Method and camera for correcting geometric imaging errors in image recording

Abstract

In order to accurately compensate for influences on the imaging accuracy of the camera occurring during the operation of the camera, deformations (V) of the transparent cover (5) and / or the sensor layer (4) caused by a pressure difference (D) between an internal pressure (p1) in the sensor recess (3) and an ambient pressure (p2) of the camera (10) are modeled by a deformation model (20), the current prevailing pressure difference (D) at the time of image recording (B') by the camera (10) is determined by the internal pressure (p1) and the ambient pressure (p2) of the camera (10) known at the time of image recording (B'), and the deformation model (20) is used to determine the deformations (V) of the transparent cover (5) and / or the sensor layer (4) caused by a pressure difference (D) between the internal pressure (p1) in the sensor recess (3) and the ambient pressure (p2) of the camera (10) known at the time of image recording (B'). It is intended that, using the determined pressure difference (D), the deformation of the transparent cover (5) and / or the sensor layer (4) at the time of image recording (B') is determined, and from the determined deformation (V) and the known geometric dimensions of the image sensor (1), the deviation of the image of the object point (G) relative to the actual image point (P'') occurring in the image recording (B') is determined as a geometric imaging error (A), and the determined geometric imaging error (A) in the image recording (B') is corrected to determine a corrected image recording (B).

Description

[Technical field]

[0001] The invention relates to a method for correcting geometrical imaging errors of a camera with an image sensor, the image sensor being constituted by a sensor housing, a sensor layer which is arranged in a sensor recess of the sensor housing and has a number of image points, and a transparent cover which seals the sensor housing from the outside, whereby object points of an object recorded by the camera are imaged onto the image points of the sensor layer. The invention also relates to a correspondingly configured camera.

[0002] In digital cameras, image sensors, for example CCD (Charge Coupled Device) sensors or active image point sensors (also called CMOS sensors), are used to record digital images. The image sensor is arranged in the image plane of the digital camera and converts the incident light rays (photons) into electrical signals that are captured and represent the digital image. The image sensor typically consists of a suitable sensor housing, usually made of ceramic material, in which an electro-optical sensor consisting of a thin electro-optical sensor layer is arranged and a transparent cover is provided that closes the sensor housing. The sensor layer is usually glued and connected to the housing. To enable a clear image recording, it is essential that the image sensor is arranged in a defined position and orientation with respect to the optical system of the digital camera in order to enable a precise imaging of the object recorded by the digital camera in the image plane. According to Non-Patent Document 1, for example, a positional shift of an undeformed filter relative to the image plane of an image sensor leads to a shift (offset) in the imaging of object points in the image plane. In particular, applications of digital cameras in the fields of large-format photography, geomatics or photogrammetry place high demands on this imaging accuracy.

[0003] It is therefore already known to position an image sensor in a digital camera with high precision. For example, US Pat. No. 6,299,433 describes a sensor support adapted to an image sensor in order to enable high precision positioning of the image sensor in a digital camera using the sensor support. US Pat. No. 6,299,433 describes an orientation of an image sensor cover relative to the image sensor. In this case, the disadvantage is that the initial orientation or positioning is fixed and cannot be changed later.

[0004] Therefore, it is already known to movably or adjustably arrange an image sensor in a digital camera. This is shown, for example, in US Pat. No. 5,399,363. On the one hand, such an adjustment is structurally complex. On the other hand, such a mechanical adjustment can hardly achieve a high-precision orientation or positioning for the above-mentioned applications. Among other things, a deviation from a target position must also be determined during the active adjustment in order to be able to compensate for this deviation by the adjustment, which further increases the complexity.

[0005] US Pat. No. 5,399,633 follows a similar approach but with the same difficulties. From this document it is known to counteract the focus drift by performing a focus shift as a function of temperature and air pressure in order to improve the image sharpness of an aerial camera. In this case the required focus shift is modelled by means of a correction table, whereby a piezoelectric transformer device is controlled by means of an appropriate control signal in order to shift the focal plane along the optical axis in response to the focus drift. A comparable focus correction is also disclosed in US Pat. No. 5,399,633.

[0006] The effects on imaging accuracy that occur during the use of a digital camera cannot be compensated for at all with known approaches, as in US Pat. No. 5,399,623 or US Pat. No. 5,433,366, or cannot be compensated for at all with known approaches, as in US Pat. No. 5,433,623 or US Pat. No. 5,433,366, Or Patent Document 5 As such, compensation can be provided to a limited extent or at a limited cost. [Prior art documents] [Patent documents]

[0007] [Patent Document 1] International Publication No. 2021 / 069382 [Patent Document 2] European Patent Application Publication No. 1475960 [Patent Document 3] European Patent Application Publication No. 2485474 [Patent Document 4] DE 10013567 A1 [Patent Document 5] US Patent Application Publication No. 2018 / 373013 [Non-patent literature]

[0008] [Non-Patent Document 1] Nach Brauers, J., Schulte, N., and Aach, T. ("Multispectral filter wheel cameras: geometric distortion models and correction algorithms," IEEE transactions on image processing, 2008, pp. 2368-2380). Summary of the Invention [Problem to be solved by the invention]

[0009] Therefore, there is a need for a simple yet accurate correction for effects on the imaging accuracy of a digital camera that occur during operation of the digital camera. [Means for solving the problem]

[0010] This object is achieved by a method as specified at the beginning and by a digital camera as specified at the beginning, which are provided with the features of the independent claims.

[0011] It has been recognized that the image sensor, in particular the transparent cover and / or the sensor layer, is mechanically deformed by the pressure difference acting during image recording between the pressure in the sensor recess of the image sensor and the ambient pressure of the camera. This deformation shifts the image of the object point recorded by the camera in the image recording (or in the image recording in the sensor layer) in the form of a geometrical imaging error. This shift can be modeled so that it can be determined as a function of the pressure difference acting at the time of image recording and can be corrected in the image recording. For the modeling, a deformation model is used, which models the relationship between the pressure difference acting at the time of image recording and the deformation. The deformation model can be pre-calibrated for a particular image sensor, for example based on measurements on the image sensor, and can then be used during camera operation.

[0012] In this way, previously uncorrected geometric imaging errors in the image recording can be corrected, which makes it possible to improve the quality of the image recording.

[0013] Advantageously, the geometric imaging error is From the imaging of the object points to the predicted image points in the case of an undeformed transparent cover and / or an undeformed sensor layer, For the imaging of a real object point caused by the deformation, it is determined as the deviation of the imaging of the object point to the resulting real image point. This can be achieved in a straightforward manner, as these are simple geometric relationships on the image sensor that take into account the laws of refraction.

[0014] The quality of the image recording can be further improved if the geometrical imaging aberrations are also taken into account, which are caused by the optical refraction of the undeformed transparent cover, since already known geometrical imaging errors can also be corrected in the same step.

[0015] Correction of geometrical imaging errors in the image recording can be achieved in a simple manner by shifting the imaging of an object point occurring in the image recording relative to an actual image point in the corrected image recording to another image point by the geometrical imaging error.

[0016] The invention will now be explained in more detail with reference to the figures 1 to 6, which show exemplary, schematic and non-limiting advantageous embodiments of the invention. [Brief description of the drawings]

[0017] [Figure 1] FIG. 1 shows an embodiment of an image sensor comprising a sensor housing, a sensor layer, and a transparent cover. [Diagram 2] FIG. 2 shows the geometric imaging errors caused by deformation of the transparent cover. [Diagram 3] FIG. 3 shows the geometric imaging errors caused by deformations of the sensor layer. [Figure 4] FIG. 4 illustrates the use of an image sensor in a camera. [Diagram 5] FIG. 5 shows a scheme for correcting geometric imaging errors. [Figure 6] FIG. 6 shows an arrangement for determining or calibrating a deformation model. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0018] An image sensor 1 of a typical embodiment is depicted in Fig. 1. A sensor layer 4 constituting an electro-optical sensor is arranged in a sensor recess 3 of a sensor housing 2. The sensor housing 2 is usually manufactured from a ceramic material. The sensor housing 2 is sealed against the outside world with a transparent cover 5, usually made of glass. The sensor recess 3 is surrounded by the sensor housing 2 except for the part closed by the transparent cover 5. The sensor layer 4 is usually bonded to the sensor housing 2, for example at the bottom of the sensor recess 3, usually facing the transparent cover 5. The well known necessary electrical connections for electrically contacting the sensor layer 4 from the outside are not depicted in Fig. 1 for the sake of simplicity.

[0019] The sensor layer 4 comprises an array of electro-optical detectors, each of which constitutes an image point of the sensor layer 4 and therefore also of the image record B.

[0020] The sensor layer 4 may also be designed with several channels, for example as an RGB sensor. The individual electro-optical detectors of the sensor layer 4 may also be provided with different color filters in order to enable imaging of several distinct wavelength ranges of light by the image sensor 1. Implementations of such image sensors 1 are known.

[0021] Thus, based on the known optical properties of the optical system of the camera 10 with the image sensor 1 and based on the known geometry of the image sensor 1, the camera 10 and the positioning of the image sensor 1 within the camera 10, each object point G (see also FIG. 4) of an object recorded with the camera 10 with the image sensor 1 is imaged to an image point P of the sensor layer 4 in the image plane 7 and can be assigned to an image point P of the camera 10. This is illustrated in FIG. 1. The optical system of the camera 10 includes an objective lens 11 and may include other optical components such as a diaphragm, optical filters, a shutter, etc. The center of projection O of the objective lens 11 (not shown in FIG. 1) is located at a known focal length f from the image plane 7. B The sensor layer 4 of the image sensor 1 is arranged in an image plane 7.

[0022] The center of projection O does not have to be located in the center of the sensor layer 4, but the sensor layer 4 can be located off-center and / or asymmetrically with respect to the center of projection O. It is also conceivable that several image sensors 1 are arranged in the camera 10, such that the image sensor 1, or the sensor layer 4 of the image sensor 1, is arranged off-center with respect to the center of projection O and with respect to an optical axis extending perpendicularly through the center of projection O to the image plane 7.

[0023] A light ray 6 starting from an object point G of the recorded object and extending through the center of projection O refracts in the transparent cover 5 according to known laws of refraction and, depending on the known refractive indices present (e.g. the air-to-glass transition at the outer interface 5a and the glass-to-air transition at the inner interface 5b), strikes the image plane 7 not at the actual image point P but at an image point P', displaced by an offset L.

[0024] However, the objective lens 11 collects all the light rays starting from an object point G on the object being recorded and passing through the lens 11 (see FIG. 4) in a manner which is of course known per se to an image point P. However, for the sake of simplicity, only the ray 6 of this light bundle which extends through the center of projection O is depicted in FIG.

[0025] It should also be noted that the refractive index at the outer interface 5a of the transparent cover 5 is not necessarily the same as the refractive index at the inner interface 5b of the transparent cover 5 into the sensor housing 2. However, the refractive index at both optical interfaces 5a, 5b can be assumed to be known.

[0026] The offset L results from the known refractive index which follows the law of refraction and the known geometric dimensions of the image sensor 1 , in particular the thickness of the transparent cover 5 and the spacing between the inner interface 5 b of the transparent cover 5 and the sensor layer 4 .

[0027] For example, it is known to mathematically represent the effect of optical refraction at the transparent cover 5 of the image sensor 1 on the recorded image and compensate in the image record B by shifting the imaging of an object point G from image point P' in the image record B' to image point P in the image record B by an offset L. This geometric imaging error caused by refraction becomes larger the more the incidence angle δ1 of the light ray 6 deviates from the surface normal of the transparent cover 5. The offset L is therefore typically not constant over the entire image plane 7.

[0028] The image sensor 1, or sensor layer 4, is typically two-dimensional, so that the offset L at the image plane 7 is caused by refraction in two coordinate directions (e.g., x and y, see FIG. 6). The offset L (in both coordinate directions) at the image point P in the image plane 7 results from the angle of incidence δ1 of the light ray 6, the refractive index present, the geometry of the image sensor 1, and possibly also from the position of the image sensor 1 in the camera 10 (especially if the image sensor is positioned off-center with respect to the optical axis).

[0029] However, in (high) precision applications such as geomatics or photogrammetry, it has been found that, despite compensation of the refraction-induced offset L, a perceptible geometric imaging error A remains in the image B, which limits the quality of the image recording B. It has been realized that this remaining geometric imaging error A is due to the design of the image sensor 1.

[0030] After the sensor recess 3 in which the sensor layer 4 is located is sealed by the transparent cover 5, there is, for manufacturing reasons, a certain gaseous medium and a certain internal pressure p1 in the sensor recess 3. If, for example, an image recording is made from an aircraft using the image sensor 1, an ambient pressure p2 different from the internal pressure p1 prevails in the area of ​​the camera 10 with the image sensor 1 in the aircraft. Since the manufacturing of the image sensor 1 takes place at an altitude significantly lower than the flying altitude of the aircraft at the time of image recording, the internal pressure p1 confined in the sensor recess 3 is usually greater than the ambient pressure p2 when the camera 10 is used. However, the internal pressure p1 may of course be less than the ambient pressure p2 when the camera 10 is used. This can cause the transparent cover 5 to deform due to the pressure difference D. If the internal pressure p1 in the sensor recess 3 is greater than the ambient pressure p2, the transparent cover 5 will, for example, bend outwards. If the sensor housing 2 is not sufficiently fixed, the sensor housing 2, and thus also the sensor layer 4, can deform, for example, bend outwards, if the internal pressure p1 is greater than the ambient pressure p2.

[0031] The pressure difference D is preferably expressed as a pressure difference (p2-p1), as a power of the pressure difference, or as a quotient of two pressures (p2 / p1), although different expressions of the pressure difference D are possible. The pressure difference D can also be expressed as the difference between the pressure difference at a first altitude and the pressure difference at another altitude.

[0032] However, the internal pressure p1 can also be influenced by other physical effects, in particular by the temperature in the region of the image sensor 1. According to Amonton's law, the pressure of a gas trapped in a certain volume depends on the temperature T according to the relationship P1(T)=P0(1+γ0(T1-T0)). Here, p1 denotes the pressure at temperature T1, which corresponds to the prevailing internal pressure p1, T0 corresponds to an initial temperature (for example a temperature prevailing during the manufacture of the image sensor or a hypothetical specific temperature) and γ0 corresponds to the volume expansion coefficient at the initial temperature T0, which may be assumed to be known for the gas trapped in the image sensor 1. The temperature T can for example be measured by the temperature sensor 13 of the image sensor 1. Amonton's law can be expressed in the simplified formula

number

[0033] For the purposes of applying Amonton's law, the volume within the image sensor 1 can be considered to be approximately constant since the image sensor 1 only deforms slightly, but this effect can also be modelled, for example, by applying the known heat equation for an ideal gas.

[0034] The effect of the deformation caused by the pressure difference D on the image sensor 1 will be explained with reference to Figures 2 and 3, with the deformation being exaggerated in order to concretely show this effect.

[0035] In FIG. 2, a part of the transparent cover 5 is depicted in dashed lines in an initial, undeformed position. An incoming light ray 6 from an object point G is refracted at the transparent cover 5 at the two interfaces 5a, 5b according to the laws of refraction and leaves the transparent cover 5 as a light ray 6' which strikes the sensor layer 4 at an image point P'. The deformation of the transparent cover 5 caused by the pressure difference D (in the illustrated embodiment, the transparent cover 5 is bent outward) changes the orientation of the normal to the interfaces 5a, 5b. This changes the angle δ1 of the incoming light ray 6 with respect to the normal to the outer interface 5a, and also the angle δ3 of the outgoing light ray 6 with respect to the normal to the inner interface 5b. The light ray 6 is therefore refracted differently due to the deformation compared to the undeformed transparent cover 5. The outgoing light ray 6'' thus strikes the sensor layer 4 at a different image point P'', resulting in a geometric imaging error A. Therefore, the object point G is no longer imaged at an image point P′ (as expected), but at an actual different image point P″ of the sensor layer 4 .

[0036] The use of reference numbers 6', 6'' for the refracted and exiting rays is for illustrative purposes only; however, ray 6 of course remains the same.

[0037] In Fig. 3 a part of the undeformed sensor housing 2 with the undeformed sensor layer 4 is depicted with dashed lines. A light ray 6 coming from an object point G strikes the undeformed sensor layer 4 at an image point P'. Due to the deformation of the sensor layer 4 caused by the pressure difference D (in the illustrated embodiment the sensor layer 4 is bent outwards) the light ray 6 strikes the sensor layer 4 at an image point P'', which also results in a geometric imaging error A. Thus the object point G is not imaged to the image point P' (as expected) but to an actual separate image point P'' of the sensor layer 4.

[0038] A geometrical imaging error A is therefore understood to be the deviation of the imaging of an object point G from an expected image point P′ of the sensor layer 4 to an actual different image point P″ of the sensor layer 4 .

[0039] The two effects of the deformations described with reference to Figures 2 and 3 may occur together (as is usually the case) but also individually. A geometric imaging error A may therefore be caused by one of these effects or by both effects. The resulting geometric imaging error A is manifested for example as a distortion and / or aberration of the recorded image. Such a geometric imaging error A may occur at each image point P of the image plane 7 or may occur only at certain pixels P, but is not constant over the entire image plane 7.

[0040] Knowledge of the deformation V of the transparent cover 5 and / or the sensor layer 4 in the sensor housing 2 due to the pressure difference D between the internal pressure p1 in the sensor recess 3 and the ambient pressure p2 of the camera 10, which is prevailing at the time of the image recording made by the camera 10 with the image sensor 1, can be used to correct the geometrical imaging errors A occurring in the image recording B' in order to obtain a corrected image recording B. For this purpose, an evaluation device 21 can be provided, which may be part of the camera 10.

[0041] For example, the start of exposure, the end of exposure, or a point between the start of exposure and the end of exposure can be considered as image recording time B'.

[0042] The geometric imaging error A can be determined and corrected for each image point P (pixel) of the digital image, but does not need to be determined and corrected for each image point P.

[0043] In this case, with reference to Figures 4 and 5, the following can be done.

[0044] It can be assumed that the internal pressure p1 of the image sensor 1 at the time of image recording B' is known. For example, the internal pressure p1 corresponds to the prevailing pressure at the time of manufacturing the image sensor 1, which prevailing pressure can be approximated by a reference state (T0=273 K, p1=1.01325 bar). For example, the influence of any air temperature on the internal pressure p1 can be determined using Amonton's law. To this end, the current prevailing temperature at the time of image recording B' in the image sensor 1 can be known by the temperature sensor 13, and the current internal pressure p1 can be calculated.

[0045] Likewise, the ambient pressure P2 in the case of the image sensor 1 or in the case of the camera 10 at the time of image recording B' can be assumed to be known. For example, the camera 10 is fixed to a vehicle 12, for example an aircraft. The ambient pressure p2 can be known by a pressure sensor, for example in the case of a road vehicle, or can be determined by an atmospheric pressure-altitude equation, for example in the case of an aircraft, which comprises a known flight altitude h. Naturally, the ambient pressure p2 can also be known in the aircraft by means of a pressure sensor. If the camera 10 is arranged in a pressurized cabin of an aircraft, the ambient pressure p2 corresponds for example to the pressure in the pressurized cabin. However, the camera 10 does not necessarily have to be arranged in the vehicle 12.

[0046] That is, the current prevailing pressure difference D at the time of image recording B' can be assumed to be known or can be determined.

[0047] A deformation model 20 is used that models the deformation V of the transparent cover 5 and / or the sensor layer 4 caused by a pressure difference D between an internal pressure p1 in the sensor recess 3 of the image sensor 1 and an ambient pressure p2 of the image sensor 1 or camera 10. The deformation V can generally be represented by the deformation model 20 as a function f of the pressure difference D, i.e., V=f(D). The deformation model 20 represents the relationship between the pressure difference D and the deformation V of the transparent cover 5 and / or the sensor layer 4.

[0048] The deformation model 20 can be implemented, for example, in an evaluation device 21 (FIG. 5). The evaluation device 21 can be a computer or microprocessor-based hardware. Also conceivable is an embodiment as an integrated circuit, such as an FPGA (Field Programmable Gate Array) or an ASIC (Application Specific Integrated Circuit). The deformation model 20 can be implemented as software running on a microprocessor, but can also be stored in tabular form or as a formula or function.

[0049] The particular implementation of the deformation model 20 is not critical to the invention, and a wide variety of models are contemplated, as described in more detail below.

[0050] Using the deformation model 20, the deformation V of the transparent cover 5 and / or the sensor layer 4 during image capture B' can be determined from the known pressure difference D during image recording B'.

[0051] From the determined deformation V and the known geometric dimensions of the image sensor 1, it can be determined at which image point P'' of the sensor layer 4 the light ray 6 captured by the camera 10 actually strikes the sensor layer 4 and at which image point P' the light ray 6 should strike the sensor layer 4 without the deformation V. This can be easily determined, since it is a simple geometrical relationship. This determined deviation between the actual image point P'' and the expected image point P', which represents the geometrical imaging error A of the image point P, can be corrected in the image recording B' in order to correct the geometrical imaging error A. For this purpose, the imaging of the object point G is shifted from the image point P'' in the image recording B', for example, by the geometrical imaging error A, to the image point P' in the corrected image recording B. Also, the offset L, which also occurs with the undeformed image sensor 1 (described above with respect to FIG. 1), can be corrected at the same time. However, this offset L can also be corrected later. However, this offset L can also be taken into account immediately in the geometrical imaging error A, so that it is immediately corrected for the image point P by the geometrical imaging error A.

[0052] In this description, the geometrical imaging error A of an image point P is simply described as a deviation in one coordinate direction (x or y), but it is clear that due to the deformation V, the geometrical imaging error A usually occurs in both coordinate directions of the sensor 4 for an image point P. That is to say, when speaking of correcting the geometrical imaging error A, it means (and usually does) a deviation in only one coordinate direction of the sensor layer 4 or in both coordinate directions.

[0053] The correction of the geometrical imaging errors A in the image recording B' due to deformations V of the transparent cover 5 and / or the sensor layer 4 can be carried out by a correction device 22. The correction device 22 may be integrated in the evaluation device 20 or may be implemented, for example, in the form of software. However, the correction device 22 may also be hardware (computer, microprocessor-based hardware, integrated circuit) separate from the evaluation device 21.

[0054] The image record B corrected for the geometric imaging error A can then be reused as required, for example by subjecting the corrected image record B to further optical corrections, such as known deblurring, or by evaluating the corrected image record depending on the application.

[0055] The correction of the geometrical imaging error A of the image point P'' can take place immediately after the image recording B' or even at a later time. If the correction is only made at a later time, the pressure D prevailing at the time of the image recording B' is also stored together with the image recording B' in order to be able to determine the later deformation V.

[0056] To determine the deformation model 20, one may proceed as follows, with reference to FIG.

[0057] The image sensor 1 detects a certain pressure p D A measuring device 31 is arranged in or on the pressure chamber 30, by measuring the deformation V of the transparent cover 5 and / or the sensor layer 4 or of the sensor housing parts attached to the sensor layer 4.

[0058] The measurement of the deformation V can be carried out by means of any direct or indirect, preferably non-contact, measuring device 31 suitable for this purpose. Examples of suitable measuring devices are mechanical measuring devices that determine the shape by tactile scanning, e.g. micrometers or measuring gauges, optical measuring devices with image sensors, e.g. 1D, 2D, 3D laser displacement sensors, line sensors and / or pixel sensors, measuring devices with imaging sensors, whereby the deformation is determined by image evaluation, and measuring devices using a combination of the above mentioned measuring methods. A preferred embodiment of the measuring device 31 is an optomechanical measuring device, where the positioning of the image sensor 1 is performed mechanically and the measurement is performed optically. For example, the optomechanical measurement of the image sensor 1 can be performed via known industrial optical 3D measuring systems. An example of an optomechanical measuring device that can be used is the device called "Vertex 251 UC" manufactured by the American company Microview (Windsor, California).

[0059] In order to be able to carry out the deformation V of the transparent cover 5 using optical methods, it is also possible to place visible points on the transparent cover 5 during the measurement.

[0060] For a certain pressure difference D set in the pressure chamber 20, the deformation V of the transparent cover 5 and / or the sensor layer 4 is measured by the measuring device 31 at a number of measurement points M (shown in FIG. 6 for the transparent cover 5). The majority of the measurement points M are distributed in the plane of the transparent cover 5 and / or the sensor layer 4 (e.g. in the xy plane in FIG. 6) in order to image the deformation V also two-dimensionally. The positions of the measurement points M can for example be determined relative to a coordinate system on the imaging sensor 1.

[0061] However, it is not important which measurement is used to represent the deformation V. For example, the deformation V can be measured as the distance between an undeformed reference position and the actual position of the measurement point of the transparent cover 5 and / or the sensor layer 4 (e.g. in the z direction in FIG. 6). From this, the normal to the interface 5a, 5b of the deformed transparent cover 5 with respect to the transparent cover 5 can also be determined. Also, the difference between the normal to the interface 5a, 5b in the undeformed reference position and for a certain pressure difference D can be determined and represented by the deformation V. For example, the normal can be specified as a vector in the coordinate system of the image sensor 1.

[0062] The deformation V measured at the measurement points M can be stored in tabular form and used as the deformation model 20. Then, when using the deformation model 20, it is possible to interpolate between the measurement points M.

[0063] However, it is of course also possible to use as deformation model 20 a mathematical function that approximates the determined deformation V at the measurement points M. The measurement points M or the selection of the measurement points M should be interpolated by a mathematical function. For this, it may be advantageous to arrange the measurement points M in a grid in a plane. For example, a polynomial of a certain degree (e.g. a polynomial of the sixth degree) or a free-form curve such as a Bézier curve, a piecewise polynomial function (spline) can be used as the mathematical function. The approximation of certain support points (measurement points M) of a curve by such a mathematical function is sufficiently well known and does not need to be described in more detail here. This can of course also be extended to surfaces, i.e. two-dimensional approximations. For this, two-dimensional mathematical functions are used (e.g. a two-dimensional polynomial of a certain degree or a free-form surface modeling a surface).

[0064] To represent the deformation V of the entire transparent cover 5 and / or the sensor layer 4 in the deformation model 20, several mathematical functions can be used, each covering a certain area of ​​the transparent cover 5 and / or the sensor layer 4. For example, several mathematical functions distributed in the y direction are used, each mathematical function being applied to a certain y coordinate and approximating several measurement points M in the x direction to this y coordinate. This can also be done two-dimensionally, if several mathematical functions distributed in the x direction and distributed in the y direction are used. Such two-dimensional modeling can increase the accuracy of the deformation model 20.

[0065] The mathematical functions in both directions can be multiplicatively linked: at any position of the image sensor 1 covered by the mathematical functions, the deformations of each can be determined by the two mathematical functions, which are then multiplied to obtain the total effect for this position.

[0066] In an exemplary implementation of the deformation model 20, a polynomial of degree 6 (or another degree) can be used to approximate the measurement points M in a certain direction, for example the x-direction or the y-direction. In this case, a number of such polynomials that model the in-plane deformation V are obtained in the xy-plane of the image sensor 1. The measurement points M are of course appropriately selected for this purpose. The polynomials can be used to interpolate between the measurement points. The polynomials can be stored in the form of their polynomial parameters that are determined when approximating a curve.

[0067] To increase accuracy, a polynomial of degree 6 (or another degree) can be used in both directions, e.g., in the x and y directions, so that the deformation V can be modeled more accurately even at the surface.

[0068] If a mathematical function is used to approximate the deformation V, it is self-evident that the deformations V of the transparent cover 5 and the sensor layer 4, when both are modeled, are usually approximated by different functions. In this case, "different" may also mean that the same function class (e.g., a 6th order polynomial) is used, but the function parameters are different.

[0069] The deformations V of the transparent cover 5 and / or the sensor layer 4 are preferably modelled by the deformation model 20 for various pressure differences D. For this purpose, a deformation V may be measured for various pressure differences D, for example by the measuring device 31, and modelled for each pressure difference D used in the deformation model 20. In use, it is possible to interpolate between the deformations V determined for a particular pressure difference D, or the deformation V for the closest pressure difference D stored in the deformation model 20 for the current pressure difference D is taken into account.

Claims

1. A method for correcting a geometric imaging error (A) of a camera (10) having an image sensor (1), comprising the steps of: The image sensor (1) A sensor housing (2); a sensor layer (4) arranged in a sensor recess (3) of the sensor housing (2) and having a plurality of image points (P); A transparent cover (5) that seals the sensor housing (2) from the outside; It is composed of In this case, an object point (G) of the object recorded by the camera (10) is imaged onto an image point (P) of the sensor layer (4). In the method, The internal pressure (p １ ) and the ambient pressure (p ２ ) is modelled by a deformation model (20), The current prevailing pressure difference (D) at the time of image recording (B') by the camera (10) is calculated based on the known internal pressure (p １ ) and the ambient pressure (p ２ ) and By means of the determined pressure difference (D), the deformation of the transparent cover (5) and / or the sensor layer (4) at the time of image recording (B') is determined by means of a deformation model (20), From the determined deformation (V) and the known geometrical dimensions of the image sensor (1), the deviation of the imaging of the object point (G) on the actually generated image point (P'') in the image record (B') caused by the deformation is determined as the geometrical imaging error (A), the determined geometrical imaging errors (A) in the image record (B') are corrected to determine a corrected image record (B); A method comprising:

2. The geometric imaging error (A) is For the actual imaging of an object point (G) to a real image point (P'') resulting from the deformation, In the case of an undeformed transparent cover (5) and / or an undeformed sensor layer (4), it is determined as the deviation of the image of the object point (G) relative to the predicted image point (P'), 2. The method of claim 1 .

3. Geometric imaging error (A) is The offset (L) caused by the optical refraction of the undeformed transparent cover (5) is also taken into account.

3. The method according to claim 1 or 2.

4. The geometrical imaging error (A) in the image recording (B') is corrected by shifting the imaging of the object point (G) relative to the actual image point (P'') occurring in the image recording (B') to another image point (P, P') by the geometrical imaging error (A) in the corrected image recording (B). The method according to any one of claims 1 to 3.

5. A camera (10) having an image sensor (1) and an evaluation device (21) for processing an image record (B') produced by the camera (10), The image sensor (1) A sensor housing (2); a sensor layer (4) arranged in a sensor recess (3) of the sensor housing (2) and having a plurality of image points; A transparent cover (5) that seals the sensor housing (2) from the outside; It is composed of In this case, an object point (G) of the object recorded by the camera (10) is imaged onto an image point (P) of the sensor layer (4). In the camera (10), A deformation model (20) is implemented in the evaluation device (21), and the deformation model (20) is configured to calculate the internal pressure (p １ ) and the ambient pressure of the camera (10) (p ２ ) modeling the deformation (V) of the transparent cover (5) and / or the sensor layer (4) caused by a pressure difference between the The evaluation device (21) calculates the current prevailing pressure difference (D) at the time of image recording (B') by the camera (10) in relation to the known internal pressure (p １ ) and the ambient pressure (p ２ ) is determined by and determining, by means of a deformation model (20), from the determined pressure difference (D), the deformation of the transparent cover (5) and / or the sensor layer (4) at the time of image recording (B'); The evaluation device (21) determines the deviation of the image of the object point (G) resulting from the deformation (V) as a geometrical imaging error (A) relative to the actually occurring image point (P'') in the image record (B'), a correction device (22) is provided for correcting the determined geometrical imaging errors (A) in the image record (B') in order to determine a corrected image record (B); A camera characterized by:

6. The evaluation device (21) For the actual imaging caused by the deformation (V) and resulting in an actual image point (P″), determining the geometrical imaging error (A) as the deviation of the imaging of the object point (G) relative to the expected image point (P') in the case of an undeformed transparent cover (5) and / or an undeformed sensor layer (4); 6. The camera according to claim 5.

7. the evaluation device (21) also takes into account the offset (L) caused by the optical refraction of the non-deformed transparent cover (5) within the geometrical imaging error (A), 7. The camera according to claim 5 or 6.

8. a correction device (22) corrects a geometrical imaging error (A) in the image record (B') by shifting the imaging of the object point (G) relative to the actual image point (P'') occurring in said image record (B') to another image point (P, P') by the geometrical imaging error (A) in the corrected image record (B); The camera according to any one of claims 5 to 7.

Images