Computer implemented method of determining the position of an observed celestial object

The method of aligning and comparing multiple images to create a mosaic for astrometric analysis enables low-cost sensors to determine celestial object positions, addressing the challenge of insufficient star trails in small field of view sensors and enhancing space domain awareness.

GB2630007APending Publication Date: 2024-11-20THE SEC OF STATE FOR DEFENCE IN HER BRITANNIC MAJESTYS GOVERNMENT OF THE UK OF GREAT BRITAIN & NORTHERN IRELAND
View PDF 1 Cites 0 Cited by

Patent Information

Application Number
GB2022010827
Authority / Receiving Office
GB · GB
Patent Type
Applications
Current Assignee / Owner
Filing Date
2022-07-25
Publication Date
2024-11-20

AI Technical Summary

Technical Problem

Low-cost, small field of view sensors used in amateur astronomy struggle to capture sufficient star trails in a single image for effective astrometric measurements, making it challenging to determine the position of celestial objects and limiting their application in space domain awareness.

Method used

A computer-implemented method that aligns and compares multiple remotely sensed images to create an image mosaic, extracting pixel locations of background celestial object trails, and comparing them to a library of known celestial object reference frames to determine the position of observed celestial objects, enabling the use of lower-cost sensors for space domain awareness.

Benefits of technology

This method allows for the determination of celestial object positions using lower-cost sensors, enhancing space domain awareness by replicating a larger field of view image with smaller field of view images, making space-based object tracking more accessible and economically viable.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure 00000000_0000_ABST
    Figure 00000000_0000_ABST
Patent Text Reader

Abstract

Determining celestial object position by: receiving consecutively sensed images of the celestial object against a celestial background, the images being aligned on the celestial object; comparing images to obtain their relative pixel offsets; aligning the images according to their relative pixel offsets to obtain an image mosaic; extracting pixel locations of background celestial object trails from the image mosaic; comparing the pixel locations to a library of known celestial object positions to obtain a reference frame transformation that transforms the pixel locations to positions in the reference frame of the known celestial objects and determining therefrom the position of the observed celestial object. Pixel offsets may be determined between temporally adjacent image pairs. Observing equipment with narrow field of view (FOV) may be directed to a fixed point in the sky to capture long-exposure photographs in which positions of objects (eg. satellites, spacecraft, rocket bodies, collision or explosion fragments) remains fixed but background objects (eg. stars) appear as streaks due to Earth’s rotation. Sensed photographs may be stitched to create a wide FOV composite image for input into plate solving algorithms which identify photographed stars, by comparison against a catalogue, and determine their celestial positions, eg. right ascension and declination.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field of the Invention This invention relates to remote sensing of celestial objects, and more particularly to the field of space domain awareness. Background to the Invention Space domain awareness relates to the monitoring of celestial objects in the vicinity of Earth. Both active and abandoned objects currently orbit the Earth, including but not limited to abandoned spacecraft; rocket bodies; fragments from explosions and collisions; and the actively growing numberof satellites relied upon by terrestrial navigation and communication systems. Detecting and monitoring celestial objects is growing in importance in an attempt to ensure space based objects avoid collisions; to ensure risks and likelihoods of the re-entry of objects towards Earth are understood; as well as to gain intelligence insights into unknown objects appearing in Earth orbit. Telescope sensors with large fields of view are a mechanism for detecting and tracking celestial objects. In addition to viewing the celestial object itself, these sensors are able to view and capture large swathes of celestial star background behind a celestial object to assist with astrometric and photometric measurements. As well as celestial objects such as stars, telescope sensors can also be used to observe and image man-made objects such as satellites. When such sensors are imaging, for instance, geostationary satellites, stars in the celestial background of the image will typically appear as streaks across the image. This is because the frame integration time of the telescope sensor is not instantaneous, and so there is a relative motion between the background stars and the geostationary satellite during the period an image is captured. These 'star streaks' are useful for astrometric measurements, because their relative separations and geometry can be used by established plate-solving or platematching algorithms (such as those available at Astrometry.net) to ascertain World coordinates corresponding to pixel locations in the sensed image. Hence it becomes possible to determine the World coordinate (for instance Right Ascension and Declination) of the geostationary satellite for space domain awareness purposes. Indeed this process is not necessarily limited to geostationary satellite tracking, but may be used to determine the position of other satellites such as low Earth orbit (LEO) satellites and indeed other celestial objects that need tracking for space domain awareness. This extends even to applicability in tracking Asteroids for which a similar relative motion between celestial background and celestial object exists. Furthermore such technologies have potential applications in star trackers used in on-board navigation equipment in spacecraft to maintain a trajectory or orientation relative to a celestial star background. For position determination of celestial objects against a star background using remotely sensed imagery it follows that it is essential there be sufficient star trails present in sensed imagery to allow for plate-solving or astrometric techniques to be effectively applied. Large field of view telescope sensors permit this, but are relatively expensive, making the costs of establishing multiple space domain awareness systems longitudinally dispersed around the Earth, prohibitively high. Amateur astronomy defines a further sector of celestial observation that is accessible to many enthusiasts. The associated telescopes and sensors are capable of capturing digital imagery of celestial objectsand backgrounds, albeit with small field of view sensors that unfortunately are often unable to capture sufficient startrails in a single sensed image to allow plate solving to be implemented effectively. Therefore whilst satellite viewing (and indeed the viewing and image capture of other celestial objects) is achievable using relatively low cost amateur equipment, correlating the observed objects with actual World coordinates is more challenging. However the smaller lower cost optics and sensors are of themselves more easily distributable geographically and present a highly versatile solution for space domain awareness systems should techniques for their integration and processing of their imagery improve. Therefore it is an aim of the present invention to provide a computer implemented method for determining the position of an observed celestial object using remotely sensed imagery of the observed celestial object against a celestial background. Summary of the Invention According to a first aspect of the invention, there is provided a computer implemented method of determining the position of an observed celestial object using remotely sensed images of the observed celestial object against a celestial background, the method comprising the steps of: receiving from remote sensing means a plurality of remotely sensed images of an observed celestial object against a celestial background, the images being aligned onto the observed celestial object and having been sensed consecutively in time; applying an image comparison operation to the plurality of remotely sensed images to obtain their relative pixel offsets; aligning the remotely sensed images according to their relative pixel offsets to obtain an image mosaic; extracting a plurality of pixel locations of background celestial object trails from the image mosaic; and then comparing the plurality of pixel locations to a library of known celestial object reference frame positions, to obtain a reference frame transformation that transforms pixel locations in the image mosaic to positions in the reference frame of the known celestial objects, and determining therefrom the position of the observed celestial object. Determining the World coordinates of celestial objects in images obtained from remote sensing apparatuses is achievable if the arrangement of celestial objects (such as stars) in the celestial background being imaged can be unambiguously recognised. This is because the positions of such celestial objects at particular dates and times can often be predicted in mathematical models or are monitored by observation with other higher fidelity large field of view sensing systems. However this approach is dependent on being able to recognise sufficient celestial objects in a celestial background and determine their relative positions accurately. For low cost small field of view sensors this can be particularly challenging as a given image simply cannot capture enough background to, individually, allow astrometric measurements. The inventor has recognised that a larger field of view image with sufficient celestial background imaged can essentially be replicated through the mosaic of smaller field of view images that have been imaged consecutively over time. This unlocks the ability to utilise lower cost sensors and optics to generate fused imagery suitable for processing within established astrometric and plate solving techniques, rendering celestial object tracking and space domain awareness far more accessible. According to a second aspect of the invention, there is provided a computer program containing code which when executed performs the computer implemented method of the first aspect of the invention. The computer program may be installed or programmed into suitable data processing means for processing received remotely sensed imagery. According to a third aspect of the invention, there is provided computer readable storage media comprising the computer program of the second aspect of the invention. The computer readable storage media is intended to be non-volatile storage media. The computer readable storage media may be connected to or inserted into suitable data processing means in order to operate the computer program for executing the method of the first aspect of the invention. According to a fourth aspect of the invention, there is provided apparatus for remotely sensing the position of an observed celestial object against a celestial background, comprising remote sensing means for consecutively sensing a plurality of remotely sensed images aligned onto an observed celestial object against a celestial background; and data processing means configured to execute the method of the first aspect of the invention. The remote sensing means may comprise a telescope attached to a digital camera, for instance, as may be used by amateur astronomers when observing and imaging the sky. The remote sensing means may comprise a plurality of such telescopes and cameras. According to a fifth aspect of the invention, there is provided a space domain awareness system comprising the apparatus of the fourth aspect of the invention. A space domain awareness system may be local or geographically widespread (for instance longitudinally around the Earth). Such a system is generally expensive to implement owing to the requirement for large complex sensor systems or arrays. However such a system can be more readily implemented, at lower cost, by utilising lower cost and lower complexity sensing systems (such as amateur telescopes and cameras) through use of the improved image processing techniques provided by the method of the first aspect of the invention. Hence the ability to monitor space based celestial objects and perform space object tracking on a Worldwide scale becomes more accessible and economically viable. According to a sixth aspect of the invention, there is provided a star tracker for a spacecraft comprising the apparatus of the fourth aspect of the invention. A star tracker for spacecraft such as satellites enables the satellite to autonomously maintain an orientation in a reference frame based on observations of celestial background. However again it is a requirement that sufficient celestial objects (typically stars) in a celestial background are sensed by the star tracker in order to accurately convert an observed scene into a reference frame orientation for navigational corrections to be determined. Generally, larger field of view sensors can enable sufficient celestial background to be imaged for this purpose. However the ability to utilise smaller, cheaper, and lighter or less complex sensors in star trackers is attractive given the size weight and power constraints spacecraft often have. The method of the first aspect of the invention deployed in a star tracker offers the ability to utilise a different class of sensors for this purpose. An observed celestial object is intended to mean the celestial object for which a position is sought to be determined and for which the remotely sensed imagery is aligned thereto. The observed celestial object may be a natural object such as an Asteroid, planet, or star for instance, or may be a man-made object such as a communications satellite (for instance a geostationary satellite), spacecraft, space station, or debris. A celestial object is in general terms an object positioned in the sky or outer space. It follows that a celestial background relates to the region of sky or space upon which the observed celestial object is presented and observed. Practically, the celestial background describes the other celestial objects sensed by a remote sensing apparatus when imaging the observed celestial object. Typically the celestial background comprises one or more stars, but may include planets or indeed may comprise man-made objects such as satellites that are not the observed celestial object of interest. For instance, it is feasible that the observed celestial object may be a satellite and the celestial background may comprise other satellites (for which positions are known). It may therefore be possible, through the method of the invention, to determine the observed satellite position using information regarding the known positions of the background satellites. The remotely sensed images received from remote sensing means are aligned onto the observed celestial object. The term "aligned" is intended to mean that for the duration within which a given image is sensed, the optical axis of the remote sensing means is aligned / fixed / tracking the observed celestial object. From an imagery perspective, each remotely sensed image will therefore present limited, if any, evidence of motion of the observed celestial object. However the celestial background, for which the remote sensing means was not tracking or fixed to, will exhibit motion during the integration time the remote sensing means is sensing an image. Therefore from an imagery perspective, each remotely sensed image will present some evidence of motion of the celestial background. By way of example, a remotely sensed image of a geostationary satellite observed using a ground based telescope and camera in this particular manner will show a substantially point-like object representing the satellite with apparently no motion. However the image will show streaks in other regions of the image that are generated owing to the motion of stars relative to the geostationary satellite across the integration time of the camera. The plurality of remotely sensed images correspond to images taken consecutively in time. The term "consecutive" or "consecutively" in this context means the images are taken at different points in time. Small field of view sensors lack the ability to capture sufficient celestial background in individual images in isolation to enable celestial observation techniques such as plate solving or astrometry to be applied. By utilising multiple consecutive images a celestial background (that is moving relative to the observed celestial object being tracked) will passthrough the field of view of the sensor, and so multiple background celestial objects can be observed. By forming a mosaic of these image frames, or "stitching" / "stacking" these frames together, a combined image can be formed that substitutes for a larger field of view single frame image. In practice, successive images are taken so as to have a portion of the celestial background in common with each other to allow for accurate alignment of the images with each other during image comparison. It is preferable that the time difference between imaging of successive remotely sensed images is less than ten minutes for Earth based observations of geostationary satellites, for instance. The image comparison operation is an operation that compares the remotely sensed images and determines their translational pixel offsets relative each other. For instance when two remotely sensed images are compared with each other, the operation will output pixel off-sets in two dimensions that if implemented would best align the two images with each other. In effect, the image comparison operation provides the offsets that allow two images to be aligned in a common coordinate system (for instance in the pixel coordinate system of the first image, or the second image). The image comparison operation may be an image registration technique, for instance, such as an image correlation, which of themselves are well established techniques in the art of image processing. Examples of correlation techniques include cross correlation and phase correlation. Aligning the images according to their relative pixel off-sets to form an image mosaic involves applying the pixel-offsets to the respective images that were compared during image comparison, in order to align the respective images with each other to create a larger 'image' composed of the remotely sensed images stacked or stitched together. In this manner an image mosaic can be formed that comprises a greater amount celestial background than a remotely sensed image in isolation. The image mosaic will comprise background celestial object trails owing to the motion of the celestial background relative to the observed celestial object during the integration time of the sensors from which each of the remotely sensed images were obtained. For example a star will appear as a streak across a portion of the image owing to its motion relative the observed celestial object during the integration time for the image. The mosaic image contains a plurality of such object trails having offsets relative each other. The pattern of object trails, their offsets from each other, and their relative brightness, are all parameters that can be used to correlate the mosaic image with a particular observation orientation in a reference coordinate system. This is achieved by comparing these parameters to a library of such parameters for known celestial objects. For instance the positions of stars in a World coordinate system such as right ascension and declination may be provided in a library of known celestial object reference frame positions. By comparing the relative pixel positions of star trails in an image mosaic to this library, it is possible to determine the orientation in which the sensor that obtained the images was (or is) pointing. It can be inferred therefore that the optical axis of the sensor, and accordingly the observed celestial object which the optical axis was tracking, was pointing in that particular orientation in World coordinates. This technique, referred to in Astronomy as plate solving, is well established and available in tools such as Astrometry.net. As an alternative example, background celestial object trails may comprise satellite trails from satellites orbiting Earth in a known constellation. An image mosaic showing a plurality of such satellite trails could be compared to a library of known positions of satellites in the satellite constellation to determine the observation orientation of the remote sensing means. It would therefrom be possible to correlate the position of the observed celestial object. In this manner, a reference frame transformation can be obtained that describes how pixels in the image coordinate system of the image mosaic correspond to coordinates in a reference frame in which reference frame positions of known celestial objects are available. In some embodiments, the step of applying an image comparison operation comprises applying an image correlation operation to each consecutive pairing of images and in each case generating a correlation image of correlation scores; and then identifying as the relative pixel offsets for each consecutive pairing of images, the pixel offsets corresponding to an optimum value of the correlation score in the respective correlation image. Image correlation techniques are a relatively quick approach to uncovering the integer translational shift between two images that best aligns the two images with each other. Two-dimensional crosscorrelation is a well understood image correlation operation that yields a 'correlation image' of values (which may be normalised), the maximum value of which often denotes, by virtue of the associated pixel location, the pixel-wise shift necessary to align the two correlated images. Even more preferable is wherein the image correlation operation is a phase correlation. Phase correlation is more resilient to defects typical of sensed imagery (such as noise of other defects) owing to the technique's reliance on the frequency-domain representation of an image (vice the spatial representation). In embodiments utilising an image correlation operation, the pixel offsets corresponding to an optimum value of the correlation score are identified as the relative pixel offsets for the respective pairing of images. The optimum value of the correlation score in some embodiments is the maximum value of the correlation score (or, for instance, the maximum of the normalised correlation score). This is relatively easier to implement than more complex algorithms for determining an optimum value, and is a reliable value where a correlation image has an unambiguous maximum value. However in other embodiments, the step of identifying relative pixel offsets for each consecutive pairing of images comprises applying a weighting function to the correlation scores to generate weighted correlation scores, the weighting function being biased towards a predetermined pixel offset; and then identifying as the relative pixel offsets, the pixel offsets corresponding to an optimum value of the weighted correlation scores in the respective correlation image. The remotely sensed imagery idealistically will feature the observed celestial object as an apparently static celestial object showing almost no motion between consecutive images. Celestial objects in the celestial background will however exhibit motion (both within an image, evidenced as trails, and between images as a result of each image's integration time starting at different points in time). Practically however, a remote sensing means may capture imagery that between images exhibit some, albeit small, motion of the observed celestial object. This can be for a number of reasons, but by way of example, may be owing to the accuracy with which the observed celestial object has been tracked by the remote sensing means, or may be owing to gradual drifting of the observed celestial object (for instance, even geostationary satellites can drift in position slightly over a time period). When such images are correlated with each other, a 'false' maximum value of a correlation score may be uncovered that is associated with the observed celestial object having zero or small offset between the images. Applying this offset would result in an image mosaic for which the background celestial object trails are not accurately aligned. This can lead to incorrect derivation of the reference frame transformation between image and reference frame (such as World) coordinates. Hence it can be necessary to apply a weighting to the correlation scores in a correlation image to bias an algorithm for detecting an optimum score, away from scores associated with certain pixel offsets. For instance, knowing the time between successive remotely sensed images, it can be pre-calculated what the expected relative pixel offsets should be between the same celestial background object trails across two images. A weighting function, for instance a Gaussian function, can be applied to the values of the correlation image that is centred on this expected pixel offset, to bias the decision on an "optimum" value towards such an offset. False optimums therefore are less likely to be identified in the image comparison operation. Similarly, the accuracy of a given remote sensing means in tracking an observed celestial object can be predetermined. This can be converted into a pixel positional error and again be used in a weighting function to bias the decision on an optimum value of a correlation score, away from scores associated with pixel offsets that are less than or equal to the positional error. In preferred embodiments, the step of aligning the remotely sensed images according to their relative pixel offsets comprises co-adding the images. When two images have been correlated and their relative pixel offsets applied to align the images, some areas of the two images will overlap. Co-adding requires that in regions of overlap, the overlapping pixels are added together or summed. This is advantageous with respect to the signal to noise ratio of the resulting image mosaic. Noise present across the two original images will comprise an element that is intrinsically random. The two original remotely sensed images of themselves were imaged at different points in time and so their noise content is unlikely to be identical. Hence the co-adding of image pixel values is unlikely to result in any increase in noise on a pixel by pixel basis. However the true signals in the image (that is those corresponding to celestial objects) are not random. When aligned, co-adding such pixel values will result in an increase in signal for a given celestial object in the image mosaic. Therefore the image mosaic will not only exhibit stronger signals associated with celestial objects observable in the separate remotely sensed images, but may also uncover celestial objects not practically resolvable from the individual images in isolation. Therefore in these embodiments, not only can multiple sensed images be used to form a larger image mosaic for astrometric measurements and plate solving techniques, but the same embodiments also generate image mosaics of greater signal to noise (and therefore utility) than single large field of view sensors might provide. In some embodiments, the step of extracting a plurality of pixel locations of background celestial object trails from the image mosaic, comprises the step of applying a clumping operation to the image mosaic to identify a plurality of contiguous groups of pixels having pixel values greater than a background threshold value, and selecting as the pixel location of each background celestial object trail the centroid pixel of each contiguous group. Each background celestial object trail in the image mosaic will be represented by a group of pixels that are next to each other in a contiguous manner, and have a pixel value that has been predetermined to be of sufficient value to be deemed a useful signal vice simply being noise. A clumping operation seeks to identify contiguous pixels that can be regarded as being associated with an object of interest in an image. A background threshold value for a pixel value of interest can be predetermined according to statistics of the image. For instance it may be preferable for the standard deviation of pixel values of the image mosaic to be used to set a background threshold value of a multiple (2 or 3 times, for instance) of the standard deviation. Pixel values above this value can be considered to be statistically relevant as having a high likelihood of being associated with a background celestial object trail. However a single pixel value, in isolation, is unlikely to be a candidate for such a trail given that trails by definition will be represented by a group of contiguous pixels. Hence the further step of identifying a clump or group of contiguous pixels above the background threshold value is necessary to truly identified object trails in the image mosaic. It may be further necessary to set a threshold value to filter groups of contiguous pixels that have a total pixel count greater than a predetermined value. For instance a group of contiguous pixels may only be considered if having greater that N contiguous pixels, for instance. Once a group of contiguous pixels has been identified, it will be understood that the group will be spread over an area of an image, for which any individual pixel of the group could be selected as the pixel location of a background celestial object. Taking the centroid pixel of the group of pixels as the pixel location is a convenient approach to providing a single pixel location that marks the average location across the group. Weighted averaging may be used to provide the centroid pixel, as an example. In some embodiments, the step of comparing the plurality of pixel locations to a library of known celestial object reference frame positions to obtain a reference frame transformation, further comprises comparing the relative pixel values of the pixel locations to relative brightness values of the known celestial objects to obtain the reference frame transformation. Known background celestial objects may have a known associated brightness. For instance a known star will have a known brightness orflux, and thereby known brightness or flux relative to other stars. By determining both the relative positions of background celestial objects in an image, as well as their relative brightnesses (for simplicity, this may be the ratio of their pixel values), such an arrangement of object locations and relative brightnesses can be compared to candidate arrangements of known celestial objects (such as stars) to determine the actual background celestial objects being observed. Utilising relative brightness of background celestial objects further enhances the ability to discriminate which known celestial objects are being observed. In preferred embodiments, the celestial background is a star background and the background celestial object trails are star trails. Remote sensing of observed celestial objects against a star background is a particular technical field to which the invention as herein described offers significant benefit. This is because star locations can be understood through modelling and or high fidelity instruments, leading to such locations being relatively well catalogued. Sensing imagery of objects against such a background therefore allows for determination of the orientation of a remote sensing means through a look-up operation to known star locations. The invention as herein described provides a mechanism for obtaining imagery with sufficient star background to allow such a look-up operation to be performed. In particular, and in even more preferred embodiments, this allows the application of plate solving operations to the plurality of pixel locations of background celestial object trails. Plate solving refers to the comparison of a plurality of star locations in an image, to a database of star positions. Once the observed stars have been matched, the right ascension and declination of the center of the image is typically returned (which corresponds to the pointing direction of the remote sensing means, and thereby the position of the observed celestial object in a known coordinate system). Astrometry.net is an example of plate solving software that is readily available for this purpose. In some embodiments, the observed celestial object is a geostationary satellite. A geostationary satellite is a satellite that moves in a geosynchronous orbit around an object (such as Earth) such that it appears to be stationary above a fixed point on the surface of the object. Geostationary satellites may need to be monitored to ensure their positions are accurate and not deviating over time. Furthermore, suspect geostationary satellites may need to be monitored for security purposes. Moreover, in an ever increasingly cluttered orbital environment, geostationary satellites may need to be monitored to detect the potential for collisions with new orbiting satellites or debris. Preferred embodiments further comprise the steps of comparing the reference frame position of a background celestial object trail to a library of known background celestial object magnitudes and extracting therefrom a magnitude value associated with the background celestial object trail; calculating a conversion factor for converting between the magnitude value and pixel value associated with the background celestial object trail; and then applying the conversion factor to a pixel value associated with the observed celestial object, thereby determining a magnitude value for the observed celestial object. By comparing the pixel locations of the background celestial object trails with the library of known background celestial object reference frame positions, the background celestial objects can be determined. By performing a look-up operation with a library of known background celestial object magnitudes, the true magnitude of a given background celestial object being observed can be determined. This value can be compared to the pixel value in the image mosaic of that background celestial object, to obtain conversion factor for converting pixel value to magnitude value. This conversion factor can be applied to the pixel value of the observed celestial object to obtain a magnitude value for the observed celestial object. The conversion factor may be generated for a plurality of background celestial objects, and the average of the conversion factors used. This may provide a more reliable conversion factor as any errors or fluctuations in pixel values will have less of an impact on the conversion factor used for determining the magnitude value of the observed celestial object. The inventors have shown through this approach that Gaia g equivalent determinations of photometric magnitude can be obtained for observed satellites against a star background. In preferred embodiments, the reference frame transformation transforms pixel locations in the image mosaic to right ascension and declination. Such a coordinate system is typically used when articulating celestial positions relative to an Earth bound observer. Brief Description of the Drawings Embodiments of the invention will now be described by way of example only and with reference to the accompanying drawings, in which: Figure 1 provides an illustration of a prior art method of determining the position of an observed geostationary satellite against a star background; Figure 2 provides an illustration of a remotely sensed image of an observed geostationary satellite for use in the prior art method of Figure 1; Figure 3 provides an illustration of a remotely sensed image of an observed geostationary satellite not suitable for use in the prior art method of Figure 1; Figure 4 provides an illustration of an embodiment of a computer implemented method of determining the position of a geostationary satellite against a star background; Figure 5 provides an illustration of an embodiment of an image mosaic generated in the method of Figure 4; and Figure 6 provides an illustration of the position of geostationary satellites as determined by the method of Figure 4. Detailed Description Figure 1 provides an illustration of a prior art method 10 of determining the position of an observed geostationary satellite against a star background. The method 10 is shown as a flow diagram of steps normally implemented within a computer. In a first step a single remotely sensed image is received 11 into the computer system. The remotely sensed image is an image of a geostationary satellite observed from Earth against the celestial star background. The image has been sensed using a telescope and sensor array with sufficient field of view to capture a large portion of star background. The sensor array itself will have integrated received radiation from the satellite and star background over an exposure time in order to capture sufficient energy for the star background and satellite to be resolvable. In a subsequent step, the received image is processed to extract the pixel locations of star trails 12 in the image. The remote sensing means obtaining the sensed image is aligned to (or tracking) the geostationary satellite. Thereby the star background motion relative to the satellite within the exposure time of the image is evidenced by the star background appearing not as point sources but as streaks or trails across the sensed image. The extraction of star trails 12 may be manually performed (for instance by clicking the centroid pixels of star trails visible in the image) or may be automatic through suitable algorithm. Regardless, the startrail pixel locations from the extraction step 12 are passed into a plate solving operation 13 as the subsequent step for determination in World coordinates of the geostationary satellite. The plate solving operation 13 is that provided by Astrometry.net. The plate solving operation 13 receives the input startrail pixel locations and returns the astrometry word coordinate system (the transformation between image coordinate i.e. pixels and sky coordinates i.e. right ascension and declination) through comparison of the relative startrail pixel locations with known star locations catalogued by other sensors or determining through modelling. Other plate solving algorithms or tools can be used, with Astrometry.net being provided herein merely by way of example only. Following the plate solving operation 13 the World coordinates of the centre pixel of the remotely sensed image is output 14. The centre pixel of the remotely sensed image corresponds to the pointing direction of the remote sensing means that obtained the original remotely sensed image. This, by virtue of the remote sensing means tracking the geostationary satellite across the exposure time of the sensed image, corresponds to the location of the geostationary satellite. Hence, through the plate solving operation, the geostationary satellite position can be determined. Figure 2 provides an illustration of a prior art remotely sensed image 20 of an observed geostationary satellite 21 against a star background 22 suitable for use in the prior art method 10 described and shown in Figure 1. The image 20 is shown is a greyscale image with dimensions of circa 750 pixels by 600 pixels. The majority of the image 20 is black indicating that no energy was sensed by the wide field of view sensorthat obtained the image originally. Within the image 20 there is a single point like object covering a single pixel representing the geostationary satellite 21. It is shown as a bright white pixel. At various other locations in the image 20 there are visible star trails 23 of varying degrees of brightness. Each star trail 23 is shown as a substantially linear streak extending in the region of 25 pixels in lengths. These trails 23 represent individual stars for which relative motion exists between the geostationary satellite 21 and the star, hence across the exposure time for the image 20 the stars appear as star trails 23 moving across the image 20. It can be seen that the image 20 contains an abundance of star trails 23 owing to the field of view over which the image 20 was sensed. This renders the image 20 suitable for input the method 10 as described for Figure 1. Figure 3 provides an illustration of a remotely sensed image 30 of observed geostationary satellites 31, 32 against a star background 33. This image 30 is not suitable for use in the prior art method 10 described and shown in Figure 1. The image 30 is shown in grey scale form with pixel wise dimensions of circa 1400 by 1000. The image 30 has been obtained by a remote sensing means having a much narrower field of view compared to that which obtained the image of Figure 2. Such a remote sensing means may comprise an amateur telescope and digital camera. For this reason, the image 30 reveals only one star trail 34 clearly. This star trail 34 is shown substantially as a line of contiguous white pixels spanning circa 1200 pixels in length. The remainder of the image 30 is dark. The presence of only a single star trail 34 is what renders the image 30 fundamentally unsuitable for plate solving techniques available in the prior art. There is simply not enough information in the image regarding background celestial objects to allow accurate determination of the pointing direction of the remote sensing means, and thereby the image 30 does not allow the true position in World coordinates of the geostationary satellites 31, 32 to be determined. Figure 4 provides an illustration of an embodiment of a computer implemented method 40 of determining the position of a geostationary satellite against a star background. The method 40 is shown in a flow diagram format as would be implemented algorithmically within a computer system. In a first step a plurality of remotely sensed images of the geostationary satellite against a star background are received 41. The remotely sensed images have been obtained by an Earth bound remote sensing means, such as a telescope and digital camera, that has been aligned onto (i.e. is tracking) the geostationary satellite for the duration of the exposure time of the image. The images have been taken at different times and may comprise a greater number of images taken across hours of observation. Hence in such images, the geostationary satellite will appear static with relative motion of the star background causing the stars in the image to appear as trails or streaks across each image. In a second step 42 of the method 40, an image comparison operation is applied to the plurality of remotely sensed images to obtain their relative pixel offsets. The image comparison operation is a phase correlation of consecutive pairs of images, such that pixel offsets aligning each pair of images can be determined. The correlation image (or map) resulting from each image comparison operation has a Gaussian weighting applied that is centred on an anticipated relative pixel offset predetermined by known or modelled movement of the star background between successive images. In a third step 43 of the method 40, an image mosaic of the remotely sensed images is formed by aligning the images according to the determined relative pixel offsets. In this step any overlapping pixels between images are co-added to improve signal to noise ratio of the image mosaic. The image mosaic contains a far greater number of star trails than any single remotely sensed image alone. Hence in a subsequent step 44 or the method 40, it becomes possible to extract a plurality of pixel locations of background celestial object trails from the image mosaic. This step is performed by applying a clumping operation to the image mosaic that identifies pixels having pixel values greater than three times the standard deviation of the image. Of these pixels, those that are present in contiguous groups exceeding a group size greater than N pixels are determined to represent startrails. The number of pixels N may be pre-calculated, for instance by estimating the spread function of a star in the image and using the frame integration time of the remote sensing means and knowledge of relative motion between the background celestial objects and the observed celestial object to estimate the trail size anticipated. Weighted averaging of these groups is used to determine a centroid pixel for each group that represents the location of the starthat generated the star trail. It is these pixel locations that are passed into a plate solving operation 45 in a subsequent step of the method 40. The plate solving operation 45 compares the pixel locations of star trails to known locations of stars in the World coordinates right ascension and declination. The relative positioning of the pixel locations compared to the relative positioning of locations of known stars determines a transformation between pixel locations in the image and right ascension and declination. The transformation corresponds to the centre pixel of the image (or pointing direction of the remote sensing means) which of itself, is approximately directed onto the geostationary satellite being tracked. Hence the world coordinates of the geostationary satellite can be output in the subsequent step 46 of the method 40. Figure 5 provides an illustration of an embodiment of an image mosaic 50 generated according to the method 40 described and shown in Figure 4. The image mosaic 50 covers a pixel-wise space of in excess of 7500 pixels by in excess of 1800 pixels. It is comprised of thirteen separate images 51 aligned together to form a single mosaic 50. The image mosaic 50 is shown in grey scale with the majority of the image shown in black, an effect of the coaddition utilised which has reduced noise relative to signal in the image 50. The alignment of separate images 51 has produced a mosaic that extends diagonally across the pixel space. Shown in the image mosaic 50 is the location of geostationary satellites 52, 53, repeated across the mosaic. This is an artefact of the image comparison process and alignment process that produced the image mosaic. Also shown in the image mosaic as bright substantially linear trails of pixels are star trails 54. There are eight such trails 54 clearly visible across the mosaic. This is a vast improvement in comparison to Figure 3, and therefore provides opportunity for the image mosaic 50 to substitute fora wider field of view image as an input to a plate solving operation. As an additional feature in the image mosaic 50, the centroid pixel location 55 associated with each star trail 54 is shown with a 'plus' marking. These are locations 55 extracted from the image mosaic for inputting to plate solving techniques. Figure 6 provides an illustration 60 of the position of geostationary satellites as determined by the method 40 of Figure 4. The positions of a first geostationary satellite and the position of a second geostationary satellite over time is represented by respective pluralities of data points 61, 62 in the figure, with each data point having a right ascension and declination in degrees. Each data point 61, 62 has been generated by applying the method 40 shown and described for Figure 4 to a plurality of remotely sensed images of the geostationary satellites for a particular time period. The illustration 60 shows how the method can be used to track the positions of satellites over time using relatively low field of view sensor systems that are cheaper and more widely deployable than larger conventional space domain awareness systems.

Claims

1. A computer implemented method of determining the position of an observed celestial object using remotely sensed images of the observed celestial object against a celestial background, the method comprising the steps of:a) receiving from remote sensing means a plurality of remotely sensed images of an observed celestial object against a celestial background, the images being aligned onto the observed celestial object and having been sensed consecutively in time;b) applying an image comparison operation to the plurality of remotely sensed images to obtain their relative pixel offsets;c) aligning the remotely sensed images according to their relative pixel offsets to obtain an image mosaic;d) extracting a plurality of pixel locations of background celestial object trails from the image mosaic; and thene) comparing the plurality of pixel locations to a library of known celestial object reference frame positions, to obtain a reference frame transformation that transforms pixel locations in the image mosaic to positions in the reference frame of the known celestial objects, and determining therefrom the position of the observed celestial object.

2. The computer implemented method of claim 1, wherein the step of applying an image comparison operation comprises:a) applying an image correlation operation to each consecutive pairing of images and in each case generating a correlation image of correlation scores; and thenb) identifying as the relative pixel offsets for each consecutive pairing of images, the pixel offsets corresponding to an optimum value of the correlation score in the respective correlation image.

3. The computer implemented method of claim 2, wherein the image correlation operation is a phase correlation.

4. The computer implemented method of any one of claims 2-3, wherein the optimum value of the correlation score is the maximum value of the correlation score.

5. The computer implemented method of any one of claims 2-3, wherein the step of identifying relative pixel offsets for each consecutive pairing of images comprises:a) applying a weighting function to the correlation scores to generate weighted correlation scores, the weighting function being biased towards a predetermined pixel offset; and thenb) identifying as the relative pixel offsets, the pixel offsets corresponding to an optimum value of the weighted correlation scores in the respective correlation image.

6. The computer implemented method of any preceding claim, wherein the step of aligning the remotely sensed images according to their relative pixel offsets comprises co-adding the images.

7. The computer implemented method of any preceding claim, wherein the step of extracting a plurality of pixel locations of background celestial object trails from the image mosaic, comprises the step of:a) Applying a clumping operation to the image mosaic to identify a plurality of contiguous groups of pixels having pixel values greater than a background threshold value, and selecting as the pixel location of each background celestial object trail the centroid pixel of each contiguous group.

8. The computer implemented method of claim 7, wherein the background threshold value is a multiple of the standard deviation of pixel values in the image mosaic.

9. The computer implemented method of any preceding claim, wherein the step of comparing the plurality of pixel locations to a library of known celestial object reference frame positions to obtain a reference frame transformation, further comprises comparing the relative pixel values of the pixel locations to relative brightness values of the known celestial objects to obtain the reference frame transformation.

10. The computer implemented method of any preceding claim, wherein the celestial background is a star background and the background celestial object trails are star trails.

11. The computer implemented method of claim 10, wherein the step of comparing the plurality of pixel locations to a library of known celestial object reference frame positions to obtain a reference frame transformation, comprises applying a plate solving operation to the plurality of pixel locations.

12. The computer implemented method of claim 11, wherein the observed celestial object is a geostationary satellite.

13. The computer implemented method of any preceding claim, further comprising the steps of:a) comparing the reference frame position of a background celestial object trail to a library of known background celestial object magnitudes and extracting therefrom a magnitude value associated with the background celestial object trail;b) calculating a conversion factor for converting between the magnitude value and pixel value associated with the background celestial object trail; and thenc) applying the conversion factor to a pixel value associated with the observed celestial object, thereby determining a magnitude value for the observed celestial object.

14. The computer implemented method of any preceding claim, wherein the reference frame transformation transforms pixel locations in the image mosaic to right ascension and declination.

15. A computer program containing code which when executed performs the computer implemented method of any preceding claim.

16. Computer readable storage media comprising the computer program of claim 15.

17. Apparatus for remotely sensing the position of an observed celestial object against a celestial background, comprising remote sensing means for consecutively sensing aplurality of remotely sensed images aligned onto an observed celestial object against a celestial background; and data processing means configured to execute the method of any one of claims 1-14.5       18. A space domain awareness system comprising the apparatus of claim 17.

19. A star tracker for a spacecraft comprising the apparatus of claim 17.

Citation Information

Patent Citations

  • Method and system to detect orbiting objects

    WO2020191427A1