Method for capturing a scene comprising a phase object

Abstract

The invention relates to a method for capturing a scene (20) comprising a phase object (40) located in front of a textured background (21), in which method a plurality of cameras directed at the background (21) through the phase object (40) at different viewing directions are used to simultaneously record a corresponding number of camera images (12), whereupon the camera images (12) are evaluated in order to draw conclusions about the form of the phase object (40). The invention is characterized by the following steps: a) providing, for each camera (10), an inverse of its imaging function, the inverse constituting a unique assignment of each pixel of the camera image to a line of sight that is defined by all object points assigned to said pixel by the imaging function, b) identifying, in each camera image (12), the pixel (121) onto which a background object point (221) imaged in each camera image (12), namely an object point belonging to the background (21), is imaged, and calculating the particular assigned line of sight (31), c) determining a measure of a distance of the calculated lines of sight (31) from one another, d) assigning a deviation value representing said measure to the corresponding pixel (121) in at least one camera image (12), and e) repeating steps b to d for a plurality of background object points (221).

Description

[0001] Method for capturing a scene comprising a phase object

[0002] Description

[0003] Field of the invention

[0004] The invention relates to a method for capturing a scene comprising a phase object arranged in front of a textured background, in the context of which a corresponding plurality of camera images are simultaneously recorded by means of a plurality of cameras arranged spatially offset from one another and directed through the phase object to the background in different viewing directions, after which conclusions are drawn about the shape of the phase object at the time of recording by evaluating the camera images.

[0005] State of the art

[0006] Such a method is known from DE 100 10 045 02.

[0007] So-called BOS methods are generally known to those skilled in the art. BOS stands for "Background Oriented Schlieren" and describes a group of methods suitable for visualizing or, more generally, for detecting phase objects. In the context of the present description, a phase object is understood to be a spatially extended area which, in the spectral range used for observation, differs optically from its immediate surroundings essentially only in its refractive index. Typical examples of such phase objects are, for example, air vortices which differ from the ambient air in terms of their pressure distribution, and foreign gas or temperature bubbles in a primary gas which differ from the surrounding primary gas in terms of their refractive index due to their different chemical composition or density.Those skilled in the art will understand that the term "different refractive index" does not imply a uniform refractive index for the entire phase object, but rather also encompasses distributions of refractive indices that are distinguishable from their surroundings. Such phase objects cannot be detected using purely intensity-based sensor technology, such as that provided by conventional cameras.

[0008] The BOS technique is based on the fundamental insight that light traveling from a background object point to a camera takes a different path depending on whether or not it passes a phase object positioned between the background object point and the camera, and over which distance it passes this. Accordingly, the background object point is imaged onto a different image point in the camera image of a camera imaging the background. Originally, the BOS technique therefore envisaged imaging a textured background, i.e., one provided with any imageable pattern, in a reference image using a camera at a first recording time directly, i.e., without the phase object positioned between the background and the camera, and then recording a measurement image using the same spatial constellation of background and camera, with the phase object now positioned between the background and the camera, i.e.the background is recorded through the phase object. The reference image and the measured image show the same background, but with slight, usually locally different relative shifts of the image points on which the background object points are imaged. Such shifts can be determined, for example, by cross-correlating the reference and measured images. The extent of the local shifts allows conclusions to be drawn about the deflection of the light from the background object point to the image point due to the phase object. This, in turn, allows conclusions to be drawn about the shape of the phase object and the size of the refractive index deviations. This data can then be used, for example, to visualize the phase object.The disadvantages of this basic type of BOS technology are, on the one hand, the time delay between the recordings and, on the other hand, the frequently occurring practical impossibility of recording a reference image without a phase object and a measurement image with it. The generic DE 100 10 045 C2 mentioned above describes a method in which two camera images are recorded simultaneously using two spatially offset cameras. Both cameras record the same background through the phase object, although the phase object and in particular its edges are viewed by the cameras from a different angle. Of particular interest are those image points on which assigned background object points are imaged through the phase object in one camera image and past the phase object in the other camera image.Therefore, each camera image can be used as a reference image for certain areas and as a measurement image for the corresponding areas of the other camera image. However, before the actual comparison of the camera images, which reveals the influences of the phase object, they must be corrected for their different perspectives. For this purpose, a global, typically quadratic transformation function is fitted, which is intended to make one image convertible into the other image, with the exception of the phase object influences. The comparison of the transformed images should then reveal only the differences attributable to the phase object influences.However, contrary to the claim in the cited publication, this approach can only work for backgrounds with a comparatively weak depth structure: A transformation function of low polynomial order cannot represent a strongly pronounced depth structure; however, a transformation function of high polynomial order also takes into account the influences of the phase object itself and thus makes it indistinguishable from differences caused solely by parallax. The entirely conceivable choice of a transformation function based on a model that accurately reproduces the background, however, would require precise knowledge of the background and would therefore be unsuitable for unknown and / or changing, e.g., moving, backgrounds.

[0009] Task

[0010] The object of the present invention is to provide a BOS method that enables the detection of phase objects even against moving backgrounds with a pronounced depth structure. Statement of the invention

[0011] This object is achieved in conjunction with the features of the preamble of claim 1 by the following steps: a) Providing, for each camera, an inverse of its imaging function, wherein the imaging function represents a unique assignment of each object point detectable by the respective camera to a pixel of the corresponding camera image, and the inverse represents a unique assignment of each pixel of the camera image to a line of sight defined by all object points assigned to this pixel by the imaging function, b) Identifying, in each camera image, the pixel onto which a background object point imaged in each camera image, namely an object point belonging to the background, is imaged, and calculating, by means of the inverse of the respective imaging function, the respectively assigned line of sight, c) Determining a measure for a distance of the calculated lines of sight from one another,d) Assigning a deviation value representing this measure to the corresponding pixel in at least one camera image and e) Repeating steps b to d for a plurality of background object points.

[0012] Preferred embodiments of the invention are the subject of the dependent claims.

[0013] The present invention essentially relates to the evaluation of images such as can also arise with the previously known method. Specifically, these are images of a textured background recorded simultaneously at different viewing angles through a phase object. Preferably, three or four simultaneously recorded images are used. However, the method according to the invention can also be carried out with just two simultaneously recorded images. In contrast to the prior art, the evaluation of such camera images according to the invention is not based on a direct comparison of the camera images, but uses the principle of triangulation. The imaging behavior of a camera can be described by its so-called imaging function. This represents an assignment for each object point to an image point in the camera image.The explicit definition of such imaging functions is necessary and common, for example, in the field of stereoscopic shape measurement and is therefore familiar to those skilled in the art. Those skilled in the art are also aware of highly accurate calibration methods that allow the imaging function to be set very precisely. The inverse of any imaging function can be calculated, i.e. a function that calculates inversely from the pixels of the camera image to the object space. However, this inverse is not unique due to the difference in dimensions between the three-dimensional object space and the two-dimensional camera image. In particular, the inverse of the imaging function can only be used to calculate a so-called line of sight for each pixel of the camera image. This line of sight corresponds to a straight line on which all those object points lie that would be mapped to the associated pixel according to the imaging function.The distance from the camera to an actually imaged object point on the line of sight cannot be determined from the inverse assigned to a single camera. However, if the same object point is imaged by multiple cameras from different viewing angles, and assuming correct calibration of the imaging functions or cameras, the lines of sight must intersect at a common point in object space, namely the imaged object point. This is the basic assumption of triangulation. To determine the position of an object point in object space imaged by multiple cameras, the inverse assigned to the respective camera is applied to the image points to which the object point in question is imaged, and the intersection point of the resulting lines of sight is calculated.

[0014] However, if the imaging process is disrupted due to the presence of a phase object between the object point and the cameras, the basic assumption explained above, in particular the assumption of a straight light path from the object point to the image point, is no longer correct. The imaging function set up for each camera without taking the phase object into account, which assumes straight-line light propagation, no longer correctly describes the actual imaging via an angled light path through the phase object. Accordingly, the inverses of the imaging functions no longer correctly represent the real light path. The calculated line of sight and the real light path no longer match. This is particularly evident in the fact that the lines of sight calculated for each image point imaging the object point using the inverses are skew to one another, i.e. they do not intersect.The extent of this skew is a measure of the deviation of the calculated lines of sight from the actual light path and therefore a measure of the influence of the phase object on the assignment of object point and image points.

[0015] This insight is utilized within the scope of the present invention by first determining object points that are depicted in all (or at least several) camera images. For each of the corresponding pixels, the associated (ideal) line of sight is then calculated using the inverse, and the extent of skew of these lines of sight is determined. A corresponding value, referred to here as the deviation value, can then be assigned to the corresponding pixel in one or more or all camera images. If this method is performed with multiple object points in the background, such deviation values ​​can be assigned to many pixels in the camera images.

[0016] The specific mathematical determination of the deviation value can be carried out in different ways. Purely as an example, the possibility of determining a sum of the minimum distances between the lines of sight and using this as the deviation value. Alternatively, it is conceivable to define a virtual object point characterized by a minimal sum of its perpendicular distances to the lines of sight. It is also conceivable to resort to the technique of epipolar geometry, which is familiar to those skilled in the art. For example, it can be provided that the degree of skew is determined from the distance of a real image point in one camera image to the epipolar line of the corresponding real image point, i.e. the real image point assigned to the same object point, in (at least) one other camera image.The epipolar line assigned to a real pixel in a first camera image is the line that corresponds to a mapping of the line of sight assigned to said real pixel in a second camera image (or would correspond if the mathematical construct "line of sight" could be mapped in reality). In this method variant, the respective real pixel assigned to a common object point is determined as the first or second real pixel in at least a first and a second camera image. Then, in the second camera image, the line that would result here as a mapping of the line of sight of the first camera emanating from the first real pixel - calculated using the inverse of its mapping function - is determined as the epipolar line. Finally, the distance of the second real pixel from said epipolar line, e.g., its perpendicular distance, is measured and subsequently used as a measure of the skew.Of course, when calculating several epipolar lines from different cameras in one camera image, a corresponding number of distances from the real image point under consideration can be combined in a suitable manner, e.g. averaged, in order to obtain the desired measure for the wind skew.

[0017] Of course, due to the specific constellation, it can happen in individual cases that the calculated lines of sight intersect at a point for individual pixels, even though they do not correctly represent the actual light paths. In this case, instead of inferring a phase object, the position of the object point would be incorrectly estimated. Such an exceptional case can occur in the worst case when using only two cameras. However, even when using three cameras that are deliberately not arranged in a line, such a situation is practically impossible. If the calculations in question are also performed for several, especially many, object points, such cases can easily be identified as "outliers" and excluded from further analysis.

[0018] A significant advantage of the method according to the invention is that no assumptions are required regarding the nature, and in particular, the depth structure, of the background. Care should only be taken to ensure that the depth structure is not so pronounced relative to the distance from the cameras that the background objects cast shadows on each other, i.e., obscure the cameras' view of them. However, this is a general requirement for a sensible measurement setup, obvious to the person skilled in the art, and not specific to the invention.

[0019] The method according to the invention results in a map that represents the local influence of the phase object on the image of the background through the phase object. Such a map is preferably extracted as a separate data set. In other words, a further step of the method according to the invention is preferably provided, namely a step f) creating a map that is structurally analogous to the camera image, in which each map point corresponding to an image point to which a deviation value was assigned in one of the performed steps d is assigned a pixel value representing this deviation value.

[0020] The term "constructively analogous" is intended to express that the data matrix, referred to here as a map, has, at least in the relevant area, the same constellation of matrix points as the data matrices referred to here as camera images. Simply put, the preferred map of deviation values ​​represents a matrix in which the matrix points to which a deviation value is assigned are located in the same position in their matrix as – with respect to the respective camera image – the image points to which said deviation value was originally assigned in the camera image.

[0021] For further use of the representation of the phase object embodied by the map, the expert has a wide variety of options at his disposal. In particular, he can use the map to visualize the phase object.

[0022] This can be achieved particularly easily by displaying the map on a screen, with each map point being assigned a color or gray tone that corresponds to its respective pixel value according to a predefined color or gray value coding. This corresponds to a visualization of the phase object only.

[0023] Alternatively, it is also possible to overlay the map of deviation values ​​with the corresponding camera image. Such a visualization shows the phase object against the background. For example, the background can be displayed in grayscale and the phase object or deviation values ​​in color coding. Such an overlaid visualization is particularly relevant for measurements in which the spatial constellation of the phase object relative to its surroundings plays a role.

[0024] So far, only those map points have been discussed that correspond to pixels in the camera image to which a deviation value has been explicitly assigned by calculating the lines of sight. When visualizing the phase object, this leads to it being represented only point by point or with gaps, which can be unsatisfactory depending on the application. However, the associated additional computational effort is often offset by only a slightly greater gain in knowledge about the shape of the phase object. This is particularly true in cases where the refractive index distribution within the phase object is largely uniform. In particular, but not only in these cases, it is conceivable to supplement the areas on the map between those map points with an explicitly calculated deviation value by interpolation or extrapolation.In this regard, it is preferably provided that map points which correspond to image points to which no deviation value was assigned in any of the steps d carried out are each assigned a pixel value which represents an interpolated or extrapolated deviation value, wherein those map points to which a pixel value representing a deviation value was assigned in step f are used as support points for the respective interpolation or extrapolation.

[0025] As described, each calculated map corresponds – according to the so-called constructive analogy explained – to an underlying camera image. In other words, each map represents the phase object from the perspective of the respective associated camera. The specific camera image chosen by the person skilled in the art to construct the map is fundamentally irrelevant. Typically, they will select the camera image that most clearly depicts the relationship between the phase object and the background in the specific case. However, a further development of the invention provides for multiple maps to be constructed, namely based on several, preferably all, camera images. Such a data set thus represents the phase object from different perspectives. From this, a three-dimensional shape of the phase object can be calculated, knowing the relative spatial arrangement of the cameras.For this purpose, familiar stereo photography algorithms can be used and applied to the maps. This subsequently also allows a perspective representation of the phase object on a display.

[0026] However, it should be expressly noted that for effective use of the method according to the invention—at least when complete, three-dimensional capture of the phase object shape is not required—it is not necessary to position the cameras at a large spatial distance from one another. In a preferred embodiment, it is conceivable, for example, for the cameras to be arranged directly adjacent to one another, fixed to one another with a common frame, or even rigidly fixed to one another in a common housing. Such rigid arrangements avoid errors that may arise, for example, from vibrations of the structure in a harsh, for example, industrial environment.

[0027] A key feature of the method according to the invention is that the camera images, on the basis of which the calculation of the phase object is carried out as explained in detail above, are recorded simultaneously. Errors caused by movements of the phase object or the background between two temporally offset recordings are avoided. The resulting map(s) thus represents(s) the phase object at a specific time.

[0028] Recording time. However, a further development of the invention provides for repeating the basic method at one or more points in time, thus also capturing the temporal development of the phase object with regard to its shape and / or movement and, if necessary, visualizing it, for example, in the form of a video.

[0029] The nature of the background plays no role in the basic principle of the invention. It is conceivable to use a naturally occurring background or an artificially patterned background. The latter can be achieved using printing or projection techniques.

[0030] Further details and advantages of the invention will become apparent from the following specific description and drawings.

[0031] Brief description of the drawings

[0032] They show:

[0033] Figure 1: a highly schematic representation of three calibrated cameras in

[0034] Alignment to a background,

[0035] Figure 2: the structure of Figure 1 with a background mounted

[0036] Phase object, Figure 3: a schematic drawing to illustrate the relationship between line of sight, real and ideal light path,

[0037] Figure 4: an illustration of the central step of the inventive

[0038] Procedure carried out on the structure of Figures 1 and 2 and

[0039] Figure 5: an illustration of a possible visualization step for the

[0040] Process step of Figure 4.

[0041] Description of preferred embodiments

[0042] The same reference symbols in the figures indicate the same or analogous elements.

[0043] Figures 1 to 4 show, in a highly schematic representation, basic steps of the method according to the invention. Figure 5 shows an optional subsequent visualization step.

[0044] Figure 1 shows a possible setup for implementing the method according to the invention, with three cameras 10 viewing a scene 20 from different angles. For illustrative purposes, the spatial positions of the cameras 10 are shown far apart from each other. In practice, a significantly smaller spacing is usually sufficient.

[0045] In the situation illustrated in Figure 1, the scenery 20 consists solely of a background 21, which in the embodiment shown is chosen as a random, printed dot pattern. Other types of texturing of the background 21, including the use of "natural" backgrounds or the projection of a structure onto a scattering surface, are also possible. All that is required is that there is sufficient space between the background 21 and the cameras 10 to position the actual target object of the inventive method, namely a phase object, which will be discussed in more detail below. The imaging optics 11 of each camera 10 is preferably adjusted so that both the background 21 and the space in which the phase object of interest is positioned can be sharply imaged.

[0046] Each camera 10 can capture camera images 12. These are ordered data matrices of pixels, as known to those skilled in the art from CCD, CMOS, or similar cameras. For the sake of clarity, the camera images 12 are shown in front of the cameras 10 in Figures 1 to 4.

[0047] The illustrated setup of the cameras 10 can be calibrated in a manner known to those skilled in the art. This means that an imaging function can be established for each camera 10 that assigns a pixel on the respective camera image 12 to each object point in the scene 20. While the imaging function can be established based on different models, e.g., based on a so-called pinhole model, it is always based on the assumption of rectilinear light propagation along an ideal, straight light path 30. Likewise, an inverse of its imaging function can be determined for each camera 10. However, due to the change in dimension between the three-dimensional scene 20 and the two-dimensional camera images 12, this inverse is not as unambiguous as the imaging function.Rather, the inverse of the imaging function assigns a line of sight 31 to each pixel of the camera image 12, which line can be imagined as a straight line composed of all object points that would be imaged on the pixel in question according to the imaging function. Ideally, i.e. in the absence of a "disturbing" phase object, the light travels along a real light path 32 that corresponds both to the ideal light path 30 assumed by the imaging function and to the line of sight 31 resulting from the inverse. Figure 1 shows this for a background object point 211. In each camera image, this is imaged on the pixel 121 assigned by the imaging function in the respective camera image 12. The light travels along a real path 32 that is identical to the ideal path 30.The respective inverse of the mapping function defines, as explained above, a line of sight 31 for each pixel 121, which corresponds to the real and ideal light paths 30, 32 from the background object point 211 to the respective pixel 121. These lines of sight 31 intersect at a point in the scene, namely—with correct calibration—at the associated object point, i.e., in the depicted situation, at the background object point 211. Those skilled in the art will understand that these relationships apply to every object point and every pixel.

[0048] Figure 2 illustrates a situation in which a phase object 40 is positioned between the background 21 and the cameras 10. For example, this could be an extraneous gas bubble, a temperature bubble, or an area of ​​differing pressure. In the situation shown in Figure 2, for purely illustrative reasons, the phase object 40 is arranged in such a way that it only partially obscures the view of the background 21 for the two lower cameras 10, while the upper camera 10 in Figure 2 has a clear view of the background 21. In practice, however, where the cameras are positioned significantly closer together than shown in Figure 2, this situation will be rather rare. The obstruction of the clear view of the two lower cameras 10 of the background 21 is indicated in Figure 2 by a semi-transparent representation of the phase object 40.This is again purely for illustrative purposes, since it is precisely the nature of a phase object 40 to be indistinguishable from the environment in the selected spectral range with regard to its absorption properties and therefore not detectable by pure intensity measurements.

[0049] However, the path that the light takes from the object point, in particular from the background object point 211 to the image points 121, is influenced by the phase object 40. This is illustrated in the schematic representation of Figure 3. The light travels from the background object point 211 along the real light path 32, shown as a solid line, to the image point 121. Due to the phase object 40 passed on this path, this real light path 32 deviates from the ideal light path 30, shown in dashed lines, which the light would take without the phase object 40 from the background object point 211 to the ideal image point 12T, or the one predicted by the mapping function. For the real image point 121, the inverse of the mapping function calculates a straight line of sight 31, shown in dashed lines, which corresponds neither to the real light path 32 nor to the ideal light path 30.

[0050] For the sake of clarity, Figure 2 only shows the ideal light path 30 (dashed line) to the ideal image point 12T (shown hollow in Figure 2) and the line of sight 31 (solid line) calculated from the inverse of the real image point 121 (shown filled in Figure 2). Therefore, the real image point 121 is not located at the same position in the camera image 12 as the ideal image point 12T actually predicted by the mapping function. Likewise, the respective line of sight calculated by means of the real

[0051] The line of sight 31 determined using the inverse applied to image point 121 (shown as a solid line in Figure 2) does not correspond to the ideal light path 30 (shown as a dashed line in Figure 2) that would result without the presence of the phase object 40 (cf. Figure 1). This is the case for the two lower cameras 10 in the situation shown in Figure 2. For the upper camera 10, whose view of the background 21 is not impaired by the phase object 40, the calculated line of sight 31 corresponds to the line of sight expected according to the calibration or the ideal light path 30, which here coincides with the actual light path.

[0052] In any case, the distortion caused by the phase object 40 results in the lines of sight 31 calculated according to the inverses provided and applied to the real pixels 121 not intersecting but running skew to each other.

[0053] This is schematically indicated in Figure 4 in enlarged section A. The extent of the resulting wind skew is simultaneously a measure of the influence of the phase object 40 on the image of the background object point 211. The method according to the invention therefore provides for quantifying this extent of wind skew in a suitable manner and assigning a corresponding value, namely the deviation value, to the respective pixel 121 in the camera images 12. This assignment is symbolized in Figure 4 by the assignment arrows 50.

[0054] For determining the extent of skew, various approaches are available to the person skilled in the art, some of which have already been mentioned as examples in the general part of this description. Using the mapping equations, the assigned pixels on the camera images 12 can then be determined, and the distances of these pixels of the virtual object point from the real pixels 121 can be calculated and used as a basis for determining the deviation value. For example, the deviation value can be defined as the sum of the stated distances between the real pixels 121 and the calculated pixels of the virtual object point. As a further, particularly simple variant, it is conceivable to merely determine whether the calculated lines of sight 31 are skew to one another, i.e.The objective is to determine whether the free view of at least one camera to the background object point is obstructed by the phase object 40, or not, and to define the deviation value as a purely binary value. However, other approaches for determining the degree of skew of the lines of sight 31 are also available to the person skilled in the art.

[0055] The process described above for a specific background object point 211 is repeated for a plurality of background object points. This results in the assignment of deviation values ​​to a plurality of pixels in the camera images 12.

[0056] Figure 5 shows one possibility for extracting this information and a visualization of the phase object 40 based on it. For each camera image 12, a constructively analog map 14 is generated, i.e., a data matrix of ordered pixels whose arrangement corresponds, at least in the specific area of ​​interest, to the arrangement of the pixels in the camera image. Thus, each (relevant) pixel 121 in the camera image 12 is assigned a map point 141 in the map 14, wherein the map points 141 have the same relative positioning to one another as the assigned pixels 121. The corresponding deviation value is then entered in the map 14 at each map point 141 assigned to a pixel 121 to which a deviation value had previously been assigned. The result is a map that represents the shape and strength (at least in a semi-quantitative sense) of the phase object 40.Such a map can be used to visualize the phase object 40, either by directly displaying the map 14 on a display or by overlaying a map 14 with the associated camera image 12 under suitable color or gray value coding, so that the phase object 40 can be visualized alone or in relation to the background 21.

[0057] Other ways of using the maps 14 are also conceivable. For example, a three-dimensional shape of the phase object 40 can be reconstructed using stereophotographic algorithms. This is particularly possible when the viewing angles from which the cameras 10 view the scenery 20 differ significantly from one another, as shown in Figures 1 to 4.

[0058] Of course, the embodiments discussed in the specific description and shown in the figures represent only illustrative embodiments of the present invention. In light of the disclosure herein, a broad spectrum of possible variations is available to those skilled in the art. In particular, the number of cameras 10 is not subject to any significant limitation. The use of three cameras is considered particularly advantageous. When using four cameras, improved accuracy can be achieved through the resulting redundancy. The use of even more cameras increases the redundancy without, however, leading to a further significant increase in measurement accuracy. With regard to the choice of scenery, no fundamental application limitations of the invention can be identified.A possible application is the visualization of flow events in aeronautics, for example, the visualization of vortices behind the wings of aircraft or helicopter rotor blades, where the naturally textured background of the aircraft can serve as the background. Another possible application is the visualization of flows in so-called flow boxes or clean rooms in the pharmaceutical or chip manufacturing industries. The visualization of general flow and combustion processes can also be an area of ​​application for the present invention.

[0059] List of reference symbols

[0060] 10 Camera

[0061] 11 Imaging optics

[0062] 12 Camera image

[0063] 121 real pixel 121' ideal pixel

[0064] 14 map

[0065] 141 map point

[0066] 20 Scenery

[0067] 21 Background 211 Background object point

[0068] 30 ideal light path

[0069] 31 Line of sight

[0070] 32 real light path

[0071] 40 phase object

[0072] 50 Assignment arrow

[0073] A Enlarged section

Claims

Patent claims 1. A method for capturing a scene (20) comprising a phase object (40) arranged in front of a textured background (21), in which a corresponding plurality of camera images (12) are simultaneously recorded by means of a plurality of cameras (10) arranged spatially offset from one another and directed through the phase object (40) onto the background (21) from different viewing directions, after which conclusions are drawn about the shape of the phase object (40) at the time of recording by evaluating the camera images (12), characterized by the following steps: a) providing, for each camera (10), an inverse of its imaging function, wherein the imaging function represents a unique assignment of each object point detectable by the respective camera to a pixel of the corresponding camera image (12), and the inverse represents a unique assignment of each pixel of the camera image to a line of sight,which is defined by all object points that are assigned to this pixel by the mapping function, b) identifying, in each camera image (12), that pixel (121, 121') onto which a background object point (221) mapped on each camera image (12), namely an object point belonging to the background (21), is mapped, and calculating, by means of the inverse of the respective mapping function, the respectively assigned line of sight (31), c) determining a measure for a distance of the calculated lines of sight (31) from one another, d) assigning a deviation value representing this measure to the corresponding pixel (121) in at least one camera image (12), and e) repeating steps b to d for a plurality of background object points (221).

2. Method according to claim 1, characterized by the further step f) creating a map (14) which is structurally analogous to the camera image (12), in which map point (141) which corresponds to an image point (121) to which a deviation value was assigned in one of the steps d carried out, is assigned a pixel value representing this deviation value.

3. Method according to claim 2, characterized in that map points (141) which correspond to image points (121) to which no deviation value was assigned in any of the steps d carried out are each assigned a pixel value which represents an interpolated or extrapolated deviation value, such map points (141) being used as support points for the respective interpolation or extrapolation to which a pixel value representing a deviation value was assigned in step f.

4. Method according to one of claims 2 to 3, characterized in that the map (14) is displayed on a display, each map point (141) being assigned a color or gray tone which corresponds to its respective pixel value according to a predetermined color or gray value coding.

5. Method according to one of claims 2 to 4, characterized in that the map (14) is visualized on a display in superposition with the associated camera image.

6. Method according to one of claims 2 to 4, characterized in that step d is carried out for a plurality of camera images (12), in particular at least three simultaneously recorded by different cameras (10). camera images (12) and in step f a corresponding plurality of maps (14) is created.

7. Method according to claim 6, characterized in that a three-dimensional shape of the phase object (40) is calculated from the created maps (14) and with knowledge of the relative spatial arrangement of the cameras (10).

8. The method according to claim 7, characterized in that the three-dimensional shape of the phase object (40) is shown on a display.

9. Method according to one of the preceding claims, characterized in that the method is repeated at one or more times in order to record the temporal development of the phase object (40).

10. Method according to one of the preceding claims, characterized in that at least three cameras (10) are used to simultaneously record a corresponding number of camera images (12).

Images