DE10127399A1 - Method and device for the autonomous navigation of satellites - Google Patents

Abstract

The invention relates to a method and a device for surface-based navigation using optoelectronic image displacement monitoring for the autonomous navigation of satellites. The inventive method is a landmark navigation method which uses a new approach to optoelectronic navigation, involving image displacement monitoring instead of landmark monitoring. The invention also relates to a device for carrying out the method. The navigation principle is based on the fact that the image displacement in the focal plane of an image detection device situated on board and oriented towards the planets detects the relative displacement of the satellite and the surface of the planet observed. The momentary image displacement in the focal plane is defined by the position and situation of the satellite in relation to the displacement of the surface of the planet which depends on the form and the rotation of the planets. In this way, both the position and information relating to the orbit can be determined from a sequence of observed images, with a minimum of a priori information.

Description

The invention relates to a method and a device for surface-based Navigation with the help of optoelectronic image motion tracking to the autonomous Carrying out the navigation of satellites. As satellites in the sense of this doctrine are artificial missile orbiting a celestial body in an orbit understand.

Autonomous satellite navigation in the sense of a "real-time provision of satellites orbit data and the current position (spatial orientation) exclusively with on-board resources "is a prerequisite for (semi-) autonomous satellite operations such as scheduling, data allocation and orientation of the on-board devices or maintaining the satellite orbit.

For some missions, autonomous satellite operations are due to the nature of the satellite Mission itself required. This is too far for the flyby on planets or flights distant planet of the case where either the spaceship without contact with the Bo station or the signal delay is too large for a controller from there.

For satellite constellations of any height, the entire process will orbit Railway determination and control is one of the most important operating cost factors. At the same time, this also represents a significant risk factor in which any Be Drives and any failure of the ground system damage the constellation or could destroy. Autonomous operations can therefore be the operating costs be reduce considerably.

Accurate instant orbit data and data from onboard instruments (Images, readings, etc.) can be simultaneously available to the instrument data clearly referenced. In addition, the measurement requires less accuracy as in systems that work with old data because the data is losing accuracy, when extrapolated to future dates.  

For accurate orbit maneuvers and precise alignment of onboard instruments Greater accuracy of navigation results based on old data and which have to be extrapolated to later times, requires that this be the Meet real-time processing needs. For real-time systems, one is highly accurate predictions of orbit of lesser importance.

For Earth-oriented missions, the most suitable onboard navigation sources are the ones GNSS services (GNSS - Global Navigation Satellite Systems) existing systems such as GPS or Glonass, future systems such as Galileo or the ground-based DORIS. But it should be emphasized that such concepts are not completely autonomous systems because they rely on the availability of the navigation satellites that provide the required data. except GNSS services are not available in interplanetary space.

Therefore, the best sources of data for fully autonomous navigation are natural ob Sunsets, planets, etc. Observing these objects and, if possible, their edges, is a basis for determining the relative position from satellites. Such navigation systems require input data of several On board sensors. The navigation performance is enhanced by visibility and observability the reference celestial body, the amount of available a-priori reference data (data, the were collected before the start of the mission) such as ephemerides (tables in which the Position of a star in the sky or its place within its orbit around a zen tral body is indicated for a regular sequence of times), observational and shape models for planet edges. The alignment error between un Different instruments is another important source of error and makes a ge accurate onboard calibration necessary.

For satellites in orbits is therefore for device technically simple and robust Sy stem solutions the use of the next object - the surface of the circled Planets - the most promising solution.  

The traditional approach to surface-based navigation is the landmarks Navigation. She uses a camera aboard the satellite to record upper surface images, an onboard database of country position and design data marken, an on-board image processing system for the detection and determination of Landmark positions on the captured image and an onboard computer, to process the results of the measurement. The position and location of the satellite in an inertial system (coordinate system in space and time, in which the Newtonian Axioms of mechanics apply without inertial forces, such as centrifugal or corio liskräfte, occur. In practice, an inertial system can be considered as a reference system the one anchored in the Milky Way. Furthermore, each is straight and uniform on the other hand moving system likewise an inertial system) can with suitable Algorithm can be determined from the information about planets: Geometry, Form, Gra Vibration and rotation model as well as coordinate positions of all determined Landmar ken on the planet. Normally, landmark information is in the stage the mission preparation collected and are on board as a landmark database before The pictures of the landmarks (in raster or vector form) together with their own Coordinates included.

The effective use of such systems depends on the following general factors:

• - Content of the onboard landmark database
• - accuracy of the reference data, d. H. Landmark coordinates
• Robustness of landmark detection algorithms over variable Appearance of landmarks

The first factor plays an important role in the length of the step (time inter vall between two measurements) and the convergence time of the navigation algorithms. For example, there are two landmarks per revolution (as in LEO missions) and a step size of a half-round Konver Genzzeiten of typically 15 rounds.  

This is far from sufficient for some short-term operating requirements. Large step sizes also require high accuracy in the definition of Landmark positions. To collect high quality image data and the Koordi It is a non-trivial task, even for the planet He and will be even more difficult for other planets. An enlargement of the land brand database makes a big effort in the stage of data preparation necessary and requires large onboard storage capacity and effective processing facilities.

The robust and accurate measurement of landmark position becomes a non-trivial pro for changes in observation conditions (different exposure angle) and surface changes (seasonal changes of landmarks, low tide and tide for coastal processes, etc.). To overcome this problem, entwe on-board image correction algorithms using a-priori information or an extended set of landmarks for different situations become. The complexity of image processing increases system cost and complexity Incorrect or incorrect results. The need for a large amount of exak A priori information about landmarks and reliable, robust image processing Having on-board treatment is the main obstacle to a cost-effective Landmark navigation solution.

The patent US 6023291 discloses a method and a device for navigating known from satellite, where the images of landmarks and the edge of Pla be used.

However, this method has the following disadvantages:

• 1. Additional storage capacity is available for Landmar's large onboard database required;
• 2. Additional computational effort is taken for the onboard correction of Pictures necessary;
• 3. The exact coordinates of landmarks are required, the extraction and Preparation is very expensive.

The patent US 4730798 discloses a method and a device for navigating known by satellite, where the measurement of angular diameter of the planet used for determining the height of the satellites.

However, with this method it is not possible to use other orbital parameters determine. Moreover, the exact information about the shape of the planet required.

Based on the described disadvantages of the prior art, the It is the object of the invention to provide a method and a device for autonomous real-time Satellite navigation, independent of data from external sources, eg. B. a ground station, and are out with a minimum of a priori information come, converge quickly and are robust to surface defects.

According to the invention, the object is achieved by a method having the features of claim 1 and solved by a device having the features of claim 3. before Advantageous embodiments of the invention are described in the subclaims.

The teaching according to the invention discloses an alternative method for landmarks. Navigation, using a new approach optoelectronic navigation uses image motion tracking instead of landmark tracking and a Vorrich for the implementation of the procedure. This procedure also requires the observation tion of the planetary surface with an onboard image capture device, but uses other, more general data.

The navigation principle is based on the fact that the image movement in the focal plane of an on-board, directed at the planet imaging device the detected relative motion of satellite and observed planetary surface. The eyes Immediate image movement in the focal plane is due to the position and position of the satellite defined relative to the movement of the planetary surface, of the shape and the Rotation of the planet depends. This allows both the location and Informatio via the orbit from a sequence of observed images with a minimum be determined from a priori information.  

In particular, the proposed navigation method does not require any in advance saved landmark database. The simple comparison of consecutive, in short intervals of recorded images also does not depend on the light conditions or seasonal changes.

Image motion tracking is based on two-dimensional correlation analysis and he allows correlation accuracies at sub-pixel level. The special advantage spectral Image information processing lies in the independence of individual pictorial feature len. This type of image analysis is based unlike landmark tracking alone on the overall pattern of the picture. This makes correlation methods extremely robust against image disturbances and highly expedient if complex image structures lie, as is the case with satellite imagery.

Due to the high update rate, the method provides accurate In Formations to orbit and location already in the first revolutions.

The application of the proposed autonomous navigation system may be of particular interest in connection with:

• - Emergency or replacement navigation systems for remote sensing satellites Using the conventional, on-board remote sensing anyway sensors as an image capture device,
• - Intelligent image acquisition devices as navigation and position sensors for small satellite,
• - Data fusion systems (data fusion: merging of diverse data to determine the system status),
• - Emergency or replacement navigation systems for planetary missions.

Data fusion means the sharing of measurement data from different sensors for the determination of the orbit or orientation of satellites, such. Star sensors, horizon sensors, telemetry data, etc. The more data there is, the more more accurate and robust is the determination.  

Replacement or backup systems are necessary in cases when the main navigation system or positioning systems have partially or completely failed. Here you can the metrics of alternative backup systems are the only ones that you can use for the Fulfillment of missionary tasks.

The navigation approach is based on the dependence of the image movement in the focal plane ne an onboard image acquisition device of the relative movement between satellite and Planet's surface. In general, this movement is due to the variable Satellite altitude (elliptical orbit) or different surface weights movement (relative movement due to the planetary rotation) at different widths not constant. The onboard image acquisition device detects this relative movement between satellite and planet in the surface image projected onto the focal plane. These Picture movement also does not stay constant and depends additionally on the moment position of the satellite.

An advantage of the device according to the invention is that only a single Sensor is needed. A compact, robust image capture device is less sensitive against unwanted mechanical deformations and forces as an An order of multiple sensors, resulting in more reliable and accurate data.

Furthermore, the method does not require special patterns, resulting in the Bildbear simplified, since any type of planetary surfaces is equally suitable and the method tolerant of observation conditions, such as light, jah reszeitliche changes, etc., is.

The method according to the invention comes with a minimum of stored a priori Information out. No landmarks need to be saved in advance where there must be no storage space for a landmark database on board. Possibly inaccurate data on landmarks have therefore no influence on the Navigation results. When using the device according to the invention is kei ne external processing device needed.  

Furthermore, it is unnecessary for the execution of the navigation algorithm, in advance over to have special knowledge of the planetary form. The determination of the situation gives approximately good results with approximate knowledge of the planet shape. The Determining both position and location is the basis for a complete one autonomous operation of the satellite.

For the implementation of the method according to the invention is a fast Bildverarbei required. To obtain real-time data on the shadow side of the planet An infrared camera must be used. The planet must have a recognizable and stable surface structure, which is not the case for all planets of the solar system the case is. The work area is limited in the situation insofar as the Pla net must be in the field of vision of the image capture device.

Under image motion, the movement of sections of the image is overflown NEN surface in the image plane of the image acquisition device understood on the Rela the movement between the satellite and the overflown surface is due to makes noticeable that the same image section in successively aufges taken pictures in different places. The paths of this picture Sections or image blocks are determined by the spatial relationships between the image level and the observed surface area. Therefore, these Pfa de, or these image movements, the information about the 3D movement of the Bildfas sungsgerätes relative to the surface. The image movement thus allows conclusions on the Movement of the satellite over the surface overflown. This information can can be obtained from the measured image motion to later for higher quality Al gorithms can be used for position and position determination.

A combination of this concept with conventional landmark navigation is also possible. As a result, on the one hand, the accuracy can be improved and on the other hand, the landmark database to be carried on board and thus the costs be significantly reduced. The landmark measurements also support the Determination of the sixth parameter of the orbit, the right ascension Ω of the rising node (Keplerian orbital parameters). The high speed of the picture  Motion Tracking allows the implementation of landmark tracking where can be improved by the navigation performance because while passing a landmark or a group of landmarks through the observation area of the Image capture device several updates of the information about the orbital can be done.

With the knowledge of all elements of the orbit, which can be determined autonomously transformed the location into roll-pitch-yaw (RPY) coordinates into the inertial system become. Therefore, such a mixing concept even allows a completely autonomous Na vigation and orientation in the reference inertial system.

The navigation and attitude data provided by the image motion analysis may be for performance enhancements to existing onboard navigation systems with whose sensors work, are used.

With reference to the following drawings, the invention will be explained in more detail. Show

Fig. 1 shows a satellite at different positions of its orbit

Fig. 2 shows the method steps of the navigation method of the invention

Fig. 3 shows the basic structure of the device according to the invention

Fig. 4 shows the different imaging of the same surface properties in different layers / positions of the satellite

Fig. 5 shows the movement of image blocks in the image plane of the imaging apparatus

Fig. 6, the reference coordinate system of the imaging apparatus

Fig. 7 shows the principle of image motion tracking

Fig. 8 shows the definition of the navigation angle λ, χ, β

Fig. 9, the image block tracking based on the image movement between different positions

FIG. 10 is an illustration of the term "focus of contraction"

FIG. 11 is an illustration of the term "focus of expansion"

Fig. 12 shows the principle of orbit estimation

Fig. 13, the principle of correction of the raw layer

Fig. 14 shows an embodiment of the device according to the invention

Fig. 1 shows a geometric interpretation of the observation conditions. If in the on the orbit ( 2 ) about the planet ( 3 ) moving satellites ( 1 ) brought image acquisition device at defined time intervals takes several pictures so that successive images have common, overlapping image parts, the image movement as a path of movement of image parts or blocks along the focal plane of the image capture device.

The overlapping areas which are sequentially detected by the field of view of the image acquisition device at the respective time of the overflight are represented by the light cones ( 4 , 5 , 6 ). Due to the planetary rotation, the detected areas ( 7 , 8 , 9 ) of the planetary surface constantly move in the corresponding direction. As a result, the captured images in other areas ( 10 , 11 ) overlap than would be the case without lateral movement due to planetary rotation.

The satellite ( 1 ) takes a picture in the first position and flies on. During this flight, the imaged area ( 7 ) moves with the planetary surface. In the second position, the satellite ( 1 ) images the second area of the surface ( 8 ). The water area ( 8 ) has an overlapping part with the first imaged area ( 7 ). The second area ( 8 ) travels more slowly than the first area ( 7 ), because the speed of the surface depends on the latitude of the area. Due to this delay, the third imaged area ( 9 ) is shifted less relative to the second area ( 8 ) than the second area ( 8 ) relative to the first area ( 7 ). For the satellite ( 1 ) in elliptical orbit ( 2 ), the speed changes according to the altitude. Therefore, images taken at equal intervals will have different overlap proportions. In addition, the satellite rotation is reflected in a shift of the areas.

It will be appreciated that for a given observation sequence and planet ( 3 ), the overlapping image portions may be different for different corresponding positions of the satellite ( 1 ) on the orbit ( 2 ).

The overlap between two images defines the two images in common Image content. However, the process is not the image content itself, but the posi tion of the overlapping areas on the image sensor prevail.

The overlap of successive images, along with a minimum of a priori information, may be used to determine in real time the path of the satellite ( 1 ) about the planet ( 3 ) and the position (orientation) of the satellite.

The required a priori information is limited to:

• - Planetary parameters such as gravitational and rotational model and a rough Be writing the form for the determination of the situation;
• - Image capture device features such as focal length, pixel count and size, and on order of the image sensor, geometric calibration coefficients.

The gravitational model of the planet is used for the description and calculation of the Satellite movement needed around the planet. The movement model is necessary for most methods for calculating the orbit. The rotation model needs one, if the surface of the planet is used as a reference for navigation. The description of the planetary shape is necessary for the determination of the relative orientation tion of the satellite. Minimum data required is radius and flattening coefficient of the planet.

The navigation method consists of several main stages, as shown in FIG .

In the first stage, images of the planetary surface ( 12 ), which overlap at least in subregions, are continuously recorded by an on-satellite image acquisition device, whereby certain sections of the images (hereinafter called image blocks) have the same contents.

These images are further processed in the image motion tracking stage ( 13 ). The movement of image blocks between successive overlapping images in the image plane of the image acquisition device is tracked. The position vectors belonging to one and the same picture block, which describe the movement of the picture block between the times of the recording of the viewed pictures, obtained for different times, are obtained for all the picture blocks considered.

In the step of image motion analysis ( 14 ), the measured quantities required for the orbit estimation, ie the navigation angles, and the raw sheet are obtained. The image motion analysis ( 14 ) determines information about the 3D motion of the satellite from the position vectors of the considered image blocks. This information relates to navigation angles and the satellite's baseline. The navigation angles describing the movement of the satellite relative to the planetary surface depend on the position of the satellite. A series of navigation angles clearly determines the movement of the satellite and, accordingly, the orbit. The blank sheet indicates the orientation of the satellite relative to the planet and the orbit. For accurate determination of the position of the satellite, these raw layer data must be corrected in a later process step.

In the stage of orbit estimation ( 16 ), using the estimation algorithm, such as Kalman filter or batch estimation algorithm, based on orbital and orbit models, from the navigation angles, satellite position and velocity data, e.g. B. in the form of the current Keplerian order track parameters ( 15 ) determined.

The next process step is the correction of the raw layer ( 17 ). Therein, correction angles are calculated from the orbit parameters ( 15 ). These correction angles are then used to determine the position of the satellite in the roll-pitch-yaw (RPY) coordinate system ( 18 ) from the previously estimated blank.

This algorithm provides an estimate of five elements of the orbit (except right ascension Ω of the ascending node) and the location in the RPY reference coordinate system. These data are sufficient for:

• 1. Autonomous control and maintenance of the orbit (eg overcoming of Height disturbances on LEO satellites and station hold);
• 2. Nadir alignment of onboard instruments.

Fig. 3 shows the structure of an apparatus for carrying out the method. The inven tion proper device consists of the main components image acquisition device ( 56 ), high-speed processing unit ( 57 ), data processing system ( 58 ), memory for a priori data ( 59 ) and system interfaces ( 60 ).

An on-board image capture device ( 56 ) captures and forwards surface images at time intervals defined by a high speed processing unit ( 57 ).

The high-speed processing unit ( 57 ) performs the image motion tracking by storing the image data coming from the image capture device ( 56 ), extracting and storing image blocks, and then performing a correlation analysis of the image blocks in pairs, resulting in the movement of the image blocks in the image Image plane of the image acquisition device ( 56 ) is known from the position vectors of the image blocks.

The position vectors of the image blocks are forwarded to a digital data processing system ( 58 ) which performs the determination of navigation angles, the estimation of the Keplerian orbital parameters and the position of the satellite based on stored a-priori data ( 59 ).

The result data of the performed method, ie the Keplerian orbit parameters and the position of the satellite, are transmitted to system interfaces ( 60 ). These system interfaces ( 60 ) allow the interaction between the device according to the invention and the main board unit of the satellite by the exchange of commands and data.

The illustration in Fig. 4 shows the different mapping of the same Oberflächenob objects for different positions of the satellite. The image capture device ( 20 ) flies over the observed surface ( 21 ). An image sensor ( 22 ) is located in the image plane of the image capture device. The image sensor ( 22 ) images two images from the surface, the first image at time t ₀ and the second at time t ₁ . At the first time, the surface objects ( 23 , 24 ) appear on the image sensor. At the appropriate time interval, these objects can also be projected onto the image sensor at the second time.

FIG. 5 shows the image movement in the image plane between two successive shots.

Between the two times, the images of these surface objects move through the image plane from positions ( 25 , 26 ) to positions ( 27 , 28 ).

FIG. 6 shows the reference coordinate system ( 29 ) of the image acquisition device. The reference coordinate system ( 29 ) of the image acquisition device is associated with the arrangement of components of the image acquisition device. The U-axis is the optical axis of the image acquisition device. The VW plane intersects the U-axis upright in the optical center of the lens. V and W axes are parallel to the edges of the image sensor and the W axis points mostly in the direction of image motion.

The occurrence of the image movement is also shown in FIG . The image movement between two time points is determined by a set of positions of image blocks in the reference coordinate system of the image acquisition device at the two time points t ₀ and t ₁ .

The vectors q are the position vectors of image blocks in the Referenzkoordinatensy system of the image acquisition device. A q-vector passes through the lens center (a zero point of the image acquisition device's reference coordinate system) and the position of the image block on the image plane. The first vector q ₀ belongs to the first time t ₀ and the second vector q ₁ to the second time t ₁ . The tracking of the image blocks starts from defined positions.

The surface image moves during the movement of the satellite over the Planten tenoberfläche constantly in the image plane of the image acquisition device. The image motion tracking algorithm described in more detail in FIG. 7 determines the sequence of image acquisition, finding and reading out of image blocks to be matched, and their matching for accurate position measurement. Further in the text, the term frame refers to a plurality of image pixels in the image plane regardless of the number and arrangement of a plurality of image sensors. The image block is a small part of the fullscreen.

Referring to Fig. 7, the process of image motion tracking is shown in block diagram form. The algorithm realizes parallel, ie temporally overlapping, stepwise image motion tracking. A line of pursuit consists of several small steps of image tracking.

A line begins with the capture of the first frame, then reads out several reference image blocks from defined positions, saving the image blocks, extrapolation and prediction of the motion of the image blocks on the image sor, acquisition of subsequent frames, reading of several current image blocks from extrapolated positions and alignment with the reference image blocks, Ak cumulate small measured block movements, assignment of current picture blocks as reference image blocks and repeat this process until the edge of the Image sensor is reached. This means the fullest possible tracking of each image block from one edge of the image sensor to the other.

The size of the steps depends on the degree of image distortion caused by the sphäri planet, the location of the satellite and image acquisition device distortions can be. This is due to the external data about the position and location of the Satellites determined automatically.

The single tracking process (EVV) consists of tracking many paths of Image blocks. He is synchronized in time. The inputs of current external Da  The position and position of the satellite are also synchronized. This he allows a more accurate prediction of the image block movement.

During each individual tracking process a tuple with all necessary data fields is generated:

Description of the fields Code of fields Start positions of image blocks SPB0, SPB1 End positions of image blocks EPB0, EPB1 End positions of reference image blocks EPRB Image data of reference image blocks RBD Start time of the pursuit SZP0, SZP1 End time of the persecution EZP0, EZP1 Number of track steps AVS0, AVS1 Estimates for navigation angles and rawlays MW Estimates for orbit parameters and RPY attitude ULD intermediate data ZDATA

EPRB contains integer positions from which the reference image blocks are read were.

EPB0 and EPB1 contain measured positions of the tracked image blocks with Sub pixel resolution. Some fields are doubled, they contain data from two levels of individual tracking process. The image blocks are independent of each level followed by other stages. This means that the stages are temporally separated. The last Image blocks from the first stage and the first image blocks from the second stage read out from the same frame. This also means the fields EZP0 and SZP1 be come the same value.

An EVV is completed when both stages are completed. Values in the fields AVS0, AVS1 are needed to predict image frame tracking error parameters. The Error parameters are needed during the estimation of orbit parameters. The entire image tracking is done in such a way that multiple EVV simultaneously be able to walk. This parallel tracking is necessary to get an updateinter  vall the orbit filter, which is shorter than the time of the run of the Image blocks from one edge of the image sensor to the other.

The tuples of different EVV are in the series of single tracing Transactions (REVV). Each tuple is automatically initialized when the time is up tervall from the start of the last tuple the default value for the update exceeds the monitoring interval of the orbit filter.

The stages of an EVV are ended when the tracked image blocks the edge of the image reach sensors or the last adjustment of image blocks does not succeed. The end times Therefore, the persecution can not be equidistant. The REVV is constantly going for a long time analysis in memory. The old tuples are automatically deleted.

After the start ( 30 ), the algorithm waits for the next timestamp ( 31 ) and then the first frame is detected ( 32 ). The first tuple in the REVV is created ( 33 ). A set of image blocks is read from the designated positions from the first frame ( 34 ) and stored as reference image blocks in field RBD of the first tuple of the REVV. The readout positions are stored in fields SPB0, EPB0 and EPRB. The start time (defined by the timestamp) is entered in the fields SZP0 and EZP0. The field AVS0 becomes zero.

Then the main loop of the algorithm begins. At the next timestamp ( 35 ), the positions of image blocks are extrapolated from field EPRB ( 36 ). It uses data about the position and position of the satellite. The position and position of the satellites at the previous times are taken from the field ZDATA. The interval for the prediction of the image block movement is denoted Δt. The interval .DELTA.t is selected so that during this interval, the position and position of the satellites can be determined with sufficient accuracy. The Δt is automatically shortened for large image distortions. The degree of distortion of moving picture blocks relative to the original rectangular shape is calculated and compared with a preset threshold Λ ( 37 ).

If the distortion level is low enough and not disturbing the image correlation, the next timestamp is expected ( 35 ) and the prediction is repeated. If the distortions exceed the predetermined threshold Λ, the current frame is considered as the last frame where the image correlation is still possible. This is also the case when the image blocks in Δt exceed the edge of the image sensor ( 38 ).

Then, at the present time, a frame is detected ( 39 ). The current image blocks are read out from rounded extrapolated positions ( 40 ). These current image blocks are then compared to the prestored reference image blocks from field RBD to measure the displacement between them ( 41 ). If the match succeeds ( 42 ), the current processed tuple in the REVV is updated ( 44 ). The field EPBn receives the sum of measured displacements and rounded extrapolated positions of image blocks minus the value of field EPRB and plus the full value of EPBn. This means that after the update, the field EPBn contains the current positions of the first picture blocks at the current time. The field EPRB receives the rounded values of EPBn. The values of EPBn are rounded and the image data from field RBD is replaced with image data from the new positions. The EZPn field receives the current time value. The value in the field AVSn is increased by one. The field ZDATA receives the last used estimate of position and position of the satellite.

If the current image blocks reach the edge of the sensor, or if image frame matching was unsuccessful, the current stage of the current EVV is terminated and marked accordingly ( 43 ). In the second case, the fields are not updated. When the first stage in EVV is finished, the second stage starts from the same frame and therefore from the same time.

Then all unfinished tuples are checked in the REVV ( 45 ). For each EVV, the image block movement at the current time is moved out of field EZPn of the current tuple. When the image blocks reach the edge of the image sensor, the stages of tested EVV are terminated without updating. If the image distortions exceed the given threshold Λ, the corresponding EVVs are subjected to the same process with measurement of the image shift and update of the tuple. This will use the same frames for different ready EVV ge and greatly reduces the frame rate of the image acquisition device.

A new tuple is created ( 47 ) if the time interval between current and start time of the last EVV (field SZP0) exceeds the predetermined updating interval of the on-track filter. Then the oldest unfired EVV is selected as the current EVV to process ( 47 ) and a new loop begins. If there are no uninterrupted EVVs, the algorithm waits in the loop until a new tuple has been created.

This algorithm requires implementation of the image block displacement measurement in real time. During the flight observation pauses may occur. That can happen if the texture of the surface is not recognizable or the lighting is not enough to allow the image correlation (night side of the planet). However, the breaks can be used to manipulate data that is higher Measuring rate are recorded. This allows for an increased refresh rate of the order runway estimator and thus the elimination of performance degradation by the Observation breaks. The algorithm must be modified as follows. All necessary data are processed and stored in the tuple of the REVV. The necessary data are all intermediate image data and their attributes. A parallel lau fender algorithm handles the processing of the data. The algorithm monitors the REVV from the first tuple and starts editing when the image editing unit is free. Despite the large storage space required, the algorithm allows a flexible control of navigation measurements. The measurement of the image shift is performed by the 2D correlation of two image blocks, with the position of the Correlation maxima is determined with subpixel accuracy.

The degree of distortion of the image blocks is determined by the following rule: The maxima The sizes in V and W directions are determined for the extrapolated image blocks. The difference between this extrapolated and the original size of image blocks presents the degree of distortion.  

The image motion analysis determines the information about the 3D motion of the satellite from measured position vectors of picture blocks. Here are navigation angles and Rohlage of the satellite. Navigation angles depend on the position of the satellite pending. A series of navigation angles clearly determines the movement of the Sa and accordingly the orbit.

Fig. 8 shows the definition of three proposed navigation angles. The satellite is in orbit and picks up three consecutive and overlapping images at times t ₀ , t ₁ and t ₂ ; where: t ₀ <t ₁ <t ₂ . The vectors P, P and P describe the positions of the satellite at these times. The lower index contains the time of frame capture. After capturing the full screen, the position vector remains linked to the planetary surface. As the planet rotates, these vectors rotate with the planetary surface about the Z axis. At some point in time, these vectors occupy the new positions P, P and P. The upper index contains this time. The vectors P and P are rotated by the angles which are proportional to the time intervals t ₂ -t ₀ and t ₂ -t ₁ . The vectors P, P and P are in the orbit plane. But the vectors P, P and P are no longer in this plane. This also changes the angles between the vectors. Through detailed analysis one finds the set of navigational angles that describe the position of the satellite at three points in time and can be determined by the image motion analysis.

The first navigation angle λ is an angle between the vectors P and P. The Value λ is always positive.

The second navigation angle χ is an angle between two planes that passes through the vectors P, P and P, P form.

The third navigation angle β is an angle between the vectors B ₁ and B ₂ , where B ₁ = P - P and B ₂ = P - P. The value β is always positive.

The three navigation angles are related to the time t ₂ . At least three overlapping images are required to determine the navigation angles. The motion analysis algorithm consists of two parts. In the first part, the angles λ and χ can be determined and then β is determined.

Fig. 9 shows the geometric relationships. The determination of the navigation angle is based on an image movement model. This model gives the positions of moving image blocks on the second image q 1 | k as a function of a central and six rotation angles (Rohlagen):

This requires external parameters such as the altitude of the satellite h ₀ , h ₁ ; the radius of the planet R and fixed starting positions of image blocks q 0 | k. The rotation angles are the orientation of the satellite in the local coordinate system.

An optimization problem is solved by comparing the measured positions with model positions.

From this one estimates the optimal value:

This procedure is then repeated for pictures 1 and 2 and obtained

Estimates calculate two navigation and three rotation angles at time t ₂ .

The navigation angle λ is taken as ₀₁ . The angle χ is calculated with two vectors 1 | 01 and 0 | 12 from the formula

where M is a rotation matrix. The reason is that 1 | 01 and 0 | 12 are in a state or a position of the satellite, but the local coordinate systems below are different because they are also connected to planetary rotation.  

With other methods, the third navigation angle β is determined.

The motion lines of pixels intersect at a point on the image level, if the image acquisition device between two times only translational emotional. This applies to any surface shape.

When the motion vectors point to this point as shown in Fig. 10, the point is called focus of contraction ( 48 ). If the motion vectors, as shown in FIG. 11, point out of this point, the point Focus of the expansion ( 49 ) is called. If the position of the focus' (v ₀ , w ₀ ) is known, the vector of the self-motion of the image acquisition device in the reference coordinate system of the Bilfassungsfassungs between the times t ₀ and t ₁ can be determined.

In the case of contraction ( 48 ), the vector b

b = [v ₀ , w ₀ , f] ^T ,

In the case of extension ( 49 ), the vector b

b = [-v ₀ , -w ₀ , -f] ^T.

This vector indicates only the direction of movement of the image acquisition device in the reference coordinate system of the image acquisition device C ₀ . He is referred to as b 0 | 01. The vector b 1 | 01 in the reference coordinate system of the image acquisition device C ₁ can be determined by rotation transformation from b 0 | 01. For the times t ₁ and t ₂ , a pair of vectors b 0 | 12 and b 1 | 12 is calculated. The vectors b 1 | 12 and b 0 | 12 are determined for a specific position of the image acquisition device and form the angle β.

If the image capture device constantly changes its location, the image motion vectors no longer share a common point. Here is the rotation of position vectors used by image blocks at the second time until the image motion vectors on show a point.  

The REVV is constantly monitored. The tuples with finished EVV will immediately bear beitet. The determined values of navigation angles and roughness are Field saved.

The principle of orbit estimation is shown in FIG . The position of the satellite in the orbit is described at a time t ₂ by the state vector x (t ₂ ). The components of the state vector are parameters of a circulation model. Conventional orbit models exist, such as ideal six-parameter Keplerian orbits, models based on Kepler's parameters with perturbations (eg NORAD for the earth) or position / velocity models.

An orbital filter estimates the state vector of the satellite at one time based on a model of true satellite motion ( 52 ) in orbit and current measurements of that motion. The measured values must be dependent on the state vector. The measurement model ( 53 ) describes this dependence.

There are several approaches to estimating orbital parameters z. B. on the Least squares based estimator batch estimation algorithm and Kalman filter.

The estimation process is shown in general terms in FIG. 12. With instantaneous values of navigation angles, the state of the system is estimated. This is done by the correction of the extrapolated state with a correction value, which is generated from the measured values. The model of satellite motion ( 52 ) allows the state to be predicted from one point of time to another. The measuring model ( 53 ) determines the extrapolated values for the navigation angles. When correcting the state ( 54 ), the current measured values ( 50 ) (navigation angle) are compared with the extrapolated values. On the basis of this difference and the statistical parameters of measurement errors generated from the model of measurement errors ( 51 ), the correction values to the extrapolated state vector are determined in order to obtain the current estimated values of real-time parameters ( 55 ).

The REVV is constantly monitored for tuples with finished navigation angles. The Tuples are processed according to readiness. The estimated values are stored in the ULD Field saved.

In Fig. 13, the correction of the green sheet is shown in principle. The raw layer differs from the position in the roll-pitch-yaw coordinate system.

The reason for this is first the flattening and the deformation of the planet. Second Position is the planetary rotation, as this is the deviation from the position of the satellites depends on the orbit. With the data about the position of the satellite and the Deformation of the planet can be corrected to the basal layer.

The REVV is constantly monitored for tuples with finished blank data. The tuples are processed according to readiness. The determined values of the RPY position are displayed in the ULD field saved. The position data can be from both the orbit filter and also be delivered externally. Other at least required data are radius, Ab plating coefficient and rotation period of the planet. The accuracy of the correction The more detailed the shape of the planet is, the higher it is.

According to the invention can be used to carry out all the steps a compact ago direction, which works autonomously, ie without the use of foreign resources, built from standard components and therefore inexpensive to manufacture and which is able to communicate via standard interfaces with other devices such. B. to communicate with the main computer of the satellite. The device consists of ( Fig. 14)

• an image capture device ( 61 ) which projects the light from the observed surface onto the majority of photosensitive elements and produces a frame,
• an image buffer ( 62 ) which receives the full image,
• a computation unit 1 ( 63 ) which takes over all the necessary image tracking tracing calculations, updates the tuples in the REVV and controls a 2D correlator,
• a 2D correlator ( 64 ), which takes over the calculation of the 2D correlation function and determines the displacement of image blocks,
• a reference timer ( 65 ) generating the reference time stamp,
• a computing unit 2 ( 66 ) which performs the image motion analysis,
• a computing unit 3 ( 67 ) which determines the orbit parameters and RPY position,
• a REVV memory ( 68 ) containing the data from ongoing individual tracking operations,
• - An external interface ( 69 ), which connects the device with the Hauptbordrechenein unit.

The operation of the device according to the invention for autonomous satellite navigation can be summarized as follows:
The arithmetic unit 1 ( 63 ) is synchronized by the reference timer and performs the extrapolation of the image blocks based on the position and attitude of the satellite ( 70 ). The arithmetic unit 3 ( 67 ) supplies these data as correspondingly revised results of the orbit estimation or as external data from the external interface ( 69 ). The arithmetic unit 1 ( 63 ) updates the REVV content and creates new tuples if necessary. The frame data is read from the frame buffer ( 62 ).

The reference blocks stored in the REVV and read out of the frame buffer current picture blocks are forwarded in pairs in the 2D correlator. This returns the shift value for each pair. This value is used by the arithmetic unit 1 for updating the data in the REVV.

The image acquisition device ( 61 ) is controlled by the arithmetic unit 1 so that the image recording takes place at certain times. The exposure time is determined by the Re cheneinheit 1 to ensure the sharpness of the images. The frame is stored in the frame buffer ( 61 ). The recording times are entered in the corresponding fields in the tuple of the REVV.

The REVV memory ( 68 ) contains the REVV. The arithmetic unit 1 deletes the unusable and old tuples in the REVV. The arithmetic unit 2 ( 66 ) accesses the REVV memory and constantly monitors it for the completed tuples with ready measurements of the image movement. If there are such, the arithmetic unit 2 determines the Navi gationswinkel and Rohlage. For this purpose, the data about the position and position of the satellite ( 70 ) are used. The determined values are saved again in the REVV.

The arithmetic unit 3 ( 67 ) estimates the orbit parameters and corrects the raw sheet. This is done with the tuples with finished values of navigation angles. The results of the estimation may be transmitted to the main board unit through the external interface ( 69 ) when needed.

In an advantageous embodiment come in the construction of the invention Device the following components are used:

Image capture device

Standard digital camera with a lens, sunscreen, matrix image sensors in the Image plane. Exposure control for image sensors is necessary. Interline transfer CCD Matrix image sensors are preferred (SONY ICX282AQF).

Lens type Biogon® T * 4.5 / 38 CF (Zeiss).

Full Screen Buffer

A number of memory chips (eg dynamic or static RAM).

Arithmetic unit 1, 2, 3

Standard computer with CPU, program and data memory, external interfaces (eg. PCI or memory mapped type).

Example, space-suited RHPPC single board computer, Honeywell.

2D correlator

A special unit is based on FFT DSP, e.g. DSP-24, DSP Architectures Inc. or on optical correlator technology, eg. B. SOCAP, TU-Dresden, Germany; MROC ™, Litton Data Systems, USA.

reference timer

Standard quartz-stabilized pulse generator

REVV memory

A number of memory chips (eg dynamic or static RAM).

External interface

PCI bus, serial interface (RS232) or similar  

LIST OF REFERENCE NUMBERS

1

satellite

2

orbit

3

planet

4

.

5

.

6

Viewing areas of the satellite (light cone)

7

.

8th

.

9

captured areas of the planet surface

10

.

11

Overlaps of the detected areas

12

Pictures of the planet surface

12

a image data

13

Level of the image processing process

13

a position vectors

14

Level of image motion analysis

14

a Rohlage

14

b navigation angle

15

Keplerian orbit parameters

16

Level of orbit estimation

17

green sheet

18

Position of the satellite in the roll-pitch-yaw coordinate system

19

position data

20

Image capture device

21

surface

22

image sensor

23

.

24

surface properties

25

.

26

Positions of images of the surface objects at time t0

27

.

28

Positions of images of the surface objects at time t1

29

Reference coordinate system of the image acquisition device

30

Start the algorithm for image motion tracking

31

Waiting for timemark

32

Capture a full screen

33

Create a new tuple in the REVV

34

Reading a set of image blocks

35

Waiting for timestamp

36

Moving block movement in Δt

37

Comparison with the threshold Λ

38

Check if all image blocks are on the sensor

39

Capture a full screen at the current time

40

Cut out image blocks from predicted positions as current blocks

41

Measuring the displacement between reference blocks and current blocks

42

Check if the measurement was successful

43

Finish the current level of the current tuple

44

Updating and possibly terminating the current tuple in REVV

45

Review and update all unfinished tuples

46

Create a new tuple in REVV if needed

47

Marking the oldest unfinished tuple as current

48

Focus of contraction

49

Focus of expansion

50

navigation angle

51

Model of measurement errors

52

Orbit model

53

measurement model

54

Correction of the condition

55

Estimates of orbit parameters

56

Onboard image capture device

56

a image data

57

High speed processing unit

57

a position vectors

58

Data processing system

58

a Keplerian orbit parameter and RPY attitude

59

A priori data

59

a planetary parameter, properties of the image acquisition device

60

System Interfaces

61

Image capture device

62

image buffer

63

Arithmetic unit 1

64

2D correlator

65

reference timer

66

Arithmetic unit 2

67

Arithmetic unit 3

68

REW memory

69

External interface

70

Data about the position and position of the satellite

Claims (10)

A method of determining the position and attitude of a satellite using image motion information consisting of
a stage of image acquisition in which an image acquisition device aboard the satellite continuously takes up images of the planetary surface that overlap, at least in subregions, whereby certain sections of the images (hereinafter called image blocks) have the same content,
a step of image motion tracking in which the movement of image blocks between successive overlapping images in the image plane of the image acquisition device is tracked, whereby the position vectors of one and the same image block belonging at different times represent the movement of the image block between the times of the acquisition of the viewed images be obtained for all considered image blocks,
a level of image motion analysis that contains information about the 3D motion of the satellite, namely the navigation angles needed for orbit estimation, which describe the motion of the satellite relative to the planetary surface and which depend on the position of the satellite, and the roulette ground Orientation of the satellite relative to the planet and to the orbit called, be determined from the position vectors of the considered image blocks,
a stage of the orbit estimation in which, using an estimation algorithm based on orbital and orbit models, from the navigation angles, satellite position and velocity data, e.g. In the form of the current Keplerian orbital parameters,
a step of correction of the raw layer, in which correction angles are calculated from the orbit parameters and then the position of the satellite in the roll-pitch-yaw (RPY) coordinate system is determined from the previously estimated raw layer with the aid of these correction angles. 2. The method according to claim 1, characterized in that the calculation of the parameters of satellite motion the image movement as di uses direct input quantities. 3. Apparatus for determining the position and attitude of a satellite using image motion information consisting of
an image capture device for taking surface images at defined time intervals,
a high-speed processing unit for driving an image capture apparatus and performing image motion tracking, consisting of buffering image data, extracting and storing image blocks, performing correlation analysis of image blocks, and calculating position vectors of the image blocks,
a digital data processing device for performing a Bildbewe analysis, consisting of the acquisition of navigation angles and Rohla ge of the satellite, the estimation and filtering of parameters of the movement and the orbit of the satellite such as Keplerian orbit parameters, and the correction of the raw layer to the RPY position and
an interface for interaction with the main board unit of the satellite. 4. Apparatus according to claim 3, characterized in that they are made
an image capture device which projects the light from the observed surface onto at least one photosensitive element,
a frame buffer that caches the frame,
a first arithmetic unit performing all the necessary image motion tracking data processing and updating operations and controlling a 2D correlator and the image acquisition device,
a 2D correlator, which takes over the calculation of the 2D correlation function and determines the displacement of image blocks,
a reference timer that generates the reference time stamp,
a second arithmetic unit that performs the image motion analysis,
a third arithmetic unit, which determines the orbit parameters and RPY position,
a REVV (Series of Single Trace Operations) memory which stores the data from ongoing individual tracking operations and
an external interface connecting the device to the main board unit. 5. Apparatus according to claim 4, characterized in that
the first arithmetic unit is synchronized by the reference timer and carries out the extraction of the image blocks on the basis of the position and position of the satellite,
wherein the third arithmetic unit supplies these data as correspondingly revised results of the orbit estimation or these are obtained as external data from the external interface,
wherein the first arithmetic unit updates the contents of the memory of the examined series of individual tracking operations (REVV) and creates new tuples, if necessary
wherein the frame data is read from the frame buffer,
wherein the reference blocks stored in the REVV and read from the frame buffer are forwarded in pairs to the 2D correlator,
wherein, for each pair, the 2D correlator returns the shift value that the first arithmetic unit uses to update the data in the REVV,
wherein the image acquisition device is controlled by the first arithmetic unit so that the image acquisition takes place at certain times,
wherein the exposure time is determined by the arithmetic unit 1 in order to ensure the sharpness of the images,
wherein the frame is stored in the frame buffer,
where the recording times are entered in the corresponding fields in the tuple of the REVV,
wherein the REVV memory contains the REVV,
wherein the first processing unit deletes the unusable and old tuples in the REVV,
wherein the second processing unit accesses the REVV memory and constantly monitors it for finished tuples with finished measurements of the image movement,
wherein the arithmetic unit 2 determines the navigation angle and the raw sheet, if there are tuples with finished measurements of the image motion, using the data about the position and attitude of the satellite,
whereby the determined values are stored again in the REVV,
wherein the third arithmetic unit estimates the orbital parameters and corrects the rouleaux, which follows for the tuples with finished values of navigation angles,
wherein the results of the estimation can be transmitted, as needed, through the external interface to the main board unit of the satellite. 6. Device according to one of claims 4 or 5, characterized in that the image capture device a camera with at least one matrix sensor in the Image plane is. 7. Device according to one of claims 4 or 5, characterized in that the image capture device includes a camera having a plurality of linear sensors in the picture plane is. 8. Device according to one of claims 4 to 7, characterized in that The image acquisition device specially designed lenses to compensate for Bildver contains distortions due to the curvature of the planetary surface.   9. Device according to one of claims 4 to 8, characterized in that the high-speed processing unit for performing the correlations analysis contains an optical correlator. 10. Device according to one of claims 4 to 9, characterized in that Functions of the data processing system through existing resources of Main board unit of the aircraft are running.