left point are marked and the marks are the same, the pixel point is given the same mark;

[0023] If the upper point and the left point are marked and the marks are different, the pixel point is given the same mark as the upper point, and the marks of the upper point and the left point are put into the equivalence table as equivalent marks;

[0024] If the upper point and the left point are not marked, the pixel point is given a new mark.

[0025] Further, the step of scanning the selected region in the region-based star map foreground layer according to the gray value and marking the pixel point serial number further includes:

[0026] Merge the equivalent marks in the equivalence table, and unify the equivalent marks to the mark with the smallest serial number in the equivalence table.

[0027] Further, the step of registering the same name star point based on the triangular distance constraint to the star target point to obtain the same name star point registration matrix, specifically includes:

[0028] The clue matrix is constructed by using the edge length of the star triangle and the vertex serial number of the star triangle;

[0029] The triangle registration is performed by using the edge length of the triangle in the clue matrix, and the triangle with the equal edge length is obtained;

[0030] The star point registration matrix is obtained by using the triangle with the equal edge length to perform the triangle vertex registration.

[0031] This step performs twice registration by using the star triangle, first performs the registration by using the edge length of the star triangle, and then performs the star triangle vertex registration by using the two edges to determine the vertex, so that the calculation amount is reduced under the condition of ensuring the accuracy, and the algorithm complexity is reduced.

[0032] Further, the star map registration is completed by using the affine transformation to perform the star map motion background compensation based on the star point registration matrix, and the star map registration specifically comprises the following steps:

[0033] The global motion parameter is calculated based on the affine transformation model and the star point registration matrix;

[0034] The background pixel point coordinates of the star map to be registered are calculated by using the affine transformation model based on the global motion parameter;

[0035] The background pixel point coordinates of the star map to be registered are split into a decimal part and an integer part;

[0036] The background pixel point coordinates of the star map to be registered after the splitting are processed by using the bilinear difference method, and the star map motion background compensation is performed.

[0037] This step selects one frame of star map as a reference, and corrects the pixel point coordinates of the star map of the subsequent frames to the reference frame, but since the pixel point coordinates of the reference star map are integers, the pixel point coordinates are generally changed into decimals after the rotation and translation transformation, and there is no gray scale of the decimal coordinate point in the star map to be registered, so the bilinear interpolation method is selected for the motion compensation.

[0038] The second technical scheme adopted by the present application is a star map registration system based on a triangle, comprising:

[0039] A preprocessing module is used to process the star map to obtain a star map foreground layer;

[0040] A selection module is used to scan the divided star map foreground layer in a region, and select the star target point;

[0041] A registration module is used to perform star triangle registration and star triangle vertex registration on the star target point;

[0042] The compensation module is used for calculating the global motion parameter of the homonymic star point and performing motion compensation on the star background map.

[0043] The method and system have the following advantages: firstly, the star map is preprocessed to obtain a star map foreground layer, which facilitates subsequent star selection; secondly, the star map foreground layer is divided into regions, which can avoid the selected stars being too concentrated and improve the star map registration accuracy; thirdly, the star target points are selected after scanning and marking; fourthly, the star target points are registered based on the triangle distance constraint, which can reduce the calculation complexity while ensuring the accuracy; and finally, the motion background compensation is performed, that is, the background registration is performed in addition to the star point registration, which further improves the registration accuracy. BRIEF DESCRIPTION OF DRAWINGS

[0044] Figure 1 is a flowchart of a star map registration method based on a triangle according to the present application;

[0045] Figure 2 is a structural block diagram of a star map registration system based on a triangle according to the present application. DETAILED DESCRIPTION

[0046] Embodiment One

[0047] The present application will be further described in detail below in combination with the drawings and specific embodiments. For the step numbers in the following embodiments, only the order of the steps is set for the purpose of description, and the order between the steps is not limited in any way, and the execution order of each step in the embodiments can be adaptively adjusted according to the understanding of those skilled in the art.

[0048] As shown in Figure 1 , the present application provides a star map registration method based on a triangle, which comprises the following steps:

[0049] Step S1, preprocessing the star map to obtain a star map foreground layer;

[0050] The star map is preprocessed to obtain a black and white template, wherein the white template represents the foreground layer and the black template represents the background layer. Most of the foreground layer is stars, but it is not certain that all of them are stars, so star selection needs to be performed on the foreground layer.

[0051] Step S2, performing star selection on the star map foreground layer by using a regional star selection algorithm to obtain star target points;

[0052] This step first divides the star map foreground layer into regions to obtain a plurality of small regions, and preferably divides the star map foreground layer into 3x3 nine rectangular regions, and performs star point selection on each rectangular region. The specific process is as follows:

[0053] First, from top to bottom, from left to right, the selected region in the region star map foreground layer is scanned in turn, and the pixel point with the gray value of 0 is scanned backward, and no marking is performed. If the pixel point with the gray value of 1 is marked, the specific marking rule is as follows:

[0054] If only one of the upper point and the left point is marked, the pixel point is assigned the same mark as the marked pixel point;

[0055] If the upper point and the left point are both marked and the marks are the same, the pixel point is assigned the same mark;

[0056] If the upper point and the left point are both marked and the marks are different, the pixel point is assigned the same mark as the upper point, and the marks of the upper point and the left point are put into the equivalence table as equivalent marks;

[0057] If the upper point and the left point are both unmarked, the pixel point is assigned a new mark.

[0058] Secondly, the equivalent marks in the equivalence table are merged, and the equivalent marks are unified into the mark with the smallest serial number in the equivalence table;

[0059] Secondly, the serial numbers of all the marks are rearranged, and the missing serial numbers in the middle are supplemented to form a set of continuous mark serial numbers, each of which represents a connected domain;

[0060] Secondly, the number of pixel points contained in each connected domain is calculated and recorded as S i ;

[0061] Secondly, the maximum gray value of each connected domain is calculated and recorded as M i ;

[0062] Secondly, the selected stars are selected according to the selection criteria of the number of pixel points contained in each connected domain and the maximum gray value, and the selection criteria are as follows:

[0063]

[0064] Among them, represents the mean value of the gray value of the whole image, represents the variance of the gray value of the whole image. The selected star set V' is recorded as V';

[0065] Finally, the selected stars in V' are sorted from large to small, and the first 6 largest stars are selected.

[0066] Step S3, based on the triangular distance constraint, the same name star point registration is performed on the star target point, and a same name star point registration matrix is obtained;

[0067] In the selection of stars, the area information of star points is used to screen them, and the position information of star points is used to label them. However, in the case of movement of observation equipment, the position information and area size of star points will change. Therefore, it is not accurate to register the same-named star points only according to the area information and position information of a single star point. However, in the movement process of observation equipment, the relative positions between star points are basically unchanged within a certain range of movement. Considering this, the same-named star points can be registered by using the triangles formed by star points. The specific steps are as follows:

[0068] Firstly, a clue matrix is constructed by using the side length and vertex number of the star triangle composed of star target points.

[0069] After the selection of regional star points and the extraction of the center of mass, N * star target points and their center of mass coordinates to be registered are selected in each small region. Preferably, in the simulation, each frame of star map is divided into 9 small regions according to the format of 3x3, N * =6 is set for each region, and the registration of the same-named star points is performed in a regional manner, that is, the registration of point sets between maps is converted into the registration between corresponding regions, so as to ensure the registration efficiency.

[0070] Let the center of mass coordinates of the 6 star target points in the region be C i , where i=1, 2, …6 represents the number of star target points, and let the set of all triangles composed of the 6 star target points be I, which is specifically represented as:

[0071]

[0072] Where Δ i,j,k represents a triangle with star target points numbered i, j, and k as vertices, which is called a reference triangle. The number of reference triangles in I is n, and it is easy to know that:

[0073]

[0074] The feature information of the reference triangle is extracted, including the three side lengths and three vertices of the reference triangle. A clue matrix of the reference triangle is constructed, which is denoted as Z. One region corresponds to one clue matrix, and the clue matrix Z is divided into 20 rows. Each row records the side length information and vertex information of a reference triangle. The specific format is as follows:

[0075]

[0076] Where l ij is the distance between the i-th star target point and the j-th star target point.

[0077] The feature information of the reference triangle of the corresponding region of the adjacent frame star map is extracted, including the three side lengths and three vertices of the reference triangle of the corresponding region, a reference triangle clue matrix of the corresponding region is constructed, and the reference triangle clue matrix of the corresponding region is denoted as Z'.

[0078]

[0079] Wherein, l i'j' is the distance between the star target point of the i' corresponding region and the star target point of the j' corresponding region.

[0080] Secondly, the triangle registration is performed by using the side length of the triangle in the clue matrix, and the triangle with equal side length is obtained.

[0081] The first three columns of side length information in Z and Z' are extracted, and are rearranged in the order from small to large, to obtain the side length information matrix L and L' of the reference triangle formed by the star target points of the corresponding regions of the adjacent two frames of star maps, and the specific expression is:

[0082]

[0083] The reference triangle Δ i,j,k in the first row of the matrix L is paired, the difference between the three side lengths in all reference triangles in the matrix L' and the three side lengths of Δ i,j,k is calculated respectively, a difference matrix D is constructed, and the specific expression is:

[0084]

[0085] The minimum value in the difference matrix is obtained, denoted as d min , and the specific expression is:

[0086]

[0087] Wherein, i', j', k' are the star target point numbers of the corresponding regions of the adjacent frames of star maps.

[0088] Whether d min reaches the standard is judged. Preferably, the change amount of the relative distance between stars less than 1% is defined as the judgment standard, and the specific expression is:

[0089]

[0090] Wherein, Δ i'j'k' : Δ 123 indicates that the pairing is successful, and []: Δ 123 indicates that the pairing is unsuccessful.

[0091] Finally, the same-named star point matching matrix is obtained by using the triangle vertex matching of the triangles with equal side length.

[0092] When the triangle Δ i'j'k' is matched with the triangle Δ 123 , the matching of the triangle vertices is continued. i'j' , l i'k' , l i'k' and l 12 , l 13 , l 23 are in one-to-one correspondence, and the intersection of l i'j' and l i'k' is i', and the intersection of l 13 is 1, so i' corresponds to 1, and correspondingly, j' corresponds to 2, and k' corresponds to 3. The star numbers of the same-named point pairs are input into the matching matrix P, that is,

[0093]

[0094] When all the triangles in L are traversed, that is, all the reference triangles in L are matched, the repeated same-named star point matching numbers in the matching matrix are deleted, and the final same-named star point matching matrix P is obtained.

[0095] In step S4, based on the same-named star point matching matrix, the star map motion background compensation is performed by using the affine transformation, and the star map matching is completed.

[0096] The star map motion background compensation refers to matching the star map background to the reference frame, that is, in the matching process, a frame of star map is selected as the reference frame, the pixel point coordinates of the subsequent frames of star map are corrected to the reference frame by using the global motion parameters of the star map, then the pixel point coordinates in the star map to be matched are solved by using the affine transformation model, and finally the star map background matching is performed by using the bilinear difference method. The specific process is as follows.

[0097] Firstly, the global motion parameters are calculated based on the affine transformation model and the same-named star point matching matrix.

[0098] Before the star map motion background compensation, the motion parameters of the star points in the front and rear frames of star map, that is, the motion parameters of the rotation and translation of the star points, need to be known. The two kinds of parameters are solved by using the affine motion transformation model, and the model structure is as follows.

[0099]

[0100] In the formula, R is the parameter for describing the rotation motion of the star map, and is specifically as follows.

[0101]

[0102] Wherein, T is a parameter describing the translation movement of the star map, specifically:

[0103]

[0104] Wherein, (x, y) is the coordinate of a star point in the previous frame of star map, (x', y') is the coordinate of the star point in the converted next frame of star map. Equation (11) is expressed in the form of equation group:

[0105]

[0106] From equation (14), it can be seen that this equation contains 6 unknowns, and at least 6 equations are needed to solve it. In order to ensure the accuracy of the solution, more equations are generally needed to participate in the calculation, that is, more than 3 pairs of star point pairs are needed to participate in the calculation, so the least square method is used to solve it. Since the structure of the two equations in equation (14) is the same, the same solving method can be used to solve it. In the solving process, the first equation can be solved first, and the second equation can be solved by the same method, which requires the same star point pairs in each region. Here, all the same star point pairs in each region are combined to obtain a total same star point pair matrix, denoted as M, wherein n is the number of all same star point pairs, specifically:

[0107]

[0108] According to equation (14), we have:

[0109] X = HY (16)

[0110] Wherein, the coefficient matrix is:

[0111]

[0112] The known vector X is:

[0113] X = [x'1 … x' n ] T (18)

[0114] Wherein, the solved parameter Y is:

[0115] Y = [a1 b1 c1] T (19)

[0116] From equation (16), we have:

[0117] Y = (H T H) -1 H T X (20)

[0118] According to formula (20), the motion parameters a1, b1 and c1 can be solved, and the motion parameters a2, b2 and c2 can be solved in the same way, so that the global motion parameters R and T of the star map are obtained.

[0119] Secondly, based on the global motion parameters, the background pixel point coordinates corresponding to the star map to be registered frame after the affine transformation model is calculated.

[0120] The affine transformation model used is:

[0121]

[0122] Wherein, the coordinate (x', y') is the background pixel point coordinate of the star map to be registered frame, and (x, y) is the background pixel point coordinate of the reference frame.

[0123] Secondly, the background pixel point coordinates of the star map to be registered frame are divided into integer part and decimal part; specifically expressed as:

[0124]

[0125] Wherein, x1 and y1 represent the integer part, and x2 and y2 represent the decimal part.

[0126] Finally, the bilinear difference method is used to process the split star map to be registered frame background pixel point coordinates, and the star map motion background motion compensation is carried out. Specifically:

[0127] Let the gray value of the star map to be registered at (x', y') be f(x', y'), and the gray value at (x, y) after registration be f'(x, y), then:

[0128] f'(x, y) = f(x', y') = (1-x2)(1-y2)f(x1, y1) + x2(1-y2)f(x1+1, y1) + (1-x2)y2f(x1, y1+1) + x2y2f(x1+1, y1+1)

[0129] As shown in Figure 2 A star map registration system based on triangle, comprising:

[0130] A preprocessing module for processing the star map to obtain the star map foreground layer;

[0131] A selection module for scanning the divided star map foreground layer by region and selecting the star target point;

[0132] A registration module for star triangle registration and star triangle vertex registration of the star target point;

[0133] The compensation module is used to calculate the global motion parameters of the same name star point and to perform motion compensation on the star background map.

[0134] Embodiment two

[0135] On the basis of the above embodiments, considering the scanning of the sub-regional star map foreground layer in the star selection stage, the scanning manner can also be from top to bottom and from right to left. If the pixel point has a gray value of 0, continue to scan downward without marking. If the pixel point has a gray value of 1, mark it. The specific marking rules are as follows:

[0136] If only one of the upper point and the right point is marked, the pixel point is assigned the same mark as the marked pixel point.

[0137] If both the upper point and the right point are marked and the marks are the same, the pixel point is assigned the same mark.

[0138] If both the upper point and the right point are marked and the marks are different, the pixel point is assigned the same mark as the upper point, and the marks of the upper point and the right point are put into the equivalence table as equivalent marks.

[0139] If neither the upper point nor the right point is marked, the pixel point is assigned a new mark.

[0140] This embodiment provides another pixel point scanning manner and marking manner.

[0141] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, but not to limit them. Although the present application has been described in detail with reference to the above embodiments, those skilled in the art should understand that they can modify the above embodiments or equivalently replace some or all of the technical features, and these modifications and replacements do not make the present application exceed the scope of the technical solutions described in the above embodiments.

Claims

1. A star map registration method based on triangles, characterized in that, Includes the following steps: Step S1: Preprocess the star map to obtain the foreground layer of the star map; Step S2: Use a regional star selection algorithm to select stars in the foreground layer of the star map to obtain star target points; Step S3: Based on the triangular distance constraint, register the target star points with corresponding star points to obtain the corresponding star point registration matrix; Step S4: Based on the registration matrix of the same star points, perform star map motion background compensation by affine transformation to complete star map registration; Record the star chart to be registered in The gray value at that location is After registration The gray value at that location is Then we have: in, and Represents the integer part. and Indicates the decimal part; Specifically, it includes: Step S21: Divide the star map foreground layer to obtain a regional star map foreground layer; Step S22: Scan the selected areas in the foreground layer of the regional star map in sequence, and mark the pixel number according to the gray value; Step S23: Rearrange all the marker numbers, fill in the missing numbers in the middle, and form a set of consecutive marker numbers, where each marker number represents a connected component. Step S24: Calculate the number of pixels and the maximum gray value of each connected component; Step S25: Select the star target points according to the pixel count standard and the maximum gray value standard of each connected component; Specifically, it includes: Step S31: Construct a clue matrix using the side lengths and vertex indices of the stellar triangle formed by the stellar target points; Step S32: Use the side lengths of the triangles in the clue matrix to perform triangle registration, and obtain triangles with corresponding equal side lengths; Step S33: Use triangles with corresponding side lengths to perform triangle vertex registration to obtain the same star point registration matrix.

2. The star map registration method based on triangles according to claim 1, characterized in that, The step of sequentially scanning the selected regions in the foreground layer of the regional star map and marking the pixel numbers according to their grayscale values ​​is specifically as follows: Scan the selected regions in the foreground layer of the regional star map sequentially from top to bottom and from left to right; If the grayscale value of a pixel is 0, continue scanning; if the grayscale value of a pixel is 1, mark that pixel.

3. The star map registration method based on triangles according to claim 2, characterized in that, If the grayscale value of a pixel is 0, the scanning continues; if the grayscale value of a pixel is 1, the marking rule for that pixel is: If the top point and the left point have only one label, then assign the same label to that pixel. If both the top and left points are marked and have the same marking, then assign the same marking to that pixel. If both the top point and the left point are marked, and the marks are different, then assign the pixel the same mark as the top point, and put the marks of the top point and the left point into the equivalence table as equivalent marks; If neither the top nor the left point is marked, then a new mark is assigned to that pixel.

4. The star map registration method based on triangles according to claim 3, characterized in that, After the step of sequentially scanning the selected regions in the foreground layer of the regional star map and marking pixels according to grayscale values, the following is also included: Merge the equivalence tags in the equivalence table and unify the equivalence tags into the tags with the smallest ordinal numbers in the equivalence table.

5. The star map registration method based on triangles according to claim 1, characterized in that, The step of star map registration, which involves using an affine transformation to compensate for background motion based on a registration matrix of stellar points, specifically includes: Step S41: Calculate global motion parameters based on the affine transformation model and the registration matrix of the same star points; Step S42: Based on global motion parameters, use an affine transformation model to calculate the coordinates of the background pixels of the star map reference frame after transformation, and the coordinates of the background pixels of the star map to be registered frame. Step S43: Split the coordinates of the background pixels of the star map to be registered frame into integer and fractional parts; Step S44: Use bilinear interpolation to process the coordinates of the background pixels of the split star map to be registered frame and perform star map motion background compensation.

6. A star map registration system based on triangles, characterized in that, For performing the triangle-based star map registration method as described in claim 1, the method includes: The preprocessing module is used to preprocess the star map to obtain the foreground layer of the star map; The selection module is used to select stars in the foreground layer of the star map using a regional star selection algorithm to obtain star target points. The registration module is used to register stellar target points with corresponding stellar points based on triangular distance constraints, and to obtain the stellar point registration matrix. The compensation module, based on the registration matrix of the same star points, uses affine transformation to compensate for the motion background of the star map and complete the star map registration.

Images