A fast star pattern recognition method and system based on adaptive star catalog

Abstract

The application discloses a kind of fast star map identification method and system based on adaptive star table, which comprises the following steps: using block, hierarchical extraction method to generate GAIA star table file from original GAIA star table data;GAIA star table file and Tycho2 star table file are read into memory, and star information is stored in partition, and star table index is established;The field of view information of input optical image is input, and according to the input information, the type of star table to be used is selected;According to the pointing and field of view information of input optical image, the required star information is read according to the star table index;Fast sorting method is used to sort image stars and star table stars, and the brightest multiple stars are selected for fast star table matching;Through the voting of matching result, the star mapping with the highest matching accuracy is found out as the successfully identified star.The application is suitable for optical image processing research field, and can be used for fast astronomical positioning of space debris and near-earth asteroids, to gain valuable time for emergency response of space events.

Description

Technical Field

[0001] This invention belongs to the fields of space target monitoring and optical image processing, and specifically relates to a fast star map recognition method and system based on an adaptive star catalog. Background Technology

[0002] To address the hazards of space debris, it is necessary to closely monitor all space targets, acquire their spatial position information in real time, and then issue timely warnings for on-orbit satellites at risk of collision. When observing space targets using optical telescopes, the images generally contain a large number of stars in addition to the space targets themselves. Obtaining the celestial position of space targets typically employs astronomical positioning methods: first, the "identity" information of the stars in the image is identified, and their celestial positions are obtained; then, based on the one-to-one correspondence between the image positions and celestial positions of the stars, a transformation model from image coordinates to celestial coordinates (called the film model) is established; finally, the image position of the space target is input into this model to calculate its celestial position (centered on the observation station, the space object lies on an imaginary celestial sphere, usually represented by right ascension and declination). These three steps are respectively called star map identification, calculation of the astronomical positioning model, and calculation of the celestial position. Among these, there are mature and internationally accepted methods for calculating the astronomical positioning model, such as the WCS model, which also includes algorithms and open-source code for calculating the celestial position. Star map identification also has popular methods such as triangle matching and pyramid matching, but these are generally relatively slow.

[0003] Star map recognition, also known as star catalog matching, involves using a star catalog, which stores information on the celestial positions, proper motions, and brightness of all stars accumulated by astronomers through long-term observation, in tabular form. The latest GAIA DR2 catalog contains over 1 billion stars. The celestial positions of stars change very little in the short term (generally less than 0.1 arcseconds per year, with only about 400 stars changing by 1 arcsecond or more per year, and the fastest-moving known stars changing by no more than 10 arcseconds per year). Therefore, astronomical positioning typically calculates the high-precision celestial positions of space targets based on their relative positions to stars in an image. The first step is to identify the stars in the image and obtain their celestial positions. Usually, the stars in the image are arranged into a certain shape (such as a triangle or quadrilateral) and compared with the polygons formed by stars in the star catalog. If the error is less than a certain threshold, a match is considered successful, and the corresponding stars are identified. Because images often contain a large number of stars at high density, triangle matching and pyramid matching methods require constructing a large number of triangles and polygons, making the matching calculation similar to an exhaustive search, which consumes a significant amount of time. Therefore, the computational speed of star map recognition becomes a bottleneck for rapidly processing optical images.

[0004] Star image recognition can be divided into two steps: first, extracting all stars within the corresponding field of view from the star catalog based on image information, and then constructing polygons for matching. Different types of star catalogs have vastly different star field densities. For example, the Tycho II catalog has a star field density of approximately 150 stars / square degree near its declination of 0 degrees, suitable for large field-of-view telescopes; the GAIA catalog has an average density exceeding 15,000 stars / square degree, suitable for small field-of-view telescopes. Extracting all stars from the GAIA catalog within the field of view would be computationally intensive. From a matching method perspective, the more stars identified, the higher the accuracy of the subsequent film model calculation. Therefore, many astronomical positioning software programs employ exhaustive search methods, resulting in long computation times. However, experimental studies have found that due to the inherent error in calculating the positions of stars in the image, the accuracy of the film model initially increases with the number of stars used, reaching a peak and then slightly decreasing. Therefore, reducing the number of stars to be matched can significantly improve computational speed. Summary of the Invention

[0005] To address the aforementioned problems, this invention provides a rapid star map recognition method and system based on an adaptive star catalog, used to quickly identify stars in an image, obtain the celestial positions of these stars, thereby establishing a film model and calculating the celestial positions of space targets.

[0006] The objective of this invention is achieved through the following technical solution: a fast star map identification method based on an adaptive star catalog, comprising the following steps: Step 1: Generate GAIA star catalog files from the original GAIA star catalog data using a block-based and layered extraction method; Step 2: Read the GAIA star catalog file and Tycho II star catalog file into memory, partition and store the star information, and build the star catalog index; Step 3: Input the field of view information of the optical image, and select the star catalog type to use based on the input information; Step 4: Based on the pointing and field of view information of the input optical image, read the required star information according to the star catalog index; Step 5: Use the quick sorting method to sort the stars in the image and the star catalog, and select the brightest stars for quick star catalog matching; Step 6: Vote based on the matching results to find the star map with the highest matching accuracy, and select it as the successfully identified star.

[0007] Preferably, in step 3, the method for selecting the star catalog includes the following steps: Step 3.1: Obtain the field of view and pixel scale information of the current image, and preset the first threshold; Step 3.2: If the field of view and pixel scale of the current image are greater than the first threshold, select the Dichotomy II star catalog; otherwise, select the GAIA star catalog.

[0008] Preferably, in step 4, the method for reading the required stellar information based on the star catalog index includes the following steps: Step 4.1: Calculate the lower boundary of the right ascension ra of the field of view based on the direction of the field of view center, the declination dec, and the size of the field of view v. min upper boundary of right ascension ra max Declination lower boundary dec min and the upper boundary of declination dec max ; Step 4.2, for those belonging to right ascension ra min -ra max Declination min -dec max The sub-regions within the range are searched one by one; Step 4.3: For each sub-region, find its starting and ending indices in the star catalog array based on the star catalog index; Step 4.4: From the starting position to the ending position of the sub-sky region, determine whether each star is within the field of view and meets the magnitude threshold; Step 4.5: Store the star data of all sub-regions that meet the conditions in the buffer.

[0009] Preferably, in step 5, the star catalog matching method includes the following steps: Step 5.1: Use the quick sorting method to sort the stars in the image and the star catalog, and select the brightest stars for quick star catalog matching; Step 5.2: Based on the telescope's direction, dynamically solve for a simplified conversion model from pixels to angles in the current image; Step 5.3: Based on the dynamic transformation model, calculate the estimated celestial positions of the stars in the image, and then construct a multi-feature triangle that includes the dot product of the long and short sides, the ratio of the long and short sides, and the features of the long side. Step 5.4: First, use the bisection method to search for the ratio of the long side to the short side, quickly find several candidate triangles that are closest to the current triangle to be matched, then determine whether other features meet the conditions, and calculate the matching error.

[0010] Preferably, in step 5.1, a recursive approach is used for quicksort, as follows: Each time, a pivot number is selected, and the array is divided into two parts: elements smaller than the pivot number are placed in the left queue; elements larger than the pivot number are placed in the right queue; the arrays in the left and right queues are processed recursively until the array to be processed is empty, and then the process returns to the previous level.

[0011] Preferably, in step 5.2, solving the simplified transformation model includes the following steps: Step 5.2.1: Obtain the orientation of the image center.az0 , up and down el0 Right Ascension ra0 and declination dec0 ; Step 5.2.2: Initialize the transformation matrix parameters using the pixel scale bar dpp; Step 5.2.3: Set the positions of 5 simulated stars in the image, assuming the image resolution is [resolution value missing]. x_size * y_size Calculate the pixel positions of these 5 simulated stars S0-S5; Step 5.2.4: Set the right ascension and declination celestial sphere position of the simulated star S0 as ( ra0 , dec0 ); Step 5.2.5: Calculate the azimuth and elevation positions of the remaining simulated stars; Step 5.2.6: Based on the station coordinates and observation time, convert the obtained azimuth and elevation positions of the simulated star to obtain its right ascension and declination positions. ra , dec ); Step 5.2.7: Calculate the input and output data of the CD transformation matrix from the pixel positions and right ascension and declination positions of the simulated stars; Step 5.2.8: Calculate the input and output data of 5 sets of CD transformation matrices from 5 simulated stars, input them into the least squares equation solving function, and solve for the CD matrix parameters of the simplified transformation model.

[0012] Preferably, in step 5.3, the method for constructing multiple feature triangles includes at least the following steps: Step 5.3.1: Convert the pixel positions of the stars into celestial angular positions; Step 5.3.2: Calculate the side length of the line connecting any two selected stars; Step 5.3.3: Select any three stars to form a triangle, and determine the longest side, the shortest side, and the middle side; calculate the ratio of the longest side to the shortest side, the ratio of the middle side to the shortest side, and the ratio of the longest side to the middle side; calculate the dot product of the longest side and the middle side. Step 5.3.4: Extract the seven features of the triangle: the longest side, the shortest side, the middle and longest side, the ratio of the longest side length to the shortest side length, the ratio of the middle and longest side length to the shortest side length, the ratio of the longest side length to the middle and longest side length, and the dot product of the longest side and the middle and longest side length. Store the seven features of all triangles into an array.

[0013] Preferably, step 5.4 includes at least the following steps: Step 5.4.1: Sort the star triangles in the image and star catalog from largest to smallest according to the ratio of their longer and shorter sides; Step 5.4.2: Select one triangle from the sorted star catalog stellar triangles and match it with the triangle in the image; Step 5.4.3: For the current triangle to be matched, use the bisection method to find the triangle with the smallest error from the sequence of length-to-short side ratios of triangles in the image; Step 5.4.4: Using the triangle with the smallest ratio error of the longer and shorter sides as a reference, search for triangles with errors less than a specified threshold before and after it, and store them in the buffer. Step 5.4.5: For each triangle found, determine whether other features are less than the threshold, and record the error value of the dot product of the length, middle and side. Step 5.4.6: Select 7 triangles whose feature errors all satisfy the threshold and have the smallest length-middle-side dot product error, and calculate the linear correlation between the triangle and the triangle to be matched. Step 5.4.7: Set a second threshold. If the linear correlation between the two triangles is less than 1, the match is considered successful and the matching result is stored; otherwise, the match fails and the next star table triangle is matched. Step 5.4.8: Perform the above operation on each star table triangle until the number of matched triangles meets a certain requirement.

[0014] Preferably, step 6 includes at least the following steps: Step 6.1: Construct the voting matrix VM and initialize it to 0; Step 6.2: Traverse each matched triangle and vote on the stars it contains; Step 6.3: After voting for each matching triangle, select the star map that meets the voting requirements from the voting matrix; Step 6.4: Save these star maps to an array and sort them from highest to lowest vote count.

[0015] In addition to providing a fast star map identification method based on an adaptive star catalog, this invention further provides a system for implementing the above method, the system comprising: The star catalog generation module generates customized GAIA star catalog files that can be read into memory, based on the original GAIA star catalog data. The star table storage module stores the GAIA star table file and Tycho 2 star table file in memory according to the designed index structure; The star catalog selection module selects the appropriate star catalog type from the star catalog storage module based on the image information. The star catalog extraction module extracts star data that matches the image from the star catalog based on information such as the direction and field of view in the image; The star catalog matching module uses a fast sorting method to sort the stars in the image and the star catalog, and selects the brightest stars for fast star catalog matching.

[0016] The voting module identifies stars in the image based on the matching results and obtains their right ascension and declination information.

[0017] The star catalog generation module generates customized GAIA star catalog files that can be read into memory, based on the original GAIA star catalog data.

[0018] The star table storage module stores the GAIA star table file and Tycho 2 star table file in memory according to the designed index structure.

[0019] The star catalog selection module selects the appropriate star catalog type from the star catalog storage module based on the image information.

[0020] The star catalog extraction module extracts star data that matches the image from the star catalog based on information such as the direction and field of view in the image.

[0021] Select a typical stellar module and use quicksort to find the brightest stars.

[0022] The transformation model solving module dynamically generates a simplified transformation model based on the information in the current image.

[0023] Construct a multi-feature triangle module to generate multi-feature triangles based on a dynamic transformation model.

[0024] The hierarchical fast triangle matching module uses features with high classification efficiency to quickly pair stars in the image with stars in the star catalog.

[0025] The voting module identifies stars in the image based on the triangle matching results and obtains the right ascension and declination information of the stars in the image.

[0026] Compared with the prior art, the present invention has the following advantages: This invention provides a fast star map recognition method and system based on an adaptive star catalog. The adaptive star catalog algorithm is suitable for various types of telescope images and significantly reduces computational load by filtering stars. A multi-feature triangle matching algorithm based on a dynamic transformation model ensures matching accuracy while employing hierarchical fast matching to improve matching speed. The combined efforts of star catalog filtering and star catalog matching facilitate rapid computation for star map recognition.

[0027] The present invention provides a fast star map identification method based on an adaptive star catalog, which is simple in process and computationally efficient. It is applicable to the field of optical image processing research and can be used for rapid astronomical positioning of space debris and near-Earth asteroids, thus gaining valuable time for emergency response to space events. Attached Figure Description

[0028] Figure 1 This is a flowchart of the fast star map recognition method based on an adaptive star catalog in an embodiment of the present invention; Figure 2 This is a custom representation of the GAIA star in an embodiment of the present invention; Figure 3 This is a schematic diagram of the memory index established in an embodiment of the present invention. Detailed Implementation

[0029] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of the embodiments of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0030] like Figure 1 As shown, the technical solution of the present invention provides a fast star map recognition method based on an adaptive star catalog, which improves the star map recognition speed in both star extraction and star matching, including the following steps: Step 1: Using a block-based and layer-based extraction method, generate the GAIA star catalog file from the original GAIA star catalog data; Step 2: Read the GAIA star catalog file and Tycho II star catalog file into memory, store the star information in partitions, and create a star catalog index to facilitate quick retrieval of the required star data based on the index later. Step 3: Based on the field of view information of the input optical image, determine the type of star catalog to use (GAIA catalog or Tycho catalog). Step 4: Based on the pointing and field of view information of the input optical image, read the required star information according to the star catalog index; Step 5: Use the quick sorting method to sort the stars in the image and the star catalog, and select the brightest stars for quick star catalog matching; Step 6: Vote based on the matching results to find the star map with the highest matching accuracy, and select it as the successfully identified star.

[0031] In this invention, the star catalog includes multiple star catalogs, including the GAIA star catalog; in step 1, the block and layer extraction method greatly reduces the amount of GAIA star catalog data used; and the memory indexing method greatly improves the efficiency of star catalog data reading. In one embodiment of the present invention, step 1, the step of customizing the star table file, is as follows: The GAIA star catalog is very large, requiring customized catalog files for typical stars, such as... Figure 2As shown, it includes the following sub-steps: (1) Divide the entire sky area into several partitions at certain intervals, such as 30 degrees, 90 degrees, and 180 degrees. The specific partition size is determined based on the size of the generated partition file, with the maximum file size not exceeding 4GB (some file systems do not support files larger than 4GB).

[0032] (2) Divide the current partition into several blocks at intervals of 0.1 degrees. If the telescope's field of view is less than 0.1 degrees, even smaller blocks can be set.

[0033] (3) For the current block, read all star information from the original GAIA star catalog file according to the right ascension and declination range.

[0034] (4) Extract the five columns of data from the extracted star information according to the fields: right ascension, declination, magnitude, right ascension proper motion, and declination proper motion.

[0035] (5) Sort the extracted star data according to magnitude and select the 5, 20 and 40 brightest stars.

[0036] (6) Perform the above operations on all partitions and blocks to finally generate multiple star table files (selecting the 5 brightest stars will generate 1 sky area file star_file0, selecting the 20 brightest stars will generate 4 files star_file1-star_file4, and selecting the 40 brightest stars will generate 8 files star_file11-star_file18).

[0037] In one embodiment of the present invention, step 2 involves establishing a memory index as follows: Store the star catalog data in memory and create indexes (e.g.) Figure 3 As shown in the diagram, this can significantly improve the reading speed of stars in the star catalog. It consists of the following sub-steps.

[0038] (1) Create a star table structure array star_table, each structure contains the right ascension, declination, magnitude, right ascension proper motion, and declination proper motion fields of a star.

[0039] (2) Create two-dimensional index arrays start_index and end_index to record the starting and ending positions of star data in star_table for each sub-partition. The entire sky is divided into 360*180 sub-partitions, of which 360 are right ascension sub-partitions (corresponding to the first dimension of the array) and 180 are declination sub-partitions (corresponding to the second dimension of the array).

[0040] (3) Read all the data in the star table file and temporarily store it in the temporary structure array temp_star_table.

[0041] (4) According to the right ascension from 0 to 360 degrees and the declination from -90 to 90 degrees, in units of 1 degree, find the star data belonging to the sub-region in temp_star_table, copy it to star_table, and record the number of stars in the sub-region. At the same time, update the index subscript of the sub-region in start_index and end_index.

[0042] (5) Store the contents of star_table into the hard disk file star_map.smp, ​​and store the contents of index start_index and end_index into the index file star_index.sta.

[0043] In one embodiment of the present invention, in step 3, the Tycho II star catalog contains more than 2.5 million stars, and the data can all be stored in memory. Under certain conditions, this star catalog is used preferentially for star image identification. The conditions for using the Tycho II star catalog are: low star field density and a sufficiently large field of view. Star field density can be simply distinguished by the pixel scale of the image, or by the ratio of the number of stars extracted from the image to the size of the field of view (its value is affected by the telescope's movement speed, observation exposure time, weather, etc., and fluctuates considerably). The star catalog selection operation is divided into the following sub-steps: Step 3.1: Obtain the field of view and pixel scale information of the current image, and preset the first threshold; Step 3.2: If the field of view and pixel scale of the current image are greater than the first threshold, then the Tycho II star catalog is selected; otherwise, the GAIA star catalog is selected. In this embodiment, if the field of view is greater than 2°×2° and the pixel scale is greater than 2" / pixel, the Tycho II star catalog is selected; otherwise, the GAIA star catalog is used. If Tycho 2 star catalog is used, the astronomical positioning program reads the corresponding star_map.smp file and index file when it starts. If the GAIA star catalog is used, the astronomical positioning program first reads the star_map.smp file and index file corresponding to star_file0 when it starts. If astronomical positioning fails using the star_file0 star catalog, select star_file1-star_file18 star catalog files to retry astronomical positioning based on the current image center direction.

[0044] When processing images from a single telescope, the field of view and pixel scale are fixed, and therefore the star catalog used is also fixed. However, when processing images from multiple telescopes or using multiple GAIA star catalog files, to avoid system delays caused by repeated switching of catalog files, multiple astronomical positioning programs can be launched simultaneously (or a distributed computing approach can be used), distributing optical image processing tasks to the appropriate processing processes for astronomical positioning.

[0045] In one embodiment of the present invention, step 4 involves rapidly filtering typical stellar information from the memory star catalog based on the direction of the field of view center and the size of the field of view for the input optical image, including the following steps: Step 4.1: Calculate the lower boundary of the right ascension ra of the field of view, pointing from the center of the field of view to the right ascension ra, declination dec, and field size v. min upper boundary of right ascension ra max Declination lower boundary dec min upper boundary of declination (dec) max The formula is as follows:

[0046]

[0047]

[0048]

[0049] Step 4.2, for those belonging to right ascension ra min ~ra max Declination min ~dec max The range of sub-heavens, that is, the set of right ascension sub-heavens, is [ The set of declination sub-regions is [ Search one by one; Step 4.3: For each sub-region, find its starting and ending indices in the star table array star_table based on the star table index; Step 4.4: From the starting position to the ending position of the sub-sky region, determine whether each star is within the field of view and meets the magnitude threshold; Step 4.5: Store the star data of all sub-regions that meet the conditions in the buffer. This allows for quick filtering to obtain the brightest and most abundant star data in the star catalog.

[0050] In one embodiment of the present invention, in step 5, regarding star matching, the brightest N (20 or 40) stars are selected to construct a multi-feature triangle based on a dynamic transformation model. Then, a fast matching method based on binary search is used to quickly identify stars in the image. In this embodiment, typical stars are first selected, and the stars in the image and the star catalog are quickly sorted by brightness, with the brightest N stars selected for each. Generally, the 20 brightest stars in the star catalog are selected and matched with the 40 brightest stars in the image, and in most cases, the matching is successful (the brightness estimation of the image stars is not very accurate, so there is some redundancy in the selection). For a few images with dense star fields, the 40 brightest stars in the star catalog can be selected for matching.

[0051] In this embodiment, a quick sorting method is used to sort the stars in the image and the star catalog, and the method for selecting the brightest stars for quick star catalog matching is as follows: Step 5.1: The quick sorting method is used to sort the stars in the image and the star catalog. The brightest stars are selected for quick star catalog matching, which improves the matching speed while ensuring matching accuracy. Step 5.2: Based on the telescope's direction, dynamically solve the simplified conversion model from pixels to angles in the current image. This dynamic conversion model effectively improves the success rate of star map recognition by altazimuth telescopes. Step 5.3: Based on the dynamic transformation model, calculate the estimated celestial positions of the stars in the image, and then construct a multi-feature triangle that includes the dot product of the long and short sides, the ratio of the long and short sides, and the features of the long side. Step 5.4: First, use the bisection method to search for the ratio of the long side to the short side, quickly find several candidate triangles that are closest to the current triangle to be matched, then determine whether other features meet the conditions, and calculate the matching error.

[0052] In one embodiment of the present invention, in step 5.1, the stars in the star catalog are recursively sorted in ascending order of magnitude, as follows: The quicksort algorithm uses a recursive approach. Each time, a pivot element is selected, and the array is divided into two parts: elements smaller than the pivot are placed in the left queue, and elements larger than the pivot are placed in the right queue. The left and right queues are then processed recursively until the array to be processed is empty, at which point the algorithm returns to the previous level. Specifically, it includes the following sub-steps: (1) Quickly sort the stars in the star catalog according to their magnitude from smallest to largest.

[0053] The quicksort algorithm uses a recursive approach. Each time, a pivot element is selected, and the array is divided into two parts: elements smaller than the pivot are placed in the left queue, and elements larger than the pivot are placed in the right queue. The left and right queues are then processed recursively until the array to be processed is empty, at which point the algorithm returns to the previous level.

[0054] (2) Select the first N1 stars and put them into the buffer zone.

[0055] (3) Quickly sort the stars in the image according to their magnitude from smallest to largest.

[0056] (4) Select the first N2 stars and put them into the buffer zone.

[0057] In one embodiment of the present invention, in step 5.2, the positions of stars in the image and the positions of stars in the star catalog need to be converted to the same coordinate system. Stars in the star catalog are represented using celestial coordinates; therefore, the stars in the image can be quickly converted to celestial coordinates using a simplified conversion model (relative to the unobtained film model), and then a triangle is constructed for comparison. This simplified conversion model sets the reference point as the image center, considers image rotation and scaling, and has a total of four parameters. For equatorial telescopes, the model parameters are relatively fixed, and the set values ​​are directly returned. However, for altazimuth telescopes, the conversion model parameters differ greatly for different celestial regions; therefore, it is necessary to dynamically solve the model parameters based on the current telescope pointing direction. In this embodiment, solving the simplified conversion model includes the following steps: Step 5.2.1: Obtain the orientation of the image center. az0 , up and down el0 Right Ascension ra0 and declination dec0 ; Step 5.2.2: Initialize the transformation matrix parameters using the pixel scale bar dpp. The transformation matrix CM is initialized as follows:

[0058] Step 5.2.3: Set the positions of 5 simulated stars in the image, assuming the image resolution is [resolution value missing]. x_size * y_size The pixel positions of these five simulated stars S0-S5 are calculated as follows:

[0059]

[0060]

[0061]

[0062]

[0063] Step 5.2.4: Set the right ascension and declination celestial sphere position of the simulated star S0 as ( ra0 , dec0 ); Step 5.2.5: Calculate the azimuth and elevation positions of the remaining simulated stars; assuming the pixel position of a certain simulated star is... Calculate its azimuth and elevation positions using the formula below. :

[0064]

[0065]

[0066]

[0067]

[0068]

[0069]

[0070]

[0071]

[0072] Step 5.2.6: Based on the station coordinates and observation time, convert the obtained azimuth and elevation positions of the simulated star to obtain its right ascension and declination positions. ra , dec For coordinate transformation methods, please refer to relevant astronomy textbooks; Step 5.2.7: Using the pixel positions and right ascension / declination positions of the simulated stars, calculate the input and output data of the CD transformation matrix according to the following formula: In this embodiment, the input is: dx , dy (See the formula above for the calculation method).

[0073] The output is: and .

[0074]

[0075]

[0076]

[0077]

[0078] Step 5.2.8: Five sets of input and output data for the CD transformation matrix are obtained from five simulated stars. These data are then input into the least squares equation solving function (see relevant linear algebra tutorials) to solve for the CD matrix parameters of the simplified transformation model, as shown below:

[0079] In one embodiment of the present invention, in step 5.3, the features used for triangle matching typically include side length, included angle, and combinations of side length and included angle. Using side length as a feature can remove most of the false matches caused by similar triangles. The triangle features we use include the dot product of the long and short sides, the ratio of the long and short sides, and features of the long side. In this embodiment, the method for constructing multi-feature triangles includes at least the following steps: Step 5.3.1: If the input is an image of stars, the pixel positions of the stars need to be calculated using the following formula ( px , py Convert to celestial angular position ( ra , dec ):

[0080]

[0081]

[0082]

[0083]

[0084]

[0085]

[0086]

[0087]

[0088] Step 5.3.2: Calculate the length L of the line connecting any two selected stars i and j, using the following formula, where the celestial position of star i is... The celestial position of star j is :

[0089] Step 5.3.3: Form a triangle with any three stars and calculate the characteristics of the triangle. By comparing the lengths of the three sides, determine the longest side, the shortest side, and the medium-longest side. Then calculate the ratio of the longest side to the shortest side (referred to as the length-to-short side ratio), the ratio of the medium-longest side to the shortest side, and the ratio of the longest side to the medium-longest side. Calculate the dot product of the longest side and the medium-longest side using the following formula. Where A, B, and C are the three vertices of the triangle, CB is the longest side, and CA is the medium-longest side:

[0090] Step 5.3.4: Extract the seven features of the triangle: the longest side, the shortest side, the middle and longest side, the ratio of the longest side length to the shortest side length, the ratio of the middle and longest side length to the shortest side length, the ratio of the longest side length to the middle and longest side length, and the dot product of the longest side and the middle and longest side length. Store the seven features of all triangles into an array.

[0091] In one embodiment of the present invention, in step 5.4, due to certain errors in the calculation of the positions of stars in the image, as well as factors such as star density, errors may occur when matching the star triangles in the image and the star triangles in the star catalog. Here, a hierarchical matching method is adopted, prioritizing the use of features with strong classification capabilities for matching, selecting several candidate triangles that meet the threshold conditions, and then judging whether each feature meets the matching conditions. In this embodiment, the matching process includes at least the following steps: Step 5.4.1: Quickly sort the star triangles in the image and star catalog according to the ratio of their longer and shorter sides from largest to smallest; Step 5.4.2: Select a triangle from the sorted star catalog triangles and match it with the image triangle; prioritize triangles with a larger ratio of their longer side to their shorter side. Step 5.4.3: For the current triangle to be matched, use the binary search method to find the triangle with the smallest error from the sequence of length-to-short side ratios of triangles in the image. The binary search algorithm is a fast search algorithm for finding a specific element in an ordered array. Its idea is to continuously divide the ordered search table in half, quickly narrowing the search area, and thus finding the target element. See relevant tutorials for the specific algorithm; Step 5.4.4: Using the triangle with the smallest ratio error of its longer and shorter sides as a benchmark, search for triangles with errors less than a specified threshold before and after it, and store them in a buffer. Because triangle matching often involves errors, it is difficult to accurately match based solely on the ratio of the longer and shorter sides; a certain range needs to be defined first. Step 5.4.5: For each triangle found, determine whether other features are less than the threshold, and record the error value of the dot product of the length, middle and side. Step 5.4.6: Select seven triangles whose feature errors all satisfy the threshold and have the smallest length-middle-side dot product error, and calculate their linear correlation with the triangle to be matched. The linear correlation is calculated using the following formula. relate The celestial positions of the three stars in the star catalog triangle are ( ). ), ( ), ( The celestial positions of the three stars in the triangle in the image are ( ). ), ( ), ( ):

[0092]

[0093]

[0094]

[0095] Step 5.4.7: Set a second threshold. If the linear correlation between the two triangles is less than 1, the match is considered successful and the matching result is stored; otherwise, the match fails and the next star table triangle is matched. Step 5.4.8: Perform the above operation on each star table triangle until the number of matched triangles meets a certain requirement (e.g., more than 1000).

[0096] In one embodiment of the present invention, in step 6, based on the triangle matching results, a vote is taken on the star maps corresponding to the triangles, and the star map with the highest number of votes is considered a successfully identified star. In this embodiment, the voting calculation process includes at least the following steps: Step 6.1: Construct the voting matrix VM, initializing it to 0. Assuming there are N images of stars and M catalog stars participating in the matching, the voting matrix is ​​an N-row, M-column two-dimensional array; Step 6.2: Traverse each matched triangle and vote on the stars it contains. Since the storage positions of stars in the triangles are strictly defined according to their shapes, the stars in the image triangles and the star catalog triangles can be matched one-to-one. Assume the three stars in the image triangle are... , , The three stars in the star catalog triangle are , , Then the image of stars Corresponding star catalog stars Image of stars Corresponding star catalog stars Image of stars Corresponding star catalog stars Voting matrix , , Increment the corresponding element by 1 (cast 1 vote); Step 6.3: After each matching triangle has voted, select the star map from the voting matrix that meets certain requirements (e.g., greater than 3 votes); Step 6.4: Save these star maps to an array, sort them by vote count from highest to lowest, and return the star map results. The star map recognition is now complete.

[0097] The present invention will be further described below with reference to specific embodiments.

[0098] Functional and performance experiments were conducted using optical images from the 1.2-meter telescope in Jilin and the 36-cm telescope in Korla, respectively.

[0099] The hardware and software configuration of the test computer is shown in Table 1.

[0100] Table 1. Hardware and software environment for testing.

[0101] Example 1 The Xinjiang 25cm telescope has a field of view of approximately 3.3 degrees and an image resolution of 3056x3056. This telescope is suitable for star chart identification using the Tycho star catalog.

[0102] In terms of functional testing, 678 images were selected, and the star map was successfully recognized in 677 of them, with a success rate of 99.9%.

[0103] In terms of performance testing, the total time for single-threaded computation of 678 images was 1.85 seconds, with an average time of approximately 2.729 milliseconds per image.

[0104] Example 2 The Korla 36cm telescope has a field of view of approximately 2.66 degrees and an image resolution of 4096x4096. This telescope is suitable for star chart identification using the Tycho catalog.

[0105] In terms of functional testing, 639 images were selected, and star map recognition was successful in 544 images, with a success rate of approximately 85%. Some images from Korla were of poor quality, with fewer than 5 stars extracted from them. There were 597 images with more than 5 stars, and based on this number, the success rate was 91.1%.

[0106] In terms of performance testing, the total time for single-threaded computation of 639 images was 25.249 seconds, with an average time of approximately 39.514 milliseconds per image.

[0107] Example 3 The Jilin 1.2-meter telescope has a field of view of approximately 0.37 degrees and an image resolution of 2048x2048. This telescope is suitable for star chart identification using the GAIA star catalog.

[0108] In terms of functional testing, 422 images were selected, and the star map was successfully recognized in 421 of them, with a success rate of 99.8%.

[0109] In terms of performance testing, the total time for single-threaded computation of 422 images was 1.684 seconds, with an average time of approximately 3.991 milliseconds per image.

[0110] In addition to providing a fast star map identification method based on an adaptive star catalog, the present invention further provides a system for implementing the above method, the system comprising: The star catalog generation module generates customized GAIA star catalog files that can be read into memory, based on the original GAIA star catalog data.

[0111] The star table storage module stores the GAIA star table file and Tycho 2 star table file in memory according to the designed index structure.

[0112] The star catalog selection module selects the appropriate star catalog type from the star catalog storage module based on the image information.

[0113] The star catalog extraction module extracts star data that matches the image from the star catalog based on information such as the direction and field of view in the image.

[0114] Select a typical stellar module and use quicksort to find the brightest stars.

[0115] The transformation model solving module dynamically generates a simplified transformation model based on the information in the current image.

[0116] Construct a multi-feature triangle module to generate multi-feature triangles based on a dynamic transformation model.

[0117] The hierarchical fast triangle matching module uses features with high classification efficiency to quickly pair stars in the image with stars in the star catalog.

[0118] The voting module identifies stars in the image based on the triangle matching results and obtains the right ascension and declination information of the stars in the image.

[0119] The above are preferred embodiments of the present invention. It should be noted that, for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A fast star map recognition method based on an adaptive star catalog, characterized in that: The method includes the following steps: Step 1: Generate GAIA star catalog files from the original GAIA star catalog data using a block-based and layered extraction method; Step 2: Read the GAIA star catalog file and Tycho II star catalog file into memory, partition and store the star information, and build the star catalog index; Step 3: Input the field of view information of the optical image, and select the star catalog type to use based on the input information; Step 4: Based on the pointing and field of view information of the input optical image, read the required star information according to the star catalog index; Step 5: Use the quick sorting method to sort the stars in the image and the star catalog, and select the brightest stars for quick star catalog matching; Step 6: Vote based on the matching results to find the star map with the highest matching accuracy, and select it as the successfully identified star.

2. The fast star map recognition method based on an adaptive star catalog as described in claim 1, characterized in that: In step 3, the method for selecting the star catalog includes the following steps: Step 3.1: Obtain the field of view and pixel scale information of the current image, and preset the first threshold; Step 3.2: If the field of view and pixel scale of the current image are greater than the first threshold, select the Dichotomy II star catalog; otherwise, select the GAIA star catalog.

3. The fast star map recognition method based on an adaptive star catalog as described in claim 2, characterized in that: Step 4, the method for reading the required stellar information based on the star catalog index, includes the following steps: Step 4.1: Calculate the lower boundary of the right ascension ra of the field of view based on the direction of the field of view center, the declination dec, and the size of the field of view v. min upper boundary of right ascension ra max Declination lower boundary dec min and the upper boundary of declination dec max ; Step 4.2, for those belonging to right ascension ra min -ra max Declination min -dec max The sub-regions within the range are searched one by one; Step 4.3: For each sub-region, find its starting and ending indices in the star catalog array based on the star catalog index; Step 4.4: From the starting position to the ending position of the sub-sky region, determine whether each star is within the field of view and meets the magnitude threshold; Step 4.5: Store the star data of all sub-regions that meet the conditions in the buffer.

4. The fast star map recognition method based on an adaptive star catalog as described in claim 3, characterized in that: In step 5, the star table matching method includes the following steps: Step 5.1: Use the quick sorting method to sort the stars in the image and the star catalog, and select the brightest stars for quick star catalog matching; Step 5.2: Based on the telescope's direction, dynamically solve for a simplified conversion model from pixels to angles in the current image; Step 5.3: Based on the dynamic transformation model, calculate the estimated celestial positions of the stars in the image, and then construct a multi-feature triangle that includes the dot product of the long and short sides, the ratio of the long and short sides, and the features of the long side. Step 5.4: First, use the bisection method to search for the ratio of the long side to the short side, quickly find several candidate triangles that are closest to the current triangle to be matched, then determine whether other features meet the conditions, and calculate the matching error.

5. The fast star map recognition method based on an adaptive star catalog as described in claim 4, characterized in that: In step 5.1, a recursive approach is used for quicksort, as detailed below: Each time, a pivot number is selected, and the array is divided into two parts: elements smaller than the pivot number are placed in the left queue; elements larger than the pivot number are placed in the right queue; the arrays in the left and right queues are processed recursively until the array to be processed is empty, and then the process returns to the previous level.

6. The fast star map recognition method based on an adaptive star catalog as described in claim 5, characterized in that: In step 5.2, solving the simplified transformation model includes the following steps: Step 5.2.1: Obtain the orientation of the image center. az0 , up and down el0 Right Ascension ra0 and declination dec0 ; Step 5.2.2: Initialize the transformation matrix parameters using the pixel scale bar dpp; Step 5.2.3: Set the positions of 5 simulated stars in the image, assuming the image resolution is [resolution value missing]. x_size * y_size Calculate the pixel positions of these 5 simulated stars S0-S5; Step 5.2.4: Set the right ascension and declination celestial sphere position of the simulated star S0 as ( ra0 , dec0 ); Step 5.2.5: Calculate the azimuth and elevation positions of the remaining simulated stars; Step 5.2.6: Based on the station coordinates and observation time, convert the obtained azimuth and elevation positions of the simulated star to obtain its right ascension and declination positions. ra , dec ); Step 5.2.7: Calculate the input and output data of the CD transformation matrix from the pixel positions and right ascension and declination positions of the simulated stars; Step 5.2.8: Calculate the input and output data of 5 sets of CD transformation matrices from 5 simulated stars, input them into the least squares equation solving function, and solve for the CD matrix parameters of the simplified transformation model.

7. The fast star map recognition method based on an adaptive star catalog as described in claim 6, characterized in that: In step 5.3, the method for constructing multiple feature triangles includes at least the following steps: Step 5.3.1: Convert the pixel positions of the stars into celestial angular positions; Step 5.3.2: Calculate the side length of the line connecting any two selected stars; Step 5.3.3: Select any three stars to form a triangle, and determine the longest side, the shortest side, and the middle side; calculate the ratio of the longest side to the shortest side, the ratio of the middle side to the shortest side, and the ratio of the longest side to the middle side; calculate the dot product of the longest side and the middle side. Step 5.3.4: Extract the seven features of the triangle: the longest side, the shortest side, the middle and longest side, the ratio of the longest side length to the shortest side length, the ratio of the middle and longest side length to the shortest side length, the ratio of the longest side length to the middle and longest side length, and the dot product of the longest side and the middle and longest side length. Store the seven features of all triangles into an array.

8. The fast star map recognition method based on an adaptive star catalog as described in claim 7, characterized in that: Step 5.4 includes at least the following steps: Step 5.4.1: Sort the star triangles in the image and star catalog from largest to smallest according to the ratio of their longer and shorter sides; Step 5.4.2: Select one triangle from the sorted star catalog stellar triangles and match it with the triangle in the image; Step 5.4.3: For the current triangle to be matched, use the bisection method to find the triangle with the smallest error from the sequence of length-to-short side ratios of triangles in the image; Step 5.4.4: Using the triangle with the smallest ratio error of the longer and shorter sides as a reference, search for triangles with errors less than a specified threshold before and after it, and store them in the buffer. Step 5.4.5: For each triangle found, determine whether other features are less than the threshold, and record the error value of the dot product of the length, middle and side. Step 5.4.6: Select 7 triangles whose feature errors all satisfy the threshold and have the smallest length-middle-side dot product error, and calculate the linear correlation between the triangle and the triangle to be matched. Step 5.4.7: Set a second threshold. If the linear correlation between the two triangles is less than 1, the match is considered successful and the matching result is stored; otherwise, the match fails and the next star table triangle is matched. Step 5.4.8: Perform the above operation on each star table triangle until the number of matched triangles meets a certain requirement.

9. The fast star map recognition method based on an adaptive star catalog as described in claim 8, characterized in that: Step 6 includes at least the following steps: Step 6.1: Construct the voting matrix VM and initialize it to 0; Step 6.2: Traverse each matched triangle and vote on the stars it contains; Step 6.3: After voting for each matching triangle, select the star map that meets the voting requirements from the voting matrix; Step 6.4: Save these star maps to an array and sort them from highest to lowest vote count.

10. A fast star map recognition system based on an adaptive star catalog, characterized in that: The system is used to implement the fast star map identification method of the adaptive star catalog according to any one of claims 1-9, and the system comprises: The star catalog generation module generates customized GAIA star catalog files that can be read into memory, based on the original GAIA star catalog data. The star table storage module stores the GAIA star table file and Tycho 2 star table file in memory according to the designed index structure; The star catalog selection module selects the appropriate star catalog type from the star catalog storage module based on the image information. The star catalog extraction module extracts star data that matches the image from the star catalog based on information such as the direction and field of view in the image; The star catalog matching module uses a fast sorting method to sort the stars in the image and the star catalog, and selects the brightest stars for fast star catalog matching. The voting module identifies stars in the image based on the matching results and obtains their right ascension and declination information.

Images