EP2907105A2

EP2907105A2 - Image processing method and method for tracking missiles

Info

Publication number: EP2907105A2
Application number: EP13794796.6A
Authority: EP
Inventors: Hans Wolfgang PONGRATZ; Manfred Hiebl
Original assignee: Airbus Defence and Space GmbH
Current assignee: Airbus Defence and Space GmbH
Priority date: 2012-10-12
Filing date: 2013-10-08
Publication date: 2015-08-19
Also published as: KR20150072423A; WO2014056473A2; US20150281572A1; JP2015537290A; WO2014056473A3; US9300866B2; DE102012022045A1; JP6285940B2

Abstract

A method for image processing with the following steps: a) capturing image information for a scene as electromagnetic radiation using an optical device (1, 101); b) processing the image information obtained in step a) by means of an image processing device (2, 102) to improve the signal-to-noise ratio in the image information, wherein the processing is performed in the following sub-steps: b1) dividing a raw image containing the image information into lines and columns to create a raster image; b2) overlaying a raster image element with a central raster filter element (22", 122") of a raster filter (22, 122) comprising an odd number of lines and an odd number of columns; b3) determining the brightness values of each of the raster image elements covered by the raster filter (22, 122), wherein apart from the central raster filter element (22", 122") every other raster filter element (22', 122') has an individual light attenuation characteristic; b4) summing the brightness values determined in step b3) to form a total brightness value and assigning this total brightness value to the raster image element covered by the central raster filter element (22", 122"); b5) repeating steps b2) to b4) for all remaining raster image elements; c) generating a resultant image with the same rasterization as the raw image from the individual total brightness values of the raster image elements obtained in step b). A method for tracking the course followed by launched rockets with burning engines providing considerably enhanced precision over the prior art, said method having the following characteristics: a) selecting an observation point to the side beneath the target with particularly good observation conditions, b) tracking using the temperature image of the core engine jet measured by a narrowband multispectral camera with at least three bands.

Description

Method for image processing and thus feasible method for automatic object recognition as well as observation device and method for highly accurate tracking of rocket launching missiles on large

Distances

TECHNICAL AREA

The present invention relates to a method for image processing according to the preamble of patent claim 1. It further relates to a method for automatic object recognition according to the preamble of patent claim 4 and observation devices according to claim 6 and claim 7.

BACKGROUND OF THE INVENTION

In the remote observation of objects, especially in the course of the military reconnaissance, it depends on moving objects, for example, starting rockets with burning engines, over a long distance of up to 1500 km and rockets after the end of fire with sufficient, possibly artificial,

Initially detect lighting over a distance of up to 1000 km, track it along its trajectory and identify it. Even if optical observation devices with long-focal-length lenses and high-resolution image sensors are used for observation, no high-resolution imaging in the range of a few meters is possible with an object distance of more than 500 km up to 1500 km. since the brightness distribution within the fire jet fluctuates in all directions with high frequency in a range of up to 100 m. However, despite such reduced imaging performance, it must be possible to determine whether the object being acquired is a military or civilian launcher or a decoy. Furthermore, the object detection must be able to take place over a sufficient period of time with reliable results with an accuracy of a few meters, to make the trajectory

| Confirmation copy | of the object (for example a rocket) can be measured so accurately that a stable tracking of the observation device is possible and that

if appropriate, a combat of the flying object can take place on its trajectory.

STATE OF THE ART

From the general state of the art, for such tracking purposes, large X-band radars are known, which are along an expected

Rocket flight route must be stationed close enough so that they always have a flying missile in view over the horizon. This not only requires a great deal of effort, but is often not sufficiently possible for political reasons. In addition, such radar stations can only determine the position of targets at a distance of 1000 km transversely to the viewing direction with an accuracy of several kilometers and measure their Radarrückstrahlquerschnitt, but make no accurate identification. In particular decoys are usually not distinguishable by radar from real warheads, which causes great problems.

From the state of the art satellites are still in geostationary

Orbits are known that can detect launching missiles in the mid-infrared range with telescopes. Since the discovery must be made from above against the warm earth background with a multitude of hard-to-detect false targets, these systems have to contend with a comparably low sensitivity. Due to the observation from above, the sensor also sees only the less bright and strongly fluctuating part of the engine jet. As a result, its measuring accuracy is limited to a few hundred meters, and can not be significantly improved due to the system. From the unpublished DE 10 201 1 010 337 is a camera system for detecting and tracking of moving objects located at a great distance from a high-flying aircraft, which is above the dense

Atmosphere flies, known. This system has the great advantage of launching missiles practically out of the atmosphere from below against the background of cold space

to observe and, except for stars and nearby celestial bodies whose position is known exactly, must suppress no further false targets in order to avoid false alarms.

Another great advantage of the observation from below and from the side of the

Trajectory is off, which is free from there the line of sight on the over 2000 ° K hot core area of the jet engine directly at the nozzle exit. This

Core area of the beam has a hundred times higher luminance than the more distant beam and is fixed at the nozzle exit, so does not cause any fluctuation. Thus, an ideal extremely bright (1 megawatt / m ² ) point light source with a diameter of only a few meters is available for tracing the tracks, which accurately and steadily executes the trajectory of the rocket.

The object of the invention is to provide a sensor that can locate and track this point light source over a distance of up to 1500 km with an accuracy of a few meters.

For this purpose, the following multi-spectral camera system for the near-infrared range is proposed. The multispectral camera can sequentially acquire multispectral images (e.g., at 700nm, 800nm, and 950nm) of a scene by means of a motor driven filter wheel having at least 3 narrowband (e.g., 20nm) transmission filters. From this, by conversion according to the blackbody radiation laws, a temperature image of the scene with a resolution of e.g. 50 ° K can be calculated. According to the invention, the approximately 2300 ° K hot core area of a solid rocket and its characteristic shape can clearly be distinguished from the 2100 ° K hot core area of a liquid rocket and a different shape, and with sufficient optical resolution of the camera from 1 m to 2 m in 1000 km distance also measure the size of the core area and its temperature distribution. With this data, military solid rockets can be distinguished from civil liquid rockets and different rocket types by the size, number and arrangement of the engines.

This camera system has a camera provided with a long-focal-length camera lens, which is arranged on a position-stabilized platform. This camera is equipped with a high-speed shutter and a first and a second image sensor. The captured by the camera optics light radiation can be performed either on the first or the second image sensor, wherein one of the image sensors is associated with a further telephoto optics. The camera optics further comprises a pivotable mirror, with which it is possible to scan an area line by line by pivoting the mirror, wherein the captured image signal is fed to one of the image sensors. If a target object is detected during this scanning process, the light beam is deflected to the other image sensor, which is then used for object identification and, if appropriate, for target tracking.

PRESENTATION OF THE INVENTION

Object of the present invention is to provide a method for image processing, with which it is possible, even over long distances,

For example, of several hundred kilometers, in particular to a distance of 100 km to 500 km, processed image information to be processed so that it is possible by means of this processed image information in the detected scene

to recognize contained object by means of the processed image information. Another object is to use this method for

Image processing to perform an automatic object detection.

Finally, it is also necessary to specify observation devices with which these methods can be realized. The directed to the method for image processing part of the object is achieved with the method steps specified in claim 1.

The inventive method designed in this way has the following

Method steps for: a) detecting image information of a scene as electromagnetic radiation, such as light in the visible spectrum, in the infrared spectrum or in the ultraviolet spectrum, by means of an optical device; b) preparing the image information obtained in step a) by means of a

Image processing device for improving the signal-to-noise ratio of the image information, wherein the processing is carried out in the following sub-steps:

b1) subdividing a raw image having the image information in rows and columns to create a raster image;

b2) superimposing a raster image element with a central one

Raster filter element of a raster filter having an odd number of rows and an odd number of columns; b3) determining the brightness values of each of the raster filters

covered raster image elements, except for the central

Rasterfilterelement each other raster filter element having an individual light attenuation property;

b4) summing the brightness values determined in step b3) into a sum brightness value and associating this sum brightness value with the raster pixel covered by the central raster filter element;

b5) repeating steps b2) to b4) for all remaining ones

Raster image elements;

c) generating a result image with the same screening as the raw image from the individual sum brightness values of the raster image elements obtained in step b).

ADVANTAGES

By this image processing method according to the invention the signal-to-noise ratio of the recorded raw image is improved by the

Brightness gradient of the raw image is smoothed. Furthermore, noise pixels that emerge from the background of the raw image are removed, thus smoothing the image as well. The brightness curve in the resultant image is continuous and differentiable from the raw image and the image contrast is improved, so that an object contained in the raw image emerges more clearly and clearly in the resulting image. A preferred development of this method according to the invention for

Image processing is characterized in that in step a) the image information of the scene is detected in more than one electromagnetic wavelength range in order to obtain raw images of the scene in different spectral ranges; that the steps b) and c) are performed for all raw images of the scene in order to obtain result images of different spectral regions, and that the result images of the different spectral regions are combined by superimposition into a multispectral result image.

Detecting the scene in different spectral regions by sequentially capturing images with narrowband (e.g., 20 nm) filters and the

Assembling the raw images of these different spectral regions which have been prepared in accordance with the invention to form a multi-spectral result image improves the informative value of the resulting image result.

This is especially the case when the individual result images

different spectral areas each monochrome colored in different colors are combined to multispectral result image. One

A multi-spectral image with filters selected to match the temperature of the observed body (e.g., 2300 ° K) on the short-wave flank of the blackbody radiation curve can be used to convert the multispectral color image to a temperature image. This temperature image makes it possible to have a small stable core temperature range within a much larger, possibly locally brighter, strongly fluctuating, background brightness field, such as, e.g. to find and track a missile fire tail.

It is advantageous if the acquisition of the image information of the scene in the different spectral ranges using different

Spectral range filter in a short temporal succession by means of a

High speed camera done. This makes it possible to record quasi-time-synchronous raw images of the scene in different spectral ranges compared to the movement of the observed object, which only insignificantly due to the extremely short temporal succession of the respective images differ with respect to the position of the object in the image, so that these shots easily by overlaying the multi-spectral result image

can be summarized. The directed to the method for automatic object detection part of the object is achieved by the method having the features of claim 4. It will be using the method according to the invention for

Image processing the following steps performed: the detection of the scene in step a) is performed at different angles of rotation about the optical axis of the optical device;

for each rotation angle, a result image is generated according to the steps of the method according to the invention for image processing;

the individual result images are compared with sample images of individual objects stored in an object database; and

that pattern image with the least deviation from one or more of the result images identifies the object contained in the scene and determines the position of the object in the result image.

By means of this automatic object recognition method using the image processing method according to the invention an automatic

Object identification enabled. Furthermore, by means of this automatic object recognition method, the position of the object in the result image can be determined, and thus a directional vector of the movement of the object (eg a rocket) can already be predicted with greater accuracy than in the prior art in a single scene detected and analyzed.

A preferred development of this object recognition method according to the invention is characterized in that the determination of the position of the object in the result image by the determination of the matching

Raster elements of the result image with corresponding raster elements of the pattern image is done. The directed to the observation device part of the object is achieved by the observation device with the features of claim 6 and also by the observation device with the features of claim 7. While the observation device is formed with the features of claim 6, to the inventive method for image processing realize, the observation device according to claim 7 is adapted to perform the inventive method for automatic object recognition using the method according to the invention for image processing.

In both observation devices according to the invention, an embodiment in which the image processing device has a screen-rastering module and a raster filter module is advantageous. In a first variant of the observation device according to the invention, the image screening module has a matrix-like arrangement of light-guiding elements which are arranged between the optical device and a sensor sensitive to the detected radiation. It is at least a part of

Light guide each associated with a brightness-reducing raster filter element of the raster filter module. The optical device is designed such that it images the acquired image information in a plane of entry of the image ram module as a raw image, and it is further configured such that the raw image is displaceable with respect to the entrance plane of the image raster module on the entry level. In addition, a computer unit is provided which receives a brightness signal from the sensor and on which runs a software which implements method step c) of claim 1 and preferably the method steps of one of claims 2 to 5. This advantageous embodiment of the observation device implements the method steps according to the invention in an opto-mechanical manner. Whenever "brightness" is used in this document, it is not limited to the spectrum of visible light, but also includes the intensity of the radiation in a non-visible spectrum such as in the infrared spectrum or in the ultraviolet spectrum, but without being limited thereto.

Alternatively, the method steps according to the invention can also be implemented in software, for which purpose the suitable observation device is characterized in that a visual sensor is arranged downstream of the optical device, that the optical device is designed such that it images the acquired image information in a sensor plane of the image sensor a computer unit is provided, which receives an image signal from the image sensor, and that software runs in the computer unit which implements the method steps b) and c) of claim 1 and preferably the method steps of one of claims 2 to 5, wherein the image rasterization module and the raster filter module are designed as a subroutine of the software. This advantageous embodiment of the observation device implements the method steps according to the invention in an optoelectronic manner.

The composite multispectral images are processed after the preparation of the raw single images from a larger number of superimposed

Frames, which have been recorded in different spectral colors, computationally assembled in the computer unit. The compound

Multispectral image then has a much better signal-to-noise ratio than the raw images of preferably more than 100 with a sufficient number of superimposed individual images by averaging over the many frames.

The composite multispectral images are preferably with a

Multispectral image evaluation and identification method according to FIG. 3 or FIG. 4 evaluated in the image evaluation device 25, 125. In this case, an observation of the target behavior is preferably carried out first and, in particular, the number and the trajectories of the visible objects are determined. Then a file of all flying objects and their trajectories is created, which is a safe

Recognition of all objects allowed in subsequent measurements and one

Extrapolation of their trajectories into the future allowed, in particular the calculation possible impact points of the flying object at the end of the trajectory. The composite multispectral images can also be used to observe and analyze the behavior of the objects (separation of a missile upper stage, ejection of decoys, flight maneuvers of a warhead) over time.

The composite multispectral images will continue to be in one

multispectral target image recognition of FIG. 3 or FIG. 4 a comparison with a carried and stored in the memory device 29, 129

Subjected to reference target image database. The target image recognition can thus recognize imaged targets as target objects of a certain type, and consequently

identify. The target image recognition can work more reliably and sharper when the added multispectral images are subjected to multi-stage rendering prior to processing. For this, the composite multispectral images are first

According to the invention, a normalized form is converted by first forming the orthogonal vectorially added total brightness as the brightness value for each pixel, and then normalizing all the color components with the total brightness. The entire color vector then consists of the brightness and the normalized

Color values. This can be a color coordinate system with any number

Spectral components can be defined and in this system can all

Color operations are performed multispectrum, which are defined in the RGB system only in three colors. For all color components and the brightness of each image pixel, the

FIG. 2 shows an operation in which the image is simultaneously smoothed and differentiated, in which after the averaging over a plurality of images still existing interference pixels are removed and smoothed, and in which brightness transitions and edges are accentuated, so that the

Result image sharper, more contrast and clearer in the colors and

can be evaluated more reliably. This is done by transforming all the color components and the luminance component of the normalized image with the in the example shown 5 × 5 pixel filter matrix of the raster filter 122 in the apparatus for processing images with poor signal-to-noise ratio on a digital basis according to Fig. 2 achieved. Preferably, the image is also subjected to an affine color transformation in which the target characterizing spectral components are stretched and thus better evaluated and atypical spectral components are compressed. This allows the correlation of the result images with the real target objects hidden in the images in the target image recognition a higher selectivity between real targets and false targets, which could be confused with real targets, as without this image processing.

The multispectral image recognition can be used either with the device for

3 or with higher accuracy with the apparatus for multispectral image recognition on a digital basis according to FIG. 4.

In the optical-based device (according to FIG. 3), the telescope 110 generates, via the deflection mirror 112, a real 25 × 25 pixel target image of a distant target in the plane 121 of the front surface of the optical 5 × 5

Optical fiber bundle of the image capture device 120, which occupies the same area as 5 x 5 pixels of the real target image. The scanning mirror 1 12 deflects the target image horizontally and vertically so that each center pixel of each 5 x 5 pixel block sweeps sequentially over the central light guide of the 5 x 5 optical fiber bundle. The twenty-five values for the 5 x 5 pixel blocks of each 25 x 25 pixel image are stored in the computing unit 126 for all spectral regions. This is repeated for twelve rotational positions over 360 ° of the 25 x 25 pixel image. Search areas of 15x15 pixel blocks from the target images are compared for the value of each center pixel of each 5x5 pixel block with the sought after 15x15 pixel reference images, with the differences of each nine

Coefficient values of input image search area and current reference image are formed. The position and the rotational position at which the smallest difference sum occurs and at which it falls below a certain minimum value as Position and orientation of a found target of the current reference picture class registered. Image parts that protrude from the image field are not considered. The position resolution of the image recognition is five pixels horizontally and vertically.

In the digital-based device (FIG. 4), the telescope 110 generates a real 25x25 pixel target image of a remote illuminated target in the image plane of the image capture device 120 comprising, for example, an NIR (Near Infrared) camera via the deflection mirror 112. The camera converts the light signal into a digital multispectral image with high resolution. For multispectral image recognition, characteristic values of a weighting function according to the invention are calculated in the search image (25 × 25 pixels in size) for each search pixel position, as described above. Due to the described design of the evaluation function, the number of rotational positions to be examined can be limited to twelve without loss of selectivity.

With a near-infrared sensor system configured according to one of the aforementioned examples, target detection, trajectory tracking, trajectory measurement and target observation and target identification of launching missiles can be performed even after engine burnout at distances up to 500 km.

Preferred embodiments of the invention with additional

Design details and other advantages are described below with reference to the accompanying drawings and explained.

BRIEF DESCRIPTION OF THE DRAWINGS

It shows: a schematic representation of a first observation device for carrying out the inventive

Imaging procedure; FIG. 1A shows a raster filter matrix of the observation device according to FIG. 1; FIG. a second embodiment of an observation device for carrying out the image processing method according to the invention; FIG. 2A shows a raster filter matrix of the observation device according to FIG. 2;

Fig. 3 shows a first embodiment of an observation device for

Implementation of the automatic according to the invention

Object recognition method; and

Fig. 4 shows a second embodiment of an observation device for

Implementation of the automatic according to the invention

Object recognition method. PRESENTATION OF PREFERRED EMBODIMENTS

1 shows an observation device according to the invention with an optical device 1 and an image processing device 2. The optical device 1 has a telescope unit 10 with a long-focal-length objective, which is shown only schematically in the figure. The telescope unit 10 begins

electromagnetic radiation S of an observed scene in the visible spectrum of the light and outside the visible spectrum of the light (for example, infrared and / or ultraviolet radiation). The captured by the telescope 10 radiation S is moved to a

Deflection mirror 12 passes, which is driven by a drive 14 shown only schematically in Fig. 1 for performing a two-dimensional scan movement. The vertical deflection of the deflection mirror 12 takes place in an angular range, which is defined in Fig. 1 by way of example by the upper boundary line a of a central beam SM and by the lower boundary line a 'and by a first lateral boundary line b and a second lateral boundary line b' of the center beam SM. In this area Z spanned by the boundary lines a, a 'and b, b', the deflection mirror 12 performs line-by-line scanning of the radiation S captured by the telescope 10 and in each case projects a portion of the target image Z 'determined by the area Z onto an entrance plane E of a rasterization module 20 of

Image processing device 2. The image screening module 20 has a matrix-like arrangement of light-guiding elements, not shown in detail in FIG. 1, which open at one end in the input plane E. At the other end of the light-guiding elements, a likewise matrix-type raster filter 22 is provided, which is shown only schematically in FIG. 1 and which is reproduced in cross-section in FIG. 1A as a filter matrix. This raster filter 22 contains a central raster filter element 22 "and a plurality of further individual raster filter elements 22 'surrounding it, which are each assigned to one of the further light-guiding elements surrounding the central light-guiding element. In the example shown, the rasterizing module 20 thus consists of five by five (ie twenty-five) Light guide, each of which a raster filter element 22 ', 22 "is assigned.

In Fig. 1A, a respective further raster filter element 22 'is assigned factors which reflect the light transmittance of the individual raster filter element. Accordingly, the central raster filter element 22 ", which is assigned to the central light-guiding element, has a light transmittance of 1.00, which corresponds to 100% .The light guided through the central light-guiding element is thus not attenuated by the raster filter 22. In contrast, the light is transmitted around the central light-emitting element Light guiding element arranged around further light-guiding elements with a light-damping filter effect applied, wherein the factors indicated in Fig. 1A have the following meaning:

0.03 corresponds to 3% light transmission,

0.06 corresponds to 6% light transmission,

0.09 corresponds to 9% light transmission, 0.12 corresponds to 12% light transmission,

0.15 corresponds to 15% light transmission,

0.21 corresponds to 21% light transmission. The light emitted by the individual light guide elements of the image screening module 20 through the raster filter 22 impinges on a sensor 24, which is formed, for example, by a photo light amplifier tube and which forms the sum signal of the light radiation emerging through the individual raster filter elements. This brightness or radiation intensity sum signal Ss is forwarded by the sensor 24 to a computer unit 26 of the image processing device 2.

By the scanning movement of the deflecting mirror 12 thus the incident

Radiation S pixel by pixel detected by the sensor 24 and the computer unit 26 as a radiation intensity sum signal Ss supplied. In this case, the sum of the radiation intensities of the neighboring pixels of the matrix shown in FIG. 1A is detected by the sensor 24 for each target image pixel, wherein the respective target image pixel forms the central element of the matrix. In this case, a target image pixel corresponds to a grid element of the grid of the light guide elements

or the raster filter. Each target image pixel thus becomes the

in accordance with the brightness attenuation matrix shown in FIG. 1A, the luminances of the pixels surrounding the central pixel element (target image pixel) are summarily assigned according to the matrix, so that a

in accordance with the attenuation factors of the matrix weighted light gain for the target image pixel is done.

In the computer unit 26, after a complete scan of the deflection mirror 12, all stored sum brightness values of the individual target image pixels are combined again into a result image, which is then output via an output interface 27, for example on a display device 28. The formation of the result image from the stored sum brightness values of the individual target image pixels takes place in such a way that the stored sum brightness value of a target image pixel is assigned to the same position in the result image that the target image pixel assumed in the target image. This output result image is not only brighter than the target image Z 'originally projected onto the input plane E by the light amplification described, but due to the different weighting of the respective neighboring pixels according to the matrix from FIG. 1A, this result image is also smoothed and no longer contains any noise pixels emerge from the background. As a result, the brightness curve of the result image is continuous, so that the result image is differentiable. The image contrast is also improved by the described method. Although in the example shown the brightness filter matrix (raster filter 22) is given with 5 × 5 raster filter elements and the target image Z 'spanned by the scan area is assumed to be 25 × 25 pixels, the invention can also be used for all other resolutions of the target image and for all other resolutions of the

Brightness filter matrix or the matrix of light-guiding elements can be realized.

FIG. 2 shows a modified variant of that shown in FIG

Observation device shown. The radiation S received by the scene to be observed is transmitted by the telescope unit 110 of the optical system

Device 101 is captured and guided by means of a - in contrast to FIG. 1 nichtbewegbaren- deflecting mirror 1 12 on an image sensor 121 (for example, a CCD sensor) of an image capture device 120, which is part of the image processing device 102, wherein the target image Z 'of observed scene is completely mapped on the image sensor 121.

The image capture device 120 converts the optical target image Z 'projected onto the image sensor 121 into a digital image, which is forwarded in the form of an image file to a computer unit 126. In the computer unit 126 runs from a software that analogous to that in connection with

Fig. 1 described optical image processing method prepared the obtained digital image. In this case, the recorded target image Z 'is converted into a raster image and processed in raster element or pixel by pixel, wherein the brightness value of each target pixel and, according to the filter matrix shown in Fig. 2A forming a raster filter 122 with the raster filter elements 122 ', 122 "as software subroutine, with the filtered pixels of the pixel of the target image surrounding the current target image pixel to a sum brightness value This formation of a sum brightness value is performed for each target image pixel, and from the individually obtained sum brightness values of these electronically-locked target pixel or target image elements, a result image is generated in an analogous manner as described in connection with FIG is sent via the output interface 127 to a

Display 128 issued.

Although the examples described herein are for the visible spectrum of light, they are not limited thereto. In addition to the described

Image processing of scenes recorded in the visible light spectrum, the scenes are also recorded in other spectra, for example in the infrared spectrum and in the ultraviolet spectrum and processed in the same manner as has been described with reference to FIGS. 1 and 2. The thus obtained

Result images of different spectral ranges are superimposed in the computer unit 26, 126 to form a multi-spectral image, which is output via the output interface 27, 127.

Fig. 3 shows an observation device as already described with reference to Fig. 1. An essential difference to the embodiment of FIG. 1 is that the Bildrasterungsmodul 20 having the light-guiding elements about a parallel to the longitudinal direction of the Lichtleitelemente axis of rotation X in predetermined angular increments (for example, 12 angular steps over 360 °) is rotatable as symbolically represented by the arrow 23. The telescope 10 generates via the deflection mirror 12, for example, a 25 x 25 pixel target image Z ', which is shown in Fig. 3 as a matrix with 5 x 5 fields, each field has the size of 5 x 5 pixels or raster elements and having the same size and number of raster elements as the raster filter 22.

The deflection mirror 12 deflects the radiation captured by the telescope 10 horizontally and vertically such that each center pixel of each 5 × 5 pixel block

(= Target pixel Zj) sequentially via the central light guide of the

Frame screening module 20 is performed. The brightness values for the twenty-five target picture elements z, which result in accordance with the image processing described in connection with FIG. 1, are stored in the computer unit 26. The image processing is preferably carried out not only in the visible light range, but also analogously for other spectral ranges. The result values for the other spectral ranges are also stored in the computer unit 26.

Thereafter, the rasterization module 20 is rotated further by one angular step and the described rendering steps are repeated. If this procedure for all angular steps, so for a complete revolution of the

Scanning module 20 about the axis X, carried out, a number of angle increments corresponding number of result images is obtained in different rotational positions.

The nine inner 5 x 5 pixel blocks z _{m of} the target image Z 'are now called

Defined search image area and reduced to these elements z _m target image is compared with stored in a memory device 29 reference images of the same size and the same resolution. The position and the rotational position at which the smallest difference between the search image and the stored reference image occurs is registered as the position and rotational position of a target of a reference image class contained in the monitored scene.

In this way, a method for automatic object recognition is under

Application of the described in connection with FIG. 1

Imaging process created. 4 shows a further variant of an inventive

Observation device for automatic object recognition. The

Observation device corresponds in its construction of the observation device described in connection with FIG. 2, so that the reference numerals in Fig. 4 are the same reference numerals as in Fig. 2 and thus the same components

describe. In contrast to the embodiment of Fig. 2, the

Embodiment of FIG. 4 on a reference image memory in the memory device 129 which communicates with the computer unit 126 for data exchange. The computer unit 26, 126 and the memory device 29, 129 form an image evaluation device 25 within the image processing device 2.

Also with this device, a method for object recognition is performed similar to the method described in connection with FIG. 3, wherein in the

Embodiment of FIG. 4, however, the rotation of the target image Z 'is performed virtually by means of the software running in the computer 126.

For the, preferably multispectral, image recognition, the following characteristic values are calculated in the target image Z '(25 × 25 pixels) for each search pixel position Pz:

1) The average of the individual normalized spectral components and the total brightness, weighted according to the matrix of the raster filter 122, is calculated for each one of the nine marked 5 x 5 pixel blocks z _m and for all four central pixels 123 of the corner pixel blocks in the central area from new pixel blocks In each case, the average value of the average values is calculated for the respective eight 5 × 5 pixel blocks Z arranged in a ring, which surround a respective corner pixel block.

2) The standard deviation of the individual normalized spectral components and the total brightness is calculated for each of the nine marked 5 x 5 pixel blocks z _m and for all four central pixels 123, and the average value of the standard deviations for each eight in one

Ring arranged 5 x 5 pixel blocks z _R calculated.

3) The values according to 1) and 2) are calculated in twelve rotational positions distributed over 360 °. 4) Each set of values becomes a search pixel in each rotation position

Difference formation compared with the values for the searched reference target image and the value set with the smallest absolute value of the difference sum is registered as a representative of this search pixel.

5) The target image is now decomposed into smaller subareas and the search pixel with the smallest difference sum in each subarea is searched. The value set of the search pixel with the smallest difference sum is interpreted as a recognized target image and at the considered search pixel position with a pixel resolution and the rotational position as the discovered target of the

Reference Target Type registered.

Reference signs in the claims, the description and the drawings are only for the better understanding of the invention and are not intended to limit the scope.

LIST OF REFERENCE NUMBERS

They denote:

1 optical device

2 image processing device

10 telescope unit

12 deflecting mirrors

14 drive

18 display device

20 framing module

22 raster filters

22 'raster filter element

22 "raster filter element

24 sensor

25 image evaluation device

26 computer unit

27 output interface

28 display device

29 storage device

101 optical device

102 image processing device

110 telescope unit

112 deflecting mirror

120 image capture device

121 image sensor

122 Raster filter

122 "grid filter element

125 image evaluation device

126 computer unit

127 Output interface 128 display device

129 Reference picture memory a Limit line

a 'borderline

b borderline

b 'borderline

E input level

Pz search pixel position

s electromagnetic radiation

SM center beam

S _s summation signal

X optical axis

z stretched area

Z 'target image

Zi target picture element

Z _m pixel block

ZR pixel block

Claims

claims

Process for image processing with the following steps:

a) Capture image information of a scene as electromagnetic

Radiation by means of an optical device (1, 101);

b) preparing the image information obtained in step a) by means of an image processing device (2, 102) for improving the signal-to-noise ratio of the image information, the processing being carried out in the following sub-steps:

b2) superimposing a raster image element with a central one

Raster filter element (22 ", 122") of a raster filter (22, 122) having an odd number of lines and an odd number of columns;

b3) determining the brightness values of each of the raster pixels covered by the raster filter (22, 122), each other except the central raster filter element (22 ", 122")

Raster filter element (22 ', 122') an individual

Having light attenuation property;

b4) summing the brightness values determined in step b3) into a sum brightness value and associating this sum brightness value with the raster pixel covered by the central raster filter element (22 ", 122");

b5) repeating steps b2) to b4) for all remaining ones

Raster image elements;

c) generating a result image with the same screening as that

Raw image from the individual sum brightness values of the raster image elements obtained in step b). A method of image processing according to claim 1,

characterized,

that in step a) the image information of the scene in more than one electromagnetic wavelength range is detected, so

To obtain raw images of the scene in different spectral ranges;

that steps b) and c) are performed for all the raw images of the scene in order to obtain result images of different spectral regions, and

that the result images of the different spectral regions are combined by superimposition into a multispectral result image.

A method of image processing according to claim 2,

characterized,

in that the acquisition of the image information of the scene in the different spectral ranges using different spectral range filters in a short temporal succession by means of a

High speed camera done.

Method for automatic object recognition using the

Method for image processing according to one of the preceding claims, characterized

that the capture of the scene in step a) in different

Rotation angles about the optical axis (X) of the optical device is performed;

that for each rotation angle, a result image is generated in accordance with the steps of the image processing method;

in that the individual result images are compared with pattern images of individual objects stored in an object database and that the pattern image with the least deviation from one or more of the result images identifies the object contained in the scene and determines the position of the object in the resulting image. Method for automatic object recognition according to claim 4,

characterized,

in that the determination of the position of the object in the result image is effected by the determination of the matching raster elements of the result image with corresponding raster elements of the pattern image.

Observation device with an optical device (1, 101) for capturing image information of a scene and with a

Image processing device (2, 102) for improving the signal-to-noise ratio of the acquired image information,

characterized,

in that the observation device is designed to carry out the method steps of the image preparation method according to one of Claims 1 to 3.

Image processing device (2, 102) for improving the signal-to-noise ratio of the acquired image information and an image evaluation device (25, 125)

characterized,

in that the observation device is designed to carry out the method steps of the object recognition method according to claim 4 or 5.

Observation device according to claim 6 or 7,

characterized,

in that the image processing device (2, 102) has a screen module (20) and a screen filter module (22, 122). Observation device according to claim 8,

characterized,

- That the Bildrasterungsmodul (20) comprises a matrix-like arrangement of light-guiding elements, which are arranged between the optical device (1) and a sensor for the detected radiation (24);

- That at least a part of the light guide each one

brightness-reducing raster filter element (22 ') of the raster filter module (22) is assigned;

- That the optical device (1) is designed so that it images the detected image information in an entry plane (E) of the image screening module (2) as a raw image;

- That the optical device is designed so that the raw image

with respect to the entry plane (E) of the image screening module (20) on the entry level (E) is displaceable;

- That a computer unit (26) is provided which receives a brightness signal from the sensor (24), and

- That in the computer unit (26) runs a software that the

Process step c) of claim 1 and preferably the

Method steps of one of claims 2 to 5 implemented.

Observation device according to claim 8,

characterized,

- That the optical device (101) an image sensor (121) is arranged downstream;

- That the optical device (101) is designed so that they

captured image information in a sensor plane of the image sensor (121) maps;

- That a computer unit (126) is provided which receives an image signal from the image sensor (121), and

- That in the computer unit (126) runs a software that the

Method steps b) and c) of claim 1 and preferably the method steps of one of claims 2 to 5 implemented, wherein the Frame screening module and the raster filter module (122) are designed as a subroutine of the software.

Method for tracking the track of launching rockets with burning

Engines with greatly increased accuracy over the prior art,

characterized,

- that the observation of a high above the dense atmosphere flying aircraft up with the cold space as

Background takes place and thereby significantly increases the sensitivity of the system and the susceptibility is reduced.

- that the observation is from the side and from below, so that the line of sight on the hot core area of the engine jet is hidden neither by the missile's fuselage nor by the smoke and the exhaust jet behind the missile.

that narrow-band multispectral images of the fire beam of the rocket are recorded in at least 3 bands whose spectral ranges are tuned to the temperature of the observed target, that the temperature distribution in the target image is very accurate, e.g. calculated to 50 ° K.

- That the track tracking of the rocket, the high temperature region of the core beam is used, which is firmly bonded to the engine nozzle and is not subject to fluctuations, the

Could affect measurement accuracy.

- That the optical resolution in the size of the nozzle diameter of the rocket is selected in order to determine the size, the number, and the arrangement of the rocket engines on the rocket, and thus to be able to classify the rocket according to their type.