A Star Map Registration Method Based on Semi-Analytical Fourier Transform of Star Point Coordinates

By using a semi-analytical Fourier transform method based on star point coordinates, the problems of high computational load and high storage cost for high dynamic star map registration on spaceborne platforms are solved, achieving fast and accurate star map registration and meeting the real-time computing needs of spaceborne platforms.

CN118644535BActive Publication Date: 2026-11-10INST OF OPTICS & ELECTRONICS CHINESE ACAD OF SCI
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411019243.X
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-07-29
Publication Date
2026-11-10
Estimated Expiration
2044-07-29

AI Technical Summary

Technical Problem

Existing technologies for high-dynamic star map registration on spaceborne platforms involve large computational loads and high storage costs, failing to meet real-time computing requirements. In particular, computational efficiency is low against dense star backgrounds, making it impossible to meet the real-time and high-precision requirements of spaceborne resource-constrained platforms.

Method used

A semi-analytical Fourier transform method based on star point coordinates is adopted. The star point coordinates are transformed by analytical Fourier transform algorithm, and combined with the inverse Fourier transform and the fast search strategy of spatial domain correlation function, the star map is registered quickly.

Benefits of technology

It significantly reduces computational and data storage requirements, achieves sub-pixel level registration accuracy, meets the real-time and high-precision requirements of spaceborne platforms, accelerates computation speed, and is suitable for real-time star map registration on resource-constrained platforms.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118644535B_ABST
    Figure CN118644535B_ABST
Patent Text Reader

Abstract

This invention provides a star map registration method based on semi-analytical Fourier transform of star point coordinates, belonging to the star map registration problem in the field of image processing. Compared with the traditional Fourier domain phase correlation method in image processing, the star point coordinate transformation method of this invention has the characteristics of low computational complexity, small computational load, and high accuracy (sub-pixel). Addressing the limitation of computing resources on spaceborne embedded platforms, the algorithm provided by this invention has small storage requirements and does not require complex numbers in the calculation, making it suitable for the real-time computing needs of resource-constrained platforms. This invention provides a set of fast semi-analytical Fourier transform based on star point coordinates, meeting the requirements of real-time computing and sub-pixel positioning accuracy. Its main schemes include: a fast calculation scheme for the real and imaginary parts of the star map Fourier phase correlation function, a fast calculation scheme for the inverse Fourier transform of the phase correlation function, and a fast search strategy for registration values.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention belongs to the field of image registration technology, and particularly relates to a star map registration method based on semi-analytical Fourier transform of star point coordinates. This method can significantly reduce the time consumption and computational cost required for registration between star maps with large field of view using conventional high dynamic range star sensors, ensuring the real-time computational needs of space observation platforms with limited spaceborne resources. This invention belongs to the field of star map registration in image processing. Background Technology

[0002] Image registration is the process of matching and overlaying two or more images acquired at different times, by different sensors (imaging devices), or under different conditions (weather, illumination, camera position and angle, etc.). It has been widely used in remote sensing data analysis, computer vision, image processing and other fields.

[0003] The registration process is as follows: First, feature points are extracted from the two images; then, matching feature point pairs are found through similarity measurement; next, image space coordinate transformation parameters are obtained from the matching feature point pairs; finally, image registration is performed using the coordinate transformation parameters. Feature extraction is the key to registration, and accurate feature extraction ensures successful feature matching. Therefore, finding feature extraction methods with good invariance and accuracy is crucial for matching precision.

[0004] Star maps, unlike regular images, lack texture features and feature points. Typically, bright stars extracted from star maps are used for association registration based on the principle of star group geometric invariance. This method offers high registration accuracy but requires storing a large number of star feature triangles, resulting in extremely high storage costs. Furthermore, the computational efficiency of inter-star group matching calculations is low, with significant latency. Against a dense stellar background, the storage requirements of this method increase exponentially, which is unacceptable for resource-constrained spaceborne platforms. On the other hand, Fourier phase correlation is another potential solution for star map registration. Fourier transform correlation, also known as phase correlation, utilizes the Fourier spectrum and its transform of an image to perform correlation operations for image recognition and registration. Since the Fourier spectrum and its transform contain information about image translation, rotation, and scaling, Fourier transform correlation can provide this information between images simultaneously during image recognition. Fourier methods are widely used in image registration, image recognition, rotation angle recognition, video target detection and tracking, image motion estimation and compensation, and beacon tracking.

[0005] The Fourier phase correlation method requires two Fourier transforms to calculate the translation parameters of two images: one for calculating the correlation coefficient and one for performing an inverse Fourier transform and searching for the two-dimensional spatial maxima. This computational process is extremely time-consuming. The complexity of the Fast Fourier Transform (FFT) increases rapidly with the number of pixels. Specifically, in star image recognition, the pixel size of large field-of-view star images is typically 4K. If the above algorithm is used directly for 4K resolution, the time to complete one recognition is approximately 4.4 seconds (4096×4096 pixels, Matlab2018, CPU 3.5GHz), which is far from meeting the real-time processing requirements of a satellite platform with a sampling rate of 10 Hz.

[0006] Only a few pixels with stars in the star map provide useful information; most are useless background signals. Applying a Fast Fourier Transform (FFT) based on all pixels in the original image involves a large amount of invalid computation involving multiplication by zero. Summary of the Invention

[0007] Considering that the star map has already been binarized and connected component extracted, making it easy to obtain the specific centroid coordinates of each star, replacing the transformation of all pixels with a Fourier transform involving only star coordinates can significantly reduce the computational load. Therefore, this invention provides a semi-analytical discrete Fourier expression involving only star coordinate calculation, discarding a large number of invalid pixels, and further reducing computational costs through downsampling coarse positioning and secondary fine positioning, making it suitable for applications on resource-constrained platforms. This invention provides a set of fast analytical Fourier transforms based on star coordinates, meeting the requirements of real-time computation and sub-pixel positioning accuracy. Its main schemes include: a fast analytical calculation scheme for the real and imaginary parts of the star map Fourier phase correlation function, a fast calculation scheme for the inverse Fourier transform of the phase correlation function, and a fast search strategy for registration values.

[0008] The purpose of this invention is to address the limitation of computing resources on spaceborne embedded platforms by providing a star map registration method based on semi-analytical Fourier transform of star coordinates. This method features low storage requirements and does not require complex number operations, making it highly suitable for the real-time star map registration needs of resource-constrained platforms.

[0009] The technical solution adopted in this invention is as follows: A star map registration method based on semi-analytical Fourier transform of star point coordinates includes the following steps: Step 1: Read in the reference image and the target image to be registered, and perform image preprocessing steps including denoising and extracting candidate star points to obtain the star point coordinates; Step 2: Use the analytical Fourier transform algorithm to transform the star point coordinates, substitute the star point coordinates into the calculation, and obtain the phase correlation function in the Fourier domain; Step 3: Perform an inverse Fourier transform on the phase correlation function to obtain the spatial domain correlation function; Step 4: Search for the coordinates of the first peak of the spatial domain phase correlation function, i.e., the coordinates (a, b) corresponding to the maximum value; Step 5: Use the coordinates (a, b) of the first peak as the image translation and registration amount to perform image translation calculation, thereby achieving rapid registration of the star map.

[0010] In summary, due to the adoption of the above technical solution, the beneficial effects of the present invention are: This invention presents a star map registration method based on semi-analytical Fourier transform of star point coordinates. Addressing the issues of conventional subgraph-based registration methods in high-dynamic star maps with high star movement speeds and dense star formations, which suffer from high computational and storage requirements and cannot meet the real-time computing needs of spaceborne platforms, this invention provides a fast registration method based on semi-analytical Fourier phase correlation. This method significantly accelerates the configuration speed of high-dynamic star maps, while requiring less data storage and achieving sub-pixel level registration accuracy, thus meeting the real-time and high-precision requirements of spaceborne platforms. Attached Figure Description

[0011] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0012] Figure 1 This is a flowchart of a star map registration method based on semi-analytical Fourier transform of star point coordinates according to the present invention. Figure 2 This is a registration diagram of two star images against a sparse star background and under strong noise conditions. Figure 3 This is a registration diagram to demonstrate a dense star background and a low intersection of two sets of star points. Detailed Implementation

[0013] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention. Furthermore, the technical features involved in the various embodiments of this invention described below can be combined with each other as long as they do not conflict with each other. To achieve the above objectives, this invention adopts the following technical solution.

[0014] The features and performance of the present invention will be further described in detail below with reference to embodiments.

[0015] Example 1 like Figure 1 The diagram shown is a flowchart of a star map registration method based on semi-analytical Fourier transform of star point coordinates according to the present invention, which includes the following steps: Step 1: Read in the reference image and the target image to be registered, and perform image preprocessing steps including denoising and extracting candidate star points to obtain the star point coordinates; Step 2: Use the analytical Fourier transform algorithm to transform the star point coordinates, substitute the star point coordinates into the calculation, and obtain the phase correlation function in the Fourier domain; Step 3: Perform an inverse Fourier transform on the phase correlation function to obtain the spatial domain correlation function; Step 4: Search for the coordinates of the first peak of the spatial domain phase correlation function, i.e., the coordinates (a, b) corresponding to the maximum value; Step 5: Use the coordinates (a, b) of the first peak as the image translation and registration amount to perform image translation calculation, thereby achieving rapid registration of the star map.

[0016] Step 1 includes the following steps: Step 1.1: Read in the reference image and the target image to be registered; Step 1.2: Perform mean filtering and Gaussian filtering on the two images to obtain a star map with background and noise removed; Step 1.3: The sampling adaptive threshold segmentation method is used to binarize the two images and extract connected regions; Step 1.4: Calculate and extract the centroid coordinates of the connected components of the star points using the weighted centroid method, such as... Figure 2 and 3 The right figure shows the extracted star point coordinates.

[0017] Step 2 includes the following steps: Step 2.1: Using the fast analytical Fourier transform equation (star point coordinate transformation) provided by this invention, the star point coordinates extracted from the image preprocessing in Step 1 are substituted into the calculation to obtain the real and imaginary parameters of the phase correlation function in the Fourier domain. The analytical Fourier transform equation is as follows: ; In the above formula For the extracted star point coordinates, and This is a set of star coordinates extracted from two images. and Image dimensions (number of rows and columns in pixels), and This represents the parameters in the Fourier domain. Parameters a, b, c, d, e, and f are the relevant parameters for the analytic Fourier transform.

[0018] The specific steps of step 3 are as follows: Step 3.1: Apply the phase correlation parameters obtained in Step 2 and An inverse Fourier transform is performed to obtain the spatial domain correlation function. The transformation equation employs a semi-analytical operator provided in this invention, which allows for rapid acquisition of the transformation result. The equation is expressed as: ; in, and The calculation is based on the equations provided in step 2.1. Spatial domain correlation function. Similar to the Dirac impulse function The peak position coordinates of this function correspond exactly to the translation parameter. When two images are uncorrelated, the detection results will not show a significant peak. For example... Figure 2 and 3 The left figure shows the spatial domain phase correlation function. The position indicated by the red circle represents the first peak value related to the registration amount, and the corresponding coordinates are the obtained configuration amount.

[0019] The specific steps of step 4 are as follows: Step 4.1: Calculate the spatial domain phase correlation function A simplified formula can be provided in this invention, which can be calculated using a small number of wavenumber points:

[0020] The wave number satisfies: ; as well as: ; and The upsampling interval for the Fourier domain. and When taking a larger value The peak value tends to be a narrow impulse function. When taking a smaller value, The peak value is relatively flat, but this can significantly reduce computational costs. It can be proven that this is possible without noise and coordinate positioning errors. and The minimum value can be set to 2. However, considering the error and algorithm stability, this parameter should be allocated reasonably in conjunction with computing resources.

[0021] Step 4.2: This invention provides a spatial domain upsampling fast search method, which achieves fast coarse localization by downsampling the correlation function, and then performs sub-pixel precise search centered on the coarse localization. The spatial domain downsampling process of the correlation function is as follows: ; The above formula and is the spatial domain downsampling interval. When and When the value is less than 1, the algorithm achieves sub-pixel registration accuracy, but the computational load is extremely high, especially for high-dynamic star sensors where a large search interval is required. Here, we set... as well as =3 performs a three-fold downsampling, reducing the computational load to one-ninth of the original. This step yields a coarse localization. After that The search strategy provided by this invention is as follows: Subpixel-centric search is performed. ; in, After a second local search, the registration accuracy is greatly improved without a significant increase in computational cost. Therefore, this strategy is very suitable for hardware solutions with limited computing resources, such as spaceborne embedded systems.

[0022] Step 4.3: Search for the first peak of the spatial domain correlation function obtained in Step 4.2, i.e., the x-axis coordinate (a, b) corresponding to the maximum value. The calculation process is as follows: finding the extremum of the function. ; In theory The maximum peak value was obtained at that time. The final obtained position is as follows: Figure 2 and 3 The position circled in red in the left image is the first peak position obtained through the search in step 4, representing the registration amount between the two frames.

[0023] The specific steps of step 5 are as follows: Step 5.1: Use the first peak coordinates (a, b) obtained in Step 4 as the image translation registration amount to perform image translation calculation, thereby achieving rapid registration of the star map.

[0024] Analyze the effects based on the attached diagram: Figure 1 This describes the flowchart of the star map rapid registration method based on coordinate transformation of the present invention. Figure 2 and Figure 3 This demonstration shows the registration results of two star images under different stellar backgrounds and noise conditions. The input is the stellar coordinates of the two star images, and the output is the estimated configuration. Figure 2 Against a background of sparse stars, Figure 3This is a scenario with high noise and low overlap. The left figure shows the spatial domain phase correlation function obtained by the scheme provided in this invention (steps 3 and 4), with the searched peak positions marked as gray circles (step 4.3), and the position coordinates are output as registration values. The right figure is a schematic diagram of the extracted star coordinate positions in two frames, where gray circles represent candidate stars extracted in the first frame, gray dots represent candidate stars extracted in the second frame, and the connecting lines represent the registration values ​​between overlapping stars. Figure 2 To demonstrate the registration of two star images against a sparse stellar background and under strong noise conditions, the left image shows the spatial domain phase correlation function obtained by the analytical Fourier algorithm of this invention, with the peak positions marked as gray circles. The right image shows the extracted stellar coordinates of the two frames, where gray circles represent the first frame, gray dots represent the second frame, and the connecting lines represent the registration amount between the intersecting stars. This invention calculates the translation amount (63, 83.2) between the two frames, which is consistent with the actual registration amount, proving the feasibility of this invention. Figure 3 To demonstrate registration under dense stellar backgrounds and low intersection of star point sets between two frames, the left image shows the spatial domain phase correlation function obtained by the analytical Fourier algorithm of this invention, with the peak position marked by gray circles. The right image shows the extracted star coordinate positions from the two frames, where gray circles represent the first frame, gray dots represent the second frame, and the connecting lines represent the registration amount between intersecting stars. The translation amount between the two frames (72.6, 40.6) is obtained, and the error with the actual registration amount is within one pixel, proving that this invention can still provide stable registration amounts under low intersection and strong noise conditions. Figure 2 and 3 This invention demonstrates the translation amount between two frames calculated using this method. Verification shows that this amount has an error of less than one pixel compared to the actual registration amount, and the processing time for a single frame is less than 30 milliseconds. It features high accuracy and fast calculation, meeting the real-time computing requirements of resource-constrained spaceborne platforms.

Claims

1. A star map registration method based on semi-analytical Fourier transform of star point coordinates, characterized in that: The method includes the following steps: Step 1: Read in the reference image and the target image to be registered, and perform image preprocessing steps including denoising and extracting candidate star points to obtain the star point coordinates; Step 2: Use the analytical Fourier transform algorithm to transform the star point coordinates, substitute the star point coordinates into the calculation, and obtain the phase correlation function in the Fourier domain; Step 3: Perform an inverse Fourier transform on the phase correlation function to obtain the spatial domain correlation function; Step 4: Search for the first peak coordinate of the spatial domain phase correlation function, that is, the coordinates (a, b) corresponding to the maximum value; use the spatial domain of the correlation function to perform downsampling to obtain a coarse localization, and then use the coarse localization as the center to perform sub-pixel search to obtain the first peak coordinate; Step 5: Use the coordinates (a, b) of the first peak as the image translation and registration amount to perform image translation calculation, thereby achieving rapid registration of the star map; Step 2 includes the following steps: Step 2.1: Perform star point coordinate transformation using the analytical Fourier transform equation, substitute the star point coordinates into the calculation, and obtain the real and imaginary parts of the phase correlation function in the Fourier domain. The analytical Fourier transform equation is as follows: ; In the above formula For the extracted star point coordinates, and This is a set of star coordinates extracted from two images. and Image size, i.e., the number of rows and columns in pixels. and represents the Fourier domain parameters; parameters a, b, c, d, e, and f are the relevant parameters for the analytic Fourier transform. Step 3 includes the following steps: Step 3.1: For phase correlation parameters and Performing an inverse Fourier transform yields the spatial domain correlation function; the transformation equation employs a semi-analytic operator to obtain the transformation result. The equation is expressed as: 。 2. The star map registration method based on semi-analytical Fourier transform of star point coordinates according to claim 1, characterized in that: Step 1 includes the following steps: Step 1.1: Read in the reference image and the target image to be registered; Step 1.2: Perform mean filtering and Gaussian filtering on the two images to obtain a star map with background and noise removed; Step 1.3: Binarize the two images using an adaptive threshold segmentation method and extract connected components; Step 1.4: Use the weighted centroid method to obtain the centroid coordinates of the star-point connected domain.

3. The star map registration method based on semi-analytical Fourier transform of star point coordinates according to claim 1, characterized in that: Step 4 includes the following steps: Step 4.1: Calculate the spatial domain phase correlation function A simplified formula is provided using wavenumber point calculation: The wave number satisfies: ; as well as ; and The upsampling interval for the Fourier domain; Step 4.2: The spatial domain downsampling formula for the correlation function is as follows: ; The above formula and Using the spatial domain downsampling interval, this step yields a coarse localization. After that A sub-pixel search is performed centered on the target pixel, using the following search strategy: ; in, ; Step 4.3: Search for the first peak coordinate of the spatial domain correlation function, i.e., the x-axis coordinate (a, b) corresponding to the maximum value. The calculation process is as follows: finding the extremum of the function: ; exist The maximum peak value is obtained at that time.

4. The star map registration method based on semi-analytical Fourier transform of star point coordinates according to claim 3, characterized in that: Step 5 includes the following steps: Step 5.1: Use the coordinates (a,b) of the first peak as the image translation registration amount to perform image translation calculation, thereby achieving rapid registration of the star map.