DE2842684C2 - Device for recognizing and tracking a target - Google Patents

Description

The invention relates to a device according to the features of the preamble of the main claim.

Target tracking and image processing systems consist of, among other things, a stabilized observation device for recording opto-electronic or image sensors sensitive to thermal radiation, an actual target tracking or image processing computer and a device for image display (monitor).

There are known facilities in which optical and / or electronic aids such. B. a target mark faded into a television picture or paraxially into the beam path via a target mark projector a direct view observation device is projected.

Target tracking systems, so-called tracking devices, are known which work according to a contrast tracking principle. With this principle, the The scene detected by the observation device is divided into light or dark areas in such a way that, for example all bright areas are regarded as the target, all dark areas as the background (or vice versa). as The center of gravity of a bright area recognized as a target can be calculated in a tracking reference point other embodiment, the position of a light / dark edge can be determined (DE-OS 25 47 798). At a Movement of the target, a storage signal is supplied to a servo system, which causes an adjustment of the observation device in accordance with the storage signal. Known designs of such target tracking systems are search heads with quadrant detectors or certain television tracking facilities. What all these devices have in common is the utilization of the

Information »contrast of the target against the background«, what z. B. leads to a flying object in this way can be easily followed as it is generally dark against a light, high contrast background On the other hand, such a device fails if a target with medium gray values is in front of a strong one Outlined background should be tracked Target tracking devices known which evaluate part of the structural information of the target being pursued. A known device works according to an optical method in which a spatial Fourier component of the target on their phase compared to the postponed target at a later point in time is being investigated. The servo system adjusts the observation device to ensure that the phase position is constant over time. There However, no real storage of the target structure takes place, for example the retrieval of one once lost goal impossible.

DE-AS 2106 035 is a device known to be derived from the structure of the target binary schemes are stored in a digital memory for the purpose of correlation with the digitized sensor image will. This method has the advantage over the optical method of storing Target structure information. Finally, according to DE patent application 25 04 536.7 Correlation tracker known that shows the complete structure of the target (one image), in a digital memory for later correlation with the digitized image of the sensor to have.

Now it is very important for target trackers to keep the target memory strong twisting, approaching or constantly changing goals due to internal structural changes to update.

DE-AS 21 06 035 shows that three different quantification levels are used with the aid of three different quantification levels binary targeting schemes can be obtained. From these it can be deduced whether by displacement the quantification level a more representative scheme of the target must be created, which is then automatic an adaptation of the grid scheme and the current image results.

If, according to DE-PA P 25 04 536.7, the complete image of the target is digitally saved, there is the possibility 100% refreshing of the target memory to adapt to changing target prospects.

Furthermore asked the practical experience with target tracking facilities shown that very often the target can be completely or partially obscured. the end For this reason, facilities must be available to bridge these periods of coverage. In large fire control systems are path data extrapolation circuits, so-called prediction computers, known.

It is also known for extensive fire control systems to give important system information on other screens or the usual operating mode displays (Indicator lights, data displays, etc.).

This is where the invention comes in, which is based on the object of adding a target tracking device create as many pursuit situations as possible for ground targets against a strongly structured background and ensures optimal interaction between operator and device.

This problem is solved by the features specified in the characterizing part of the main claim.

The device according to the invention has components that

a) include intelligent target memory learning processes,

b) be able to safely bridge times of target obscuration and

c) as much system-specific and always up-to-date information as possible from an anthropotechnical point of view to offer.

Further developments of the inventive concept according to the main claim are given in the subclaims.

In practice, the target memory can be updated in such a way that the operator at beginning restless wandering of the track window over the target without interrupting the track process, can intervene by hand and give a refreshing impulse. Since the correlation tracker itself by controlling the height of the maximum of the correlation function a statement about the quality of the Can make correspondence between the current target structure and the saved target, it is after the invention possible to define a scene and situation-dependent match limit and at to automatically give a refreshing impulse when it falls below the limit.

Another type of automatic learning method is continuous learning, in which from a new one TV picture always only a certain (variable) percentage of the brightness values of the target in the Target memory is added and an equal percentage is removed from it. After all, it is too conceivable, a certain percentage, e.g. 100%, of the latest TV picture in the Record target memory. In the course of a longer track process, e.g. B. with frequent target concealment or In the case of capture errors, it can happen that the previously targeted target is no longer in the center of the memory window stands, so the tracker wants to track something different than the operator. In such cases so far in tracking systems with contrast trackers with the help of an "offset" via the directional handle, an analogue one Correction variable fed to the stabilization control loop, so that although the tracker, for example, due to his way of working safely follows the top of the mast of a ship, but the line of sight on a more vulnerable one Part of the ship shows. This operating mode, sometimes referred to as »fine alignment«, has its own drift errors and is z. B. for land targets, where mostly the target center and the desired line of sight are the same, not very useful. Instead of it is conceivable to use a structure tracker through window and thus during the track process To force memory area shift within the search area to save other target parts and to be given back to the middle of the object or the middle of the residual structure that has remained due to partial concealment. These The operating mode is called »Messages«.

According to the above proposal, correlation trackers are known which, at the push of a button by the operator save a target view you have just seen and e. B. in the following television picture a correlation between Calculate this saved target structure and the currently visible one and from this the storage of the target determine. If the contrast of the sensor image is very low, a view of the target is saved is just as noisy as the original image or the higher-contrast information from the target background

This leads to a sharp drop in the height of the correlation maximum and thus to an unstable one Tracker behavior. However, since the objectives to be pursued are generally known, for example an im Approaching fighter a dark point against a gray background, it is possible to save the target (PROM) to fill such a noisy expectation and this with the current one Correlate target view.

Since the size of the expectation must be adapted to the actual goal, it is now It is necessary to save various target values in PROMs. In the simplest case, the operator could Enter a suitable target format by hand, as well as the target type (tank, aircraft, ship, etc.) would have to choose. However, it is simpler and more time-saving to automatically select the expected targets to be provided. There are several solutions to this:

Measurement results from distance measuring devices can be used can be used to directly select the PROM (LEM, Radar, etc.).

During the search and capture process, the tracker can select the PROM that corresponds to the background. Ultimately, the largest practical interest rate range shows with continuous Learning to control the level of the maximum of the correlation function, which occurs when the value falls below a predetermined threshold or threshold calculated from system parameters that triggers the prediction and the learning turns off.

The automatic return from the prediction to the learning mode is done again by the control this threshold, which has now been calculated from a different point of view, if the maximum of the correlation function exceeds this threshold.

In the F i g. 1 and 2 are the previously known possibilities of human-machine cooperation shown schematically.

According to FIG. 1, the signal from the image sensor is sent directly to the monitor 2 and from Operator observes. F i g. 2 already shows image processing insofar as from the signal line between Image sensor 1 and monitor 2 transmit signals to a target tracking device 3, tracker or similar, takes place, from which a servo device is acted upon.

According to the invention, a further step is

the closest match with the current 25 like F i g. 3 shows the image processing between sensor t

Target. Here, the height of the maximum of the correlation function is used as a selection criterion of the PROM.

During the track process, e.g. B. when approaching the target, the tracker selects when If the PROM falls below a certain match threshold, the next largest one Target structure. Is the approach so advanced that this particular target view is not more exists, the tracker switches to normal operation, i.e. saving the current one Target view to.

When removing the target, the
Normal operation worked and only when falling below a correspondence threshold or a 40 also signals from a window generator 10, contrast threshold to a character generator 11 stored in the PROM and a digital one

and turn on the monitor. In this case, all system components correspond to one another and the operator via the intermediate images and symbols from the image processing that are superimposed on the image. Finally, the track device 3 from FIG. 4 is shown in FIG. 3 shown in their essential components The central operating mode control is influenced by the control panel on monitor 2 and servo 4 controls the track process. The analog sensor signal is essentially converted into a digital signal in the ADC 6 transformed and fed to the actual image processing 7. The output signal is in a DAC 8 in an analog signal is converted and fed to a mixer 9. Go into this mixer

Point target passed over.

The case is far more difficult with tanks. Here can a constant structure cannot be assumed because the turret and weapon positions are variable. That The expected image could consist of two dark rectangles of different sizes lying one on top of the other consist of light background. The elimination of the well-structured would remain as a mathematical advantage Background from the saved target view

The target occlusion problem can be found under Use of a predictive calculator can only be managed if the tracker is capable of doing this by itself is to distinguish whether there is a real target occlusion or just a change of view Twist acts In the first case, the tracker temporarily transfers responsibility to the prediction computer, in the second case the learning of the new target view is stimulated

Image processing methods are thought of here, in which z. B. is examined whether in parts of the target image the maximum of the correlation function drops sharply, while in other parts of the picture the target structure is still well present. In further refinements it is checked whether an occlusion is in progress with the inverse actual speed of the observation device runs over the target, what standing interfering objects, like Test pattern generator 12 a. From the mixer 9, the processed sensor image arrives in the monitor 2 of the Control unit

The image of the target area saved by the tracker or converted by image processing can be at a location outside the center of the image, after the necessary reconversion (digital / analog), can be displayed again. This

So the procedure offers the possibility of a self-control, whether the operator is at the right moment on target tracking has switched, d, h = whether he is the center of the target or more the background follows The "memory window" also shows whether everyone is seeing math Operations that are carried out during the image processing process are error-free, or whether Too much information is lost Finally, you can see in the memory window whether and how often a activated learning process is carried out and whether the tracker is not off the target! down on the Background learns

F i g. 5 shows one possible arrangement of the memory window in the television picture. 13 is the track window to be pushed by the operator over the target and 14 is the memory window with the target image located in the target memory or its structure converted by image processing.
In the case of special embodiments of the invention,

the following problems arise:
There is necessarily a time delay of a few pixels between an analog image written directly on the monitor and its counterpart going through image processing. A pixel that is centrally targeted in the analog image appears shifted in the memory window. A remedy can now be provided by running the analog image through analog delay lines prior to display in order to generate congruent images. However, this is always associated with a loss of image quality.

According to the invention, this problem can be circumvented if the observer only the digital or somehow image processing image is delivered to the monitor. This representation also serves as a good one Verifiability a) of the entire image processing and b) by using z. B. digital test pattern generators the tracker or the image processing device.

For test purposes (documentation) or to warn the operator of unexpected reactions from the Systems, it is often of great use to write further information in the sensor image. if For example, if the target tracking system is manually tracking a target, the window border could be punched black, but automatically switched to white with safe automatic target tracking will. Each flicker can be meant as an operating system-specific warning.

F i g. 6 shows a block diagram of all the possibilities discussed, such as a target memory 16 in a normal target tracking device 15, which is shown to the left of the dashed line, updated can be.

The device 15 has been proposed, for example, in DE-PA P 25 04 356.7. For this purpose, the Status quo register 17 located current image of the target with a target view in the target memory 16 Corrector 18 supplied, the maximum of the correlation function determined in block 19 and then to further image processing and tracking are passed on to block 20.

The memory 16 is connected via lines 21 with a memory 22 for older target views (RAM) as well as with a fixed target view memory (PROM) 23. A control logic 24 for the sensible Interaction of the status quo register 17 and the target memory 16, while a learning logic 25 the Makes decision which target view correlates with the register 17 and in which time The order and the percentage of the target memory 16 to be refreshed (learn). The decision criteria are provided by sensors 26 to 29. In larger fire control systems there are learning logic 25 system information from the laser range finder (LEM) and radar 30 is also available. Furthermore has the operator 31 always has the option of interrupting incorrect learning processes by manual intervention or to initiate meaningful learning processes by hand at suitable moments. Also can under Under certain circumstances, the search area in the TV picture can be enlarged. After all, it is conceivable and often of practical use if the Based on their own experience, the operator triggers a prediction by hand and only then does the normal one again Enables tracking process when the target can be safely recognized by the tracker.

For this purpose 4 sheets of drawings

Claims (14)

Patent claims: 1. Device for the detection and tracking of a target, consisting of an optronic Image sensor, an image evaluation logic, an automatic tracking system for the image sensor and an operating device in which the image evaluation logic according to dzm Correlation principle by comparing the cached images of the current scene with a stored reference view of the target works and after each image a filing signal to the Automatic tracking supplies, characterized in that an electronic decision odtr learning logic (25) by information such as Similarity between saved and current target view, target speed, target concealment etc. provides for the updating or selection of different target memories (16) connected to a common data bus (21) and that devices are provided which allow for enable the operator to have important information in the monitoring monitor (2) of the control unit. 2. Device according to claim 1, characterized in that the digitally stored target view is automatically overwritten with the current target view if the correspondence between the two target views is a given or falls below the threshold value calculated from the system parameters 3. Device according to claim 1, characterized in that instead of learning after falling below a certain threshold value, new target views are stored at fixed time intervals and / or continuously to certain percentages in the saved view of the target be included. 4. Device according to claim 3, characterized in that in the case of overlap or a prediction computer (27) contradicting tracker results, the learning process is suppressed or decreased. gig is made. 5. Device according to one of claims 1 to 3, characterized in that several on one target memories (RAM) (22) located in the common memory bus system with different old target views are filled to which the correlation calculator has access and their refreshment different learning criteria takes place 6. Device according to claim 1, characterized in that automatically a selection from before permanently entered target views (PROM) (23) and the correlation calculation is carried out with them. 7. Device according to claim 1, characterized in that the automatic updating of the stored target view is always suppressed if the image processing carried out in the correlation tracker is partial or complete Has detected target obscuration or other processes. 8. Device according to claim 7, characterized in that during a total concealment of the A larger search area to find the target automatically after a certain period of time Target view is selected and, after the target view has been found, switched back to the previous search strategy. 9. Device according to claim 1, characterized in that a prediction computer (27) is provided hen is the wrong results of the tracker due to previous speed data of the target prevented and, if necessary, falls back on older target views before a fault 10. Device according to claim 1, characterized in that depending on the prediction computer (27) determined accelerations of the target relative to the line of sight and the target area with variable search area during the first capture process is scanned. 11. The device according to claim 1, characterized in that the stored, optionally by Image processing converted target view is displayed in a small part of the monitor image is 12. Device according to claim 1, characterized in that the window edge of the punched Memory window and / or the track window for displaying certain operating modes in his Color or brightness is switched. 13. Device according to claim 11, characterized in that the entire from the tracker or Image processing device processed sensor image ζ. Ε for testing purposes only after the conversion has been carried out by image processing on the monitor or the observation device is shown 14. Device according to claim 1, characterized in that for documentation purposes and for Support of the operator current system information through certain symbols, letters, Measured values or words are displayed in the image of the observation device.