Noise minimization method and noise-minimizing measurement system

Abstract

Method for minimizing noise during data acquisition with a measuring system (100) comprising the steps: Acquiring data of a reference measurement with the measuring system (100) under defined conditions, wherein the conditions include at least a total measurement time, a rate of individual measurements per time interval, an integration time per measurement, and a temperature of an array of measuring elements (102) of a measuring device (101) of the measuring system (100), wherein the data acquisition of the reference measurement comprises acquiring a plurality of individual measurements according to a reference measurement sequence; Deriving an estimator for a measure of the noise of the measuring system (100) based on measured values ​​of the reference measurement; Determining an optimized measurement sequence taking into account boundary conditions (170), which include at least one possible total measurement time and a rate of individual measurements per time interval, for intended measurements with a view to minimizing a measure of the noise, wherein the data acquisition is carried out with an imaging measuring system (100) with an array of infrared-sensitive measuring elements (102), in front of which a controllable shutter (110) is arranged such that, in a closed state of the shutter (110), none of the measuring elements (102) is exposed to radiation from a measured scene or a measured object (120), and in an open state, at least some of the measuring elements (102) are exposed to radiation from the measured scene or the measured object (120), wherein the measuring process is carried out via a A shutter sequence function is defined which, for a sequence of individual measurements acquired one after the other in time, defines which individual measurements are acquired with the shutter (110) open and which with the shutter (110) closed, wherein the estimation function for the measure of the noise is derived from a spatial-temporal autocovariance function of the measuring system (100), which is acquired on the basis of the reference measurement with the reference measurement sequence in which a sequence of individual measurements acquired one after the other in time with the shutter (110) closed.

Description

[0001] The invention relates to a method for noise minimization and a corresponding noise-minimizing measuring system, in particular a method for noise reduction in infrared measuring systems and especially in two-dimensional imaging infrared measuring systems.

[0002] It is known from the prior art that measurement systems that measure in the infrared wavelength range, particularly those that measure in the thermal infrared range, can lead to measurement errors due to interference. For example, the housing of the measuring device itself represents a radiation source. While the influence of this interference source can be reduced by actively cooling the housing, it cannot usually be completely eliminated. Other components, such as the measuring device itself and its evaluation electronics, also influence the measurement results of the measuring system. All components of a measurement signal that do not come from the signal source actually being measured are referred to here as noise components or generally as noise. Imaging measuring systems that measure the thermal radiation of an object with spatial resolution are used, for example, in exploration satellites.Of course, it is desirable to obtain measured values ​​that have only a low noise component.

[0003] It is known from the prior art, for example in CCD sensor-based measurement systems, to provide a mechanical shutter and to record measurement results with the shutter closed on the one hand and measurements with the shutter open on the other. By averaging the individual recorded measurements and calculating the difference, a significant reduction in the noise component can be achieved.

[0004] The article "MERTIS - Background Signal Removal and Signal Simulation" by Thomas Säuberlich et al., Proc. of SPIE Vol. 7453 745306, 2009, describes the MERTIS (MERcury Thermal Infrared Imaging Spectrometer) system, an advanced infrared remote sensing instrument that is part of ESA's BepiColombo mission to the planet Mercury. The technology that will enable the first thermal infrared spectrometer to be sent to Mercury is an uncooled microbolometer. This detector allows the instrument to operate in Mercury's hot environment without the need for a cryogenic cooling system. The challenge lies in calibrating the instrument.A radiometric and a spectroscopic experimental setup model of MERTIS were used to develop suitable calibration methods and derive system parameters to support the construction of an end-to-end simulation capable of processing spectra of analog planetary materials from the DLR Planetary Emissivity Laboratory (PEL) as input to generate a realistic representation of the MERTIS output signal. Calibration attempts to remove the background signal contained in the raw image datasets, which actually represents the dominant signal component. A method for measuring the background using a shutter and a method for noise reduction based on a pixel-by-pixel correlation approach are discussed and related to the residual errors of the emission spectra calculated from raw images of laboratory experiments using ground calibration datasets.The results of the error evaluation and new emission spectra from the PEL for high temperatures of planetary analog materials are input parameters for the end-to-end simulation of MERTIS. Regarding the instrument's signal-to-noise ratio (SNR), a comparison of the simulation results with the experimental data and the effect of the noise suppression method are presented.

[0005] The article "MERTIS - Shutterless Background Signal Removal" by Thomas Säuberlich et al., Proc. Of SPIE Vol. 7808 78080L, 2010, describes how MERTIS attempts to separate non-scene and scene signal components during calibration. The non-scene signal component is contained in the raw image data sets and usually represents the dominant signal contribution. The conventional method of measuring the non-scene signal components using a shutter or space view and performing time interpolation is compared with an approach that uses linear pixel-to-pixel relations, in which information from the outer regions of the image matrix is ​​used to estimate the non-scene signal components of the inner regions where additional scene signal components are present. The results of both methods are discussed with regard to noise or errors in the extracted scene information.The proposed method can be used without further modifications to the instrument and provides functional redundancy, which is important to maintain MERTIS operation in case of failure of the mechanically loaded high-speed shutter.

[0006] US 7,557,915 B2 describes a system and method for automatically obtaining spectra for samples. The method comprises a two-step process with a photobleaching phase and a spectral acquisition phase. In the photobleaching phase, a series of spectral data sets of a sample is collected. A relative difference between the background of the consecutive spectral data sets is determined and compared to a predetermined threshold. If the threshold difference is less than the relative difference between the background of the consecutive spectral data sets, the steps for acquiring a series of spectral data sets are automatically repeated. In the spectral acquisition phase, a series of Raman data sets of the sample are acquired until a target signal-to-noise ratio is reached.

[0007] US 7,605,918 B2 describes a method for operating a spectrometer. The operator of a spectrometer can specify a desired signal-to-noise ratio (SNR) to be achieved when acquiring spectra from a sample. The SNR of a single, short sample exposure is used to determine the maximum exposure time that can be achieved without overloading the spectrometer. If the desired SNR is greater than the SNR of an exposure with the maximum exposure time, multiple exposures with the maximum exposure time can be acquired and combined (e.g., averaged or summed) to obtain a spectrum with an SNR that at least approximates the desired SNR.If the desired SNR is less than the SNR of an exposure with the maximum exposure time, only a single exposure is required, and the exposure time can be scaled using the SNR of the single short sample exposure to obtain an SNR at least close to the desired one.

[0008] The article "Performance model for uncooled infrared bolometer arrays and performance predictions of bolometers operating at atmospheric pressure" by Frank Niklaus et al., Infrared Physics & Technology, 51 (2008), pages 168-177, presents a comprehensive computational model for the noise equivalent temperature difference (NETD) of infrared imaging systems based on uncooled bolometer arrays. The equations are presented in a novel and practical form. The NETD model is validated and benchmarked against published performance data from a state-of-the-art uncooled infrared bolometer array. The NETD model is used to evaluate potential improvements to the system and bolometer design.The calculation results indicate that infrared imaging systems based on uncooled bolometer arrays with a bolometer pixel pitch of 28 µm x 28 µm have the potential to achieve NETDs on the order of 12 mK. The calculations also suggest that NETDs on the order of 200 mK can be achieved with infrared imaging systems based on uncooled bolometer arrays operating in air at atmospheric pressure.

[0009] The invention is based on the object of improving the noise reduction or noise reduction for measuring systems, in particular those measuring systems which measure in the thermal infrared wavelength range and even beyond that by imaging, compared to the known methods and devices.

[0010] The invention is achieved by a method having the features of patent claim 1 and a measuring system having the features of patent claim 5. Advantageous embodiments of the invention emerge from the subclaims.

[0011] The invention is based on the idea of ​​acquiring, similarly to the prior art, measured values ​​that comprise a signal to be measured and measured values ​​that only comprise noise, for example, acquiring individual measurements with the shutter closed and, on the other hand, acquiring individual measurements with the shutter open.

[0012] The invention is based on the finding that noise reduction with simple averaging and subtraction is not optimal for all measuring systems. Rather, an optimal measuring sequence consisting of measured values ​​with signal and those without signal, i.e., for example, consisting of individual measurements recorded with the shutter open and individual measurements recorded with the shutter closed, depends on the noise characteristics of the measuring device used. Furthermore, boundary conditions often influence a measuring sequence. For example, with a measuring device installed in a satellite or a probe, only a limited number of individual measurements with signal can be recorded from one and the same object or an area of ​​the object, e.g. with the shutter open, without impairing spatial resolution because the measuring device is moved relative to the object.

[0013] In particular, a method for noise suppression on an IR measuring device is proposed, which comprises the steps of: acquiring a reference measurement under defined conditions; deriving an estimator for a noise behavior; acquiring boundary conditions for a measurement sequence; optimizing the measurement sequence taking into account the boundary conditions with a view to minimizing the noise by utilizing the estimator for the noise behavior; acquiring measurement data using the optimized measurement sequence and processing the measurement data to remove noise components.

[0014] In particular, an embodiment of a method for noise minimization during data acquisition with an infrared-sensitive measuring system is proposed, which comprises the steps: Acquiring a reference measurement with the infrared-sensitive measuring system under defined conditions, wherein the acquisition of the reference measurement comprises acquiring a plurality of individual measurements according to a reference measurement sequence; Deriving an estimation function for the noise behavior of the measuring system based on measured values ​​of the reference measurement; Determine an optimized measurement sequence, taking into account the boundary conditions for the intended measurements with a view to minimizing noise. This results in an optimal measurement sequence that is optimized in terms of noise and resource utilization.

[0015] Furthermore, a measurement system with improved noise minimization is proposed, which comprises: a measuring device with at least one infrared-sensitive measuring element; a control device which controls the acquisition of a reference measurement under defined conditions according to a measurement sequence, wherein the acquisition of the reference measurement comprises the acquisition of a plurality of individual measurements according to a reference measurement sequence; a noise characteristic prediction device which is designed to derive an estimation function for a noise behavior of the measuring system based on measured values ​​of the reference measurement; a sequence selection device which is designed to determine an optimized measurement sequence for intended measurements taking into account boundary conditions for the intended measurements with a view to minimizing noise.

[0016] In a further development of the method and the device, it is provided that the intended measurements are carried out by acquiring measurement data using the optimized measurement sequence.

[0017] According to the invention, the estimation function for the noise behavior is derived from a spatial-temporal autocovariance function of the measuring system.

[0018] Accordingly, the noise characteristic prediction device is designed to derive the estimation function of the noise behavior from a spatial-temporal autocovariance function of the measuring system.

[0019] It is intended that the measuring system comprises at least one infrared-sensitive measuring element in front of which a controllable shutter opening is arranged and the measuring sequence determines via a shutter sequence function which individual measurements are recorded with the shutter open and which individual measurements are recorded with the shutter closed.

[0020] It is advantageous to determine a measurement result of a measurement from a fusion of measured values ​​during the acquisition of which the corresponding measuring element is exposed to the radiation of a measured scene or a measured object, and measured values ​​during the acquisition of which the corresponding measuring element is not exposed to the radiation of the measured scene or the measured object.

[0021] If the measured values ​​during which the corresponding measuring element is exposed to the radiation of the measured scene or object are called scene measured values, and the measured values ​​during which the corresponding measuring element is not exposed to the radiation of the measured scene or object are called reference measured values, then the fusion of measured values ​​usually occurs by calculating the difference between the scene measured values ​​and the reference measured values. If the acquisition times or numbers of the individual measurement result types differ, a corresponding weighting is applied during the fusion to take this into account.

[0022] Furthermore, the IR measuring device is equipped with a controllable shutter and a measurement sequence is structured in such a way that it includes a sequence of data acquisitions with the shutter open and closed.

[0023] In a preferred embodiment, which also enables imaging measurements, it is provided that the measuring system is a two-dimensional imaging measuring system and comprises a two-dimensional array of measuring elements, and a controllable shutter is arranged in front of the array of measuring elements in such a way that, when the shutter is closed, none of the measuring elements is exposed to radiation from a measured scene or a measured object, and when it is open, at least some of the measuring elements are exposed to radiation from the measured scene or the measured object, wherein the measuring sequence is defined by a shutter sequence function which defines a measurement for a sequence of individual measurements recorded one after the other in time, which individual measurements are recorded with the shutter open and which with the shutter closed.

[0024] In one embodiment, the measuring device comprises an array of infrared-sensitive measuring elements arranged in a matrix comprising columns and rows. A shutter is preferably arranged in front of the measuring element(s) of the measuring device, and the control device comprises a sequence control device that controls the shutter and the measurement value acquisition of the measuring elements of the measuring device. The measuring elements are particularly preferably microbolometers. However, other detectors capable of detecting radiation intensity, in particular also in the infrared wavelength range, can also be used as measuring elements.

[0025] A measuring device with microbolometers is particularly suitable for carrying out thermal measurements in the infrared wavelength range.

[0026] According to the invention, the reference measurement is recorded using a reference sequence in which a sequence of chronologically successive individual measurements is recorded with the shutter closed.

[0027] An embodiment has proven particularly advantageous in which the estimation function for the noise behavior is formed by a root of a sum of a spatial-temporal autocovariance function for measurement results, weighted by factors dependent on the measurement sequence, depending on the time difference between the measurements of these measurement results.

[0028] For example, when used on a spacecraft, it is advantageous to be able to capture and record boundary conditions for measurement sequences of intended measurements. In one embodiment of a measurement system, the sequence selection device therefore includes an interface for capturing the boundary conditions.

[0029] The measurement system of a preferred embodiment is configured such that the sequence selection device determines an optimized sequence by determining a measure of predicted noise for different shutter sequences or shutter sequence functions and selecting the shutter sequence function that is predicted to produce minimal noise under the given boundary conditions. Processing the measurement data to remove noise components.

[0030] Both for the determination of an estimator for the noise behavior and for the further processing of the individual measurements, it has proven advantageous to correct the measured values ​​of the individual measurements with regard to systematic errors due to a row offset and / or a column offset and / or a total offset.

[0031] In one embodiment, it is provided that the row offsets and / or column offsets and / or the total offset when carrying out the intended measurements and / or the reference measurement are each determined for this individual measurement on the basis of a spatial expected value of the measured values ​​of the measuring elements of the respective column, the respective row or the array of an individual measurement, which are not exposed to radiation from the measured scene or the measured object (even when the shutter is open), and are used to correct measured values ​​for the measuring elements that are exposed to radiation when the shutter is open.

[0032] In a measuring system, it is therefore provided that a noise reduction device is designed to correct systematic errors due to a row offset and / or a column offset and / or a total offset, wherein the correction is carried out according to the equation S^ijtn:=S^ijt−Em[S^mjt]−En[S^int]+Emn[S^mnt] is carried out, where n and m are the column and row indices for measuring elements which are taken into account for estimating the row and / or column and / or total offset, where, when correcting the measured values ​​of the respective individual measurement, n and m indicate the measuring elements which are not exposed to radiation from the measured scene or the measured object (even when the shutter is open).

[0033] In the reference measurement, in which all individual measurements are recorded with the shutter closed, the row offsets and / or column offsets and / or the total offset can each be determined based on a spatial expected value of the measured values ​​of the measuring elements of the respective column, the respective row or the array of an individual measurement for this individual measurement and used for correction.

[0034] In one embodiment of the method, it is provided that a root of the spatial variance of the difference measurement formed from the individual measurements, weighted in time with respect to the shutter sequence function defining the measurement sequence, is calculated as a measure of the noise.

[0035] Thus, in one embodiment, a root of the spatial variance of a difference result of the measurement is used as a measure of the noise of a measurement, the difference result being a time-averaged sum of the individual measurements, the measured values ​​of the individual measurements recorded with the shutter closed being included in the sum with an opposite sign, and the time averaging being carried out via a weighting of the number of measurements recorded with the shutter open and the number of individual measurements recorded with the shutter closed contained in the measurement according to the shutter sequence function defining the measurement sequence.

[0036] The following function is preferably used to calculate the noise measure: σΔ≈∑t*=1T∫seq0(t*)γ¯(0)+2∑t*=1T−1∑u*=t*+1T∫seq1(t*,u*)γ¯(u*−t*), where γ¯(Δt):=1MN∑ijtγij(Δt) and γij(Δt):=1T−Δt−1∑t*=1T−Δt(S^ijt*''−S^¯ij'')(S^ijt*+Δt''−S^¯ij'') with S^¯''=1T∑t*S^ijt*'' and S^ijt'':=S^ijt−Ei[S^ijt]−Ej[S^ijt]+Eij[S^ijt], where Ŝ ijt is a signal value of a measuring element of the i-th column and the j-th row of the individual measurement at time t of the reference measurement, where seq t specifies the value of a shutter sequence function, where: seqt{0: shutter closed at time t1: shutter opened at time t and seqt− is defined as follows; seqt−=1−seqt, where the shutter sequence function represents the position of the shutter during a measurement sequence, T the total number of individual measurements, T D the number of individual measurements with the shutter open and T S specify the number of individual measurements with the shutter closed, and where: ƒseq0(t*):=(seqt*(TS)2+seqt*−(TD)2) ƒseq1(t*,u*):=(seqt*⋅sequ*(TS)2+seqt*−sequ*−(TD)2−seqt*⋅sequ*−+seqt*−⋅sequ*TSTD)

[0037] In a measuring system, it is therefore provided that the sequence selection device and the noise characteristic prediction device are designed, if necessary together with the noise reduction device, to determine the measure of the predicted noise according to the function defined therein.

[0038] A measurement is understood here as the acquisition of a large number of individual measurements that are carried out according to a measurement sequence within a total measurement time. The total measurement time is referred to as dwell time. The individual measurements are usually acquired at a constant rate of individual measurements per time. The actual acquisition time is generally the same for all individual measurements and all measurement elements. In an imaging system, the term individual imaging is often used instead of individual measurement, and the term imaging instead of the term measurement. The prefix "individual" is not always used consistently, but the distinction can be clearly deduced from the context.

[0039] It has proven essential for noise suppression to determine an estimation function that specifies the noise behavior of the IR measuring device. The following explains how such an example estimation function can be determined. First, it is assumed that a total measurement time for capturing a (total) image, or a measurement duration, which is referred to as dwell time, is divided into equal time periods, and that an individual measurement is recorded with the measuring device in each time period. If the measuring device is an imaging measuring device, then such an individual measurement represents an individual image. In each time period of the total measurement duration (dwell time), a measured value is recorded for each imaging point, which is also referred to as a pixel in the following. This means that the measured value recording with the measuring device is carried out according to a predetermined acquisition frequency.Typical total measurement times for bolometric detectors in satellites for planetary surface observations are between 100 ms and 800 ms, depending on the satellite's altitude above the planet's surface. Within such a total measurement time t. dwell is the time dependence of the measured single image or signal intensity of a pixel S̃ ijt given as the sum of several contributions and can be represented as follows: S˜ijt=sij0+SijH+SijSceneseqt+SijSchutterseqt−+δ˜ijt=Sijt+δ˜ijt, where (ijt) are the indices for column, row and time, sij0 a constant offset, SijH, SijScene and SijSchutter and the signal components of the housing of the observed object and the shutter, assumed to be noise-free, are and S ijt indicates the noise-free, actual signal and δ̃ ijt summarizes all noise components of the measured signal. The factors seqt and seqt− each represent a shutter sequence function, where the shutter aperture is also called shutter, and are defined as follows: seqt:={0: shutter closed at time t1: shutter opened at time t seqt−=1−seqt

[0040] A measurement sequence for an image can thus be regarded as a set of recorded or acquired individual images that are acquired at a constant acquisition frequency or acquisition rate f p (called frame rate) within a single total measurement time t dwell The number of individual images acquired during the total measurement time t dwell recorded, results in T=TS+TD=ƒptdwell=∑t*=1T(seqt*+seqt*−), where T S and T Dthe number of images acquired with the shutter open and the shutter closed in the measurement sequence. A differential image, ie differential measurement signals S˜ijΔ for which temporal averaging is already carried out, is defined as follows: S˜ijΔ:=∑t*=1T(seqt*TS−seqt*−TD)S˜ijt*.

[0041] If we set the measured signals S̃ ijt * the signal model of equation (1) in the definition of the difference mapping according to equation (5), the result is: S˜ijΔ=SijScene−SijSchutter︸SijΔ+∑t*=1T(seqt*TS−seqt*−TD)δ˜ijt*=SijΔ+δ˜ijΔ.

[0042] Equation (6) represents an estimator for the measured difference signal free of systematic errors (unbiased estimator) if the noise components δ ̃ ijt* each have a constant and finite temporal expectation value. In this case, the temporal expectation value of the differences of the noise components δ˜ijΔ zero, which means that the temporal expectation value of the measured difference signals S˜ijΔ corresponds to the actual difference signal. Mathematically expressed: Et[δ˜ijt]=const.ij<∞⇒Et[δ˜ijΔ]=0⇔Et[S˜ijΔ]=SijΔ.

[0043] The estimates S˜ijΔ the actual difference signals SijΔ represent the input values ​​for the calibration calculation for a measuring device, whereby the differential measurement signals S˜ijΔ can be converted into a difference radiance. If the emissivity and temperature of the shutter are known, the radiance of the observed object or scene can be calculated from the difference radiance.

[0044] In practice, the measured difference noise quantity δ˜ijΔ be different from zero in each difference image and provide noise components for the calibrated images. It is therefore necessary to calculate the remaining noise δijΔ to measure, predict and evaluate the closure sequence function seq t , to minimize the limited number of images T available in a total measurement time, and a maximum shutter speed or switching time of the shutter, etc.

[0045] One way to derive a measure of the noise, i.e. the noise intensity, is to use a data set of a reference measurement in which, in a total measurement time t dwell Individual images are captured using a corresponding shutter sequence according to a shutter sequence function, but in the actual measurement (regardless of the value of seq t ) the shutter always remains closed. Quantities related to the reference measurement are marked with a "^". According to the signal model in equation (1) and the definition of the differential image in equation (5), the following equation results from equation (6): S˜ijΔ=SijShutter−SijSchutter︸=0+∑t*=1T(seqt*TS−seqt*−TD)δ^ijt*=δ^ijΔ.

[0046] The difference mapping S^ijΔ, which is calculated from the individual images of the reference measurement, which are recorded with the shutter constantly closed, thus only includes the noise quantity δ^ijΔ. However, this still depends on the measurement sequence under consideration. For such a measurement, the thermal state of the measuring device must be kept constant, which must also apply to other measurements, so that the prerequisite of equation (7) is met, namely that the temporal expectation value of the noise quantities is constant and finite. A noise intensity σ Δ the noise quantities δ^ijΔ can be estimated using their variance or standard deviation. The reference data for the estimation are taken directly from the difference image, which is calculated for a measurement according to equation (8) using a shutter sequence function seq t is calculated. The noise intensity σ Δis estimated as σ Δ : σΔ≈σ^Δ=1MN−1∑i*=1M∑j*=1N(S^i*j*Δ−S^¯Δ)2 with S^¯Δ:=1MN∑i*=1M∑j*=1NS^i*j*Δ, where M is the number of columns and N is the number of rows in the difference mapping matrix S^ijΔ Since a constant thermal environment during the individual measurements and a constant temporal expectation value for the individual noise values ​​δ ̂ ijt .(see equation (7)), the spatial expectation value of the resulting difference mapping matrix, which is described in equation (8), should be zero. It is therefore useful to calculate the spatial mean S^¯Δ the noise quantity equal to zero, ie an expected value E[S^¯Δ] equal to zero, so that equation (9) is transformed to: σΔ≈σ^Δ=1MN−1∑i*=1M∑j*=1N(S^i*j*Δ)2 with E[S^¯Δ]=0.

[0047] By using the estimator σ̂ Δused, any deviation from the stability assumptions during the acquisition of a measurement of an image (ie a measurement sequence of individual images) is reflected in an increase in the noise level. From equation (10) it can be seen that the noise level σ Δis estimated from a sample set that includes all imaging matrix elements in the spatial imaging direction, i.e. the individual reference values ​​are taken from a single difference image. In practice, it has been shown that the spatial and temporal noise values ​​appear to be almost the same if row, column, and total offset structures are removed before or after calculating the difference image. It has been shown that in measuring devices, individual rows and columns are affected differently in terms of noise, presumably due to the acquisition electronics used. This means that the measurement results for a row and / or a column, for example, all have an increased noise value, but one that is almost uniform for the respective column or row.

[0048] The procedure just described for determining a noise figure σ̂ Δfor the imaging noise intensity can only be applied to measurements that do not contain a measurement signal from an object to be observed, ie, are recorded with the shutter closed or are all recorded with the shutter open, without any thermal radiators in the field of view of the detection device that emit a temporally or spatially fluctuating temperature signal. This means that a stationary noise process must be present. Furthermore, it has been shown that the noise intensities σ determined according to this method Δ , which are each derived from individual difference images calculated for a measurement sequence, exhibit relatively strong fluctuations. Therefore, it is necessary to perform multiple measurements for a measurement sequence and average them.

[0049] Another alternative method for determining a measure of the noise signal will be explained below and uses the data from a single measurement. This estimation is based on knowledge of the spatio-temporal autocovariance function, which serves as a compact representation of the noise characteristics or noise behavior of the measurement system and is estimated from a simple measurement of an image data cube from acquired individual images of a reference measurement, as defined in equation (8). This means that for this second approach, i.e. the second method for estimating the noise intensity described here, a measurement sequence with the shutter closed must also be acquired. The one reference measurement essentially represents the data material for any number of measurements for different shutter sequences.

[0050] In a noise prediction calculation according to this approach, only the autocovariance function and the shutter sequence function are used as input data. The acquired individual image data sets are first freed of correlated noise components before further processing. These modifications impart stationary properties to the individual data sets, making an estimate of the autocovariance function more stable. The modification procedures used to estimate the autocovariance function are later also applied to the actual measurement data, where noise reduction is performed according to the method proposed here.

[0051] It is assumed that the individual signal values ​​recorded by the matrix-arranged measuring elements exhibit a correlation with respect to the row and / or column in which they are arranged. Looking at actual measurement results from measuring devices, it becomes clear that this assumption is justified. In figures, the noise values ​​of individual measuring elements in a row or the individual measuring elements in a column are often correlated with each other. It is assumed that an expected value of the noise values ​​δ ̂ ijt of a row j at time t has a constant value across all columns. This can be expressed as: Ei[δ^ijt]=δ^jt.

[0052] This means that for the noise signals related to a line, a constant (line) expected value E i [δ ̂ ijt ] exists which has the constant value δ ̂ jtIf we consider the correlation of the noise values ​​of a column, we can see that the column expectation value E j [δ ̂ ijt ] of column i has a constant value δ ̂ it Expressed in formulaic terms, this means: Ej[δ^ijt]=δ^it.

[0053] Note that t indicates the time index. If the noise value is written as a modified noise value δ' ijt , where the individual values ​​are free from the respective line offset, ie a deviation from the line offset is specified, the following equation is obtained: δ^ijt':=δ^ijt−Ei[δ^ijt]⇒Ei[δ^ijt']=0.

[0054] From this definition together with the assumption that the expected value of the noise values ​​along a line is locally constant, it follows that the expected value of the modified noise value δ' ijt along a column at a time is always zero. A modified image data set, ie the modified signal values ​​Ŝ' ijt , which takes the line offsets into account, can be defined as: S^ijt=Sijt+δ^ijt'+Ei[δ^ijt]S^ijt':=S^ijt−Ei[S^ijt]=Sijt−Ei[Sijt]+δ^ijt'.

[0055] The ijt represents the measured reference image data, ie, signal values, the expected value for the signal of a row can be estimated using a suitable estimation function across all signal values ​​of this row measured for the different columns. In practice, an averaging is performed. If one takes the modified image data set, ie, the modified signal values ​​Ŝ'ijt , the column offsets can be taken into account analogously to the procedure according to equation (14). This results in another modified data set, ie, further modified signal values ​​Ŝ'' ijt : S^ijt'':=S^ijt'−Ej[S^ijt']=Sijt+Eij[Sijt]−Ej[Sijt]−Ei[Sijt]+δ^ijt''.

[0056] In order to estimate the column and row offsets caused by noise and avoid erroneously including signal components of an actual measurement signal, the measuring device can be operated in actual measurement mode such that some of the measuring elements are not exposed to the actual signal of the observed scene, i.e., the object. This measurement data then only measures a noise signal, which can be used to estimate the expected values ​​required according to equation (15). If the measuring elements of such a reference area of ​​the matrix are addressed via the column and row indices m, n, and those measuring elements exposed to the actual scene signal are addressed via the indices i, j, the further modified signal can be written according to the following equation: S^ijt'':=S^ijt−Em[S^mjt]−En[S^int]+Emn[S^mnt].

[0057] The fact that this approach produces meaningful results has been experimentally confirmed. The determined row and column offsets, which are determined based on the measurement elements in the reference area, can thus be transferred to the remaining row and column signals.

[0058] If the further modified measurement signal Ŝ'' is used ijt according to equation (15) or in the specific measurement case later equation (16), it turns out that a difference signal according to equation (8) takes the following form: S^ijΔ''=δ^ijΔ'', where the difference noise quantity δ^ijΔ'' in addition to the properties given in equation (7), has the following properties: Ei[δ^ijt'']=Ej[δ^ijt'']=Eijt[δ^ijt'']=0⇒Ei[δ^ijΔ'']=Ej[δ^ijΔ'']=Eijt[δ^ijΔ'']=0.

[0059] The expected values ​​of the individual noise quantities as well as the differences between them each have an expected value of zero in the imaging directions as well as in the time and cube directions. To determine the noise intensity of a difference image, the spatial variance of the difference noise quantities δ^ijΔ'' analyzed. The individual calculations are only described in outline here, as the complete calculation is quite extensive. Nevertheless, the calculation can be understood using general algebraic calculation rules according to the principles shown here. Assuming that a series of measurements (measurement sequence) is recorded under the conditions indicated in equation (8) and the individual measurement signals are modified according to equation (16), a spatial variance for a sceneless difference image is defined as: Varij(S^ijΔ'')=Varij(δ^ijΔ''):=Eij[(δ^ijΔ''−Eij[δ^ijΔ''])2].

[0060] According to equation (18) the expected value Eij[δ^ijΔ''] equal to zero, so that some terms vanish when multiplying out the brackets and the variance can be written as: Varij(δ^ijΔ'')=Eij[(δ^ijΔ'')2]=Eij[(∑t*=1T(seqt*TS−seqt*−TD)δ^ijt*'')(∑u*=1T(sequ*TS−sequ*−TD)δ^iju*'')].

[0061] In equation (20), the product of the sums can be considered as sums of elements of symmetric matrices. The total sum can be represented as consisting of two summation series, where one summation includes the diagonal elements of the matrices and the other the two remaining triangular submatrix elements. Together with the introduction of a covariance function with the definition Covij(δ^ijt*'',δ^iju*''):=Eij[(δ^ijt*''−Eij[δ^ijt*''])(δ^iju*''−Eij[δ^iju*''])]=Eij[δ^ijt*'',δ^iju*''] the spatial variance of a difference map can then be written as: Varij(S^ijΔ'')∑t*=1Tƒseq0(t*)Covij(δ^ijt*'',δ^ijt*'')+2∑t*=1T−1∑u*=t*+1Tƒseq1(t*, u*)Covij(δ^ijt*'',δ^iju*'') with ƒseq0(t*):=(seqt*(TS)2+seqt*−(TD)2)ƒseq1(t*,u*):=(seqt*⋅sequ*(TS)2+seqt*−sequ*−(TD)2−seqt*⋅sequ*−+seqt*−⋅sequ*TSTD).

[0062] The factors ƒseq0(t*) and ƒseq1(t*, u*) are values ​​that are determined exclusively by the shutter sequence function seq t , which is the basis of the measurement series. The variance of a difference image is therefore only dependent on these factors dependent on the shutter sequence function and the spatial-temporal covariance function, as defined in equation (21). If the spatial variance is known, the spatial noise intensity of a difference image can be calculated according to: σΔ=Varij(δ^ijΔ'').

[0063] In order to determine the spatial noise intensity, it is therefore necessary to obtain an estimator for the covariance function defined in equation (21). For this purpose, further assumptions are made. An element of the mapping, which is addressed by the indices i, j, is regarded as a random variable that has T temporal outcomes. In this way, the definition of equation (21) can be regarded as an autocovariance definition. In addition, the values ​​of the random variables should reflect the outcome of a stationary process, which implies that the temporal expectation of the noise quantity δ''ijt constant over time, but individual for each individual mapping element (i, j): Et[δ^ijt'']=δ^ijt''.

[0064] Furthermore, the autocovariance of the noise is assumed to be time-independent for each time interval Δt = u* - t*. This then yields: Covij(δ^ijt*'',δ^iju*'')=Eij[δ^ijt*''δ^iju*'']=Eij[δ^ij(t*+ν)''δ^ij(u*+ν)'']∀ν∈ℕ, where time invariance is exploited in the second transformation. According to the definition of Δt given above, this can be rewritten as: Eij[δ^ijt*''δ^iju*'']=Eij[δ^ijt''δ^ij(t*+Δt)'']≈Eij[γij(Δt)]≈γ¯(Δt), where γ ij (Δt) an estimator for (δ^ijt'' δij(t*+Δt)'') and γ(Δt) represents the estimator for the spatio-temporal autocovariance to be determined.

[0065] To make an estimate, it is useful to acquire an imaging cube with a large temporal extent T, since then (T - Δt) temporal event or reference measurement pairs with a time difference Δt are available to estimate y ij (Δt) provided the reference image dataset is sufficiently stationary (see equation (25)). A suitable temporal estimator based on the dataset Ŝ'' ijtcan be applied can then be defined as follows: γij(Δt):=1T−Δt−1∑t*=1T−Δt(S^ijt*''−S^¯ij'')(S^ijt*+Δt''−S^¯ij'')with S^¯ij''=1T∑t*S^ijt*''.

[0066] The spatial estimator according to equation (21) is then finally given by γ¯(Δt):=1MN∑ijγij(Δt), which represents the autocovariance function of the measuring device or system. According to equations (22), (23) and (24), the difference imaging noise strength σ Δ then approximated: σΔ≈∑t*=1Tƒseq0(t*) γ¯(0)+2∑t*=1T−1∑u*=t*+1Tƒseq1(t*,u*)γ¯(u*−t*) .

[0067] The measure of the noise intensity of a difference image given in equation (30) thus depends on the autocovariance function and the shutter sequence function. The autocovariance function can be determined from the measurement results (individual images) of a reference measurement data set acquired with the shutter constantly closed.

[0068] Individual aspects of the invention are explained in more detail below with reference to the figures. Fig. 1 the spatial mean values ​​of a large number of individual measurements without and with correction for row and column offset; Fig. 2 Representations of a measured and predicted spatio-temporal autocovariance function for real measurements and for simulated noise; Fig. 3 measured and predicted noise levels for measurement results recorded at different times, a comparison of the predicted noise levels and for simulated noise, each for a parameterized measurement sequence; Fig. 4 measured and predicted noise levels for measurement results recorded at different times and for simulated noise for two additional parameterized measurement sequences; Fig. 5 a schematic representation of a measuring system; and Fig. 6 a schematic representation of another measuring system.

[0069] The method described above was validated using a modified commercial ULIS UL 02 05 1 microbolometer detector array from ULIS, based in Veurey-Voroize, France. During all measurements, the array was maintained at a constant temperature of 30°C (+ / - 10 mK) using a thermoelectric cooling system. The same thermoelectric cooling system will be used on a future spacecraft during a Mercury observation mission. All image acquisitions were performed under thermally constant ambient conditions and with a shutter in the closed position according to equation (8). The entire image measurement series, which is used to derive the autocovariance functions (equation (29)), comprises 450 images, each comprising 160x120 pixels (i.e., image elements). The individual measurements or individual images result in a data cube.The data cube was acquired at a repetition rate of 20 individual images per second. The integration time for one measurement is set to the default value of 120 µs. The noise and autocovariance estimates were performed based on result sets of rectangular 122x100 pixel regions of the individual images, or, in other words, on a sub-image cube with dimensions of 122x100x450. The rectangular region has a fixed position in all images or the entire cube considered and represents a scene area "exposed" by the optics for a planned Mercury mission, and in particular, for a Mercury Radiometer and Thermal Infrared Spectrometer (MERTIS). The rest of the image or data cube is considered as a reference region and used to estimate the row / column / image offsets, as described above.A calculation was performed for both a measured image data cube and a simulated image data cube containing Gaussian noise. The noise intensity was estimated according to the measures from equations (10) and (30) and compared for different measurement sequences. Before calculating the values ​​of γ, ij (Δt) and γ̅(Δt) are each “removed” from the image data cubes according to equation (16).

[0070] The spatial averages of 450 individual images taken at a speed of 20 images per second at an interval of 22.5 s are shown in Fig. 1. The upper graph shows the means of the raw data, while the lower graph shows the means after row, column, and image offset removal. All offsets were estimated using a reference area outside the scene domain, with the offset removal applied to the entire image. Combined with spatial averaging, the modified dataset appears to exhibit more temporally stationary behavior and is therefore used to estimate the autocovariance function γ̅(Δt).

[0071] In Fig. Figure 2 shows an example of the estimated autocovariance function as shown in equation (29). Three examples are shown.

[0072] On the left side of the Fig. Figure 2 shows an estimate for γ̅(Δt), the spatio-temporal autocovariance function (cf. equation (29)). The individual images were acquired at a rate of 20 images per second using a version of a commercial ULIS UL 02 05 1 microbolometer detector array adapted for the MERTIS mission. The base image data set was acquired under constant thermal conditions with a fully closed shutter and was freed of offset noise structures from the rows, columns, and the entire image. The left-hand panel shows γ̅(Δt) for the results of two measurements acquired 15 months apart with the same measuring device. The two estimates of γ̅(Δt) are in very good agreement, indicating that the noise behavior of the system has not changed significantly over the given time interval. In addition, a small systematic error (SAR) is present.A bias is evident, which causes the functions to fall below zero over a longer time interval, implying a negative correlation between the temporal measurement results. It is assumed that the systematic error due to the inclusion of temporal expected value estimates, which consist of temporally correlated measurement results, is introduced into the calculation of γ. ij (Δt). Furthermore, the slow decay of γ̅(Δt) means that the system has slowly changing noise components or slow fluctuation components.

[0073] On the right side of the Fig. Figure 2 shows the result of the same calculation of γ̅(Δt) based on a dataset containing only simulated uncorrelated noise values ​​of a Gaussian distribution function with a defined mean of 243 DN (digital number unit) and a standard deviation of 2.5 DN. As expected, the function shows a variance of (2.5 DN) only at t = 0. 2 and disappears for the remaining time intervals. A systematic error cannot be detected. The autocovariance function γ̅(Δt) given above thus appears to adequately describe the noise behavior, ie, the noise characteristics of the measuring device or system.

[0074] The information related to Fig. The autocovariance function γ̅(Δt) shown in Figure 2 represents input information for the shutter sequence optimization, which is described below.

[0075] First, a simple sequence or shutter sequence function is considered. In such a sequence, a single image is captured with a closed shutter and a single image with an open shutter. The time delay Δt = s between the two individual images is variable: seqts:={0: t=t01: t=t0+s with s>0.

[0076] In this definition of the shutter sequence function, the original shutter sequence function definition of equation (2) is extended by a parameter s, so that the shutter sequence function seqts actually represents a class of sequences, where s is the sequence index and indicates a single shutter sequence function within the class. The index t still points to a single image within a specific measurement history, i.e., a measurement sequence. Zero indicates the start time t0 at which the first image is acquired. For each shutter sequence function seqts The standard deviations are estimated for both the estimator of equation (30) and that of equation (10). The results are plotted graphically as a function of the sequence index s in Fig. 3 for both MERTIS measurements, which were recorded 15 months apart, and for a simulated uncorrelated noise. The top left shows the measured noise levels and the noise levels estimated according to the determined autocovariance function against the sequence index for a first series of measurements, and the top right shows the same for the series of measurements recorded 15 months later. It can be seen that with increasing time delay Δt, the noise increases slightly. This is due to the slow decay of the corresponding autocovariance function. In analogy to the estimated function of γ̅(Δt) in Fig. 2 show the estimates of the noise intensity σ Δ Good agreement is observed between all MERTIS measurements and both noise measurements. In the case of the simulated dataset, which has uncorrelated noise components, no time dependence is evident, as expected.

[0077] In Fig. 4 is the noise intensity σ for two further different measurement sequences Δ against the sequence index, shown on the left for real measurements with a measuring system as planned for the MERTIS mission, and on the right for a simulation with uncorrelated Gaussian noise. The measured and predicted noise levels are shown. For the measurement sequence shown in the lower part of Fig. 4, the number of scene images is fixed to 4.

[0078] Here are the results for a class of sequences defined as follows: seqts:={0: t0≤t≤t0+s ∨ t0+s+5≤t≤t0+2s+51: t0+s+1≤t≤t0+s+4 with s≥0, where four individual images of the scene under investigation, i.e., the observed object, are each surrounded by 2s individual images with the shutter closed to obtain a time-interpolated estimate. Such a sequence could occur if, for example, only a very short total measurement time is available for an image, in orbit, and an attempt is made to improve the signal-to-noise ratio by increasing the number of measurements with the shutter closed by the current scene subsequence. Individual images acquired with the shutter closed can be used for noise reduction for the preceding scene images as well as for the subsequently acquired scene images.

[0079] Out of Fig. 4, bottom left, shows that for this sequence class, a noise minimum for real measurement results exists, i.e., an optimal number of measurements with the shutter closed exists around the four currently acquired scene measurements. Extending the measurement series or sequence with additional individual images with the shutter closed would not lead to an improvement, but rather to a deterioration, in contrast to the case of uncorrelated noise (cf. Fig. 4 bottom right), for which the noise level could in principle be further reduced. Below right are Fig. 4 shows the calculated and estimated noise values ​​for uncorrelated Gaussian noise.

[0080] By comparing the autocovariance functions for the actual measurements performed with the aforementioned real-world measurement setup and the case of simulated uncorrelated noise, it can be concluded that the noise minimum can only exist due to the temporal correlation of noise components. For this example, the optimal sequence would be as follows: 18 images with the shutter closed, 4 scene images with the shutter open, and 18 images with the shutter closed, at an image acquisition rate of 20 images per second. The two real measurements, taken 15 months apart, show virtually identical results. Comparing the worst-performing sequence (s=0) with the optimal sequence, a relative improvement in the signal-to-noise ratio (SNR) by a factor of approximately 1.5 is possible.

[0081] The results for another example are shown in the upper part of Fig. 4. Here, a long total measurement time (dwell time interval) of approximately 800 ms is assumed. At an acquisition rate of 20 images per second, a constant number of 16 individual images could be acquired. As a further boundary condition, it is assumed that the shutter should only be switched once during the total measurement time for an image. In addition, time interpolation should be possible, meaning that a series of individual images acquired with the shutter closed can be inserted into the series of images acquired in the next total measurement time interval. A sequence class representing this situation can be described as follows: seqts:={0: t0≤t≤t0+s ∨ t0+T≤t≤t0+s+T1: t0+s+1≤t≤t0+T−1with s∈{0...T−2} and T=16.

[0082] The question of how many images with a closed shutter and how many images with the open shutter should be acquired in a total measurement time interval in order to obtain a minimum noise level can be answered by Fig. 4 (above). On the left, the calculated and estimated noise levels after a measurement are shown, and on the right, the calculated and estimated noise levels for simulated uncorrelated Gaussian noise. As with the previous examples, a noise minimum can be seen. In contrast to the previous examples, the measurement series for the uncorrelated noise also shows a noise minimum and the same optimal sequence, which consists of 6 images with the shutter closed, 9 images with the shutter open (scene images), and 6 images with the shutter closed. The relative improvement in the signal-to-noise ratio is a factor of approximately 2.1 for the measurement device used and a factor of 2.3 for uncorrelated noise.

[0083] It thus follows that, depending on boundary conditions such as the length of the total measurement time, the speed and / or frequency with which the shutter can be opened, and other factors, an optimum can be determined for each of the resulting measurement sequence classes in order to achieve noise reduction. Once such an optimal sequence or measurement procedure has been identified, real measurements are subsequently carried out, for example during planetary observation. The actually received or recorded measurement data are preprocessed in the same way to achieve noise suppression as was done to determine the autocovariance function. This means that the measurement data are processed analogously to the procedure described according to equations (11) to (16) as mentioned above.

[0084] The advantage of the method described here is that the noise behavior of the measurement system, similar to that planned for the MERTIS mission—that is, an imaging infrared measurement system—can be represented by a simple one-dimensional autocovariance function. This function can be derived from an image cube, i.e., a data set of consecutively acquired individual images under known and constant conditions.

[0085] The measurement sequence optimization method can be applied to a wide variety of sequence classes. Since many imaging systems suffer from slowly changing noise signal components, the method described here is helpful for finding an efficient operating mode. This avoids noise components that occur when using an incorrect operating mode. A further advantage is that the estimates can be calculated very quickly, so that this method can also be carried out on board a spacecraft with such a measurement system. Furthermore, the method has the advantage that, through appropriate processing of the measurement data, these steady-state conditions can be approximated. The data preprocessing performed to find the autocovariance function, which reduces the measurement data to a steady-state case, is also performed on the real measurement data acquired later.For different measurement series or sequences, an improvement in the signal-to-noise ratio of 1.5 to 2.1 has been demonstrated. However, depending on the specific boundary conditions, many other classes of measurement sequences are possible for which an optimum can be determined using this method.

[0086] In Fig. Figure 5 schematically describes a measuring system 100. Its core is a measuring device 101. The measuring device 101 comprises a plurality of measuring elements 102 arranged in a matrix-like array 103. The measuring elements 102 are arranged in columns 104, designated by the index i or n, and rows 105, which are linked to the indices j and m. The individual measuring elements 102 are, for example, microbolometers. However, they can also be other measuring detectors that detect a radiation signal that is at least also thermally influenced.

[0087] A shutter 110 is arranged in front of the measuring device 101 or integrated therein. In an open position, this shutter exposes at least some of the measuring elements 102, so that electromagnetic radiation from an object 120 or a scene to be observed can reach the corresponding measuring elements 102a through the open shutter 110 and expose them to radiation from the object or scene. Preferably, there are several measuring elements 102b onto which no radiation from the scene or object 120 impinges, even when the shutter 110 is in the open position. In the closed position of the shutter 110, no radiation from the observed object 120, i.e., the scene, impinges on any of the measuring elements 102.

[0088] The shutter 110, which is also referred to as a shutter in English, is controlled by a sequence control device 130 in order to be able to switch it between the open state and the closed state. This occurs in a time-coordinated manner with the acquisition of individual images by the measuring device 101. Therefore, this is generally also controlled by the sequence control device 130. The measuring device 101 is generally designed such that it acquires (individual) images at regular time intervals, i.e., reads out a acquired radiation value from each measuring element 102. A acquired (individual) image thus consists of a set of measured values ​​that can be assigned to the individual measuring elements 102. The individual (individual) images are thus acquired at a acquisition frequency, for example 20 images per second.The described measurement system 100 for noise reduction comprises a noise characteristics predictor 140, to which the (individual) image data of a set of individual images of a measurement series or measurement sequence are fed. In one embodiment, it can be provided that the acquired (raw) image data are fed directly to the noise characteristics predictor 140. In another embodiment, the image data are first fed to a noise reduction device, which largely removes systematic errors caused by readout electronics 106 or other interference components from the image data. Method steps are carried out as explained above in connection with the derivation of equation (16). In this second embodiment, the noise-reduced image data are fed to the noise characteristics predictor 140.In the first embodiment, these processing steps for eliminating systematic offset errors are performed in the noise characteristic prediction device 140 itself. This device is further configured to estimate the autocovariance function γ̅(Δt), as listed in equation (29), based on the image data corrected for systematic errors. The result of the estimation is fed to a sequence selector 160.

[0089] Additionally, boundary conditions 170 are provided to the sequence selector 160 as input. These can, for example, be received from a ground station via an interface 171 in a measurement system 100 integrated into a satellite or a measuring probe. Based on the boundary conditions, the sequence selector 160 calculates the noise intensity for different sequences or sequence classes according to or analogously to equation (30). This determines an optimal measurement sequence with minimized noise. The measurement sequence thus determined is transmitted to the sequence control device 130, which then controls the shutter 110 and the measuring device 101, as already mentioned.

[0090] Under stable ambient conditions, the noise characteristic prediction device 140 only needs to be supplied with one set of (individual) image data sets from a measurement series in which the shutter is closed for all acquired (individual) images. From this single set of image data (measurement data set), also referred to as the image cube, the noise characteristic, also referred to as the noise behavior, of the measurement system 100 is determined, which can then be used to determine the optimal measurement sequence. Whenever the noise characteristic, which is embodied by the autocovariance function, is to be redetermined, such a reference or calibration data set is acquired under the control of the sequence control device. In some embodiments, an additional noise analyzer 190 is provided, which controls the acquisition of such a reference or calibration data set.

[0091] After an optimal measurement sequence for the specified boundary conditions has been determined, the sequence control device 130 controls the acquisition of image data sets for individual (overall) images. The data provided by the noise reduction device 150 is used as output signal 180 for further processing.

[0092] In Fig. 6 is another embodiment similar to that of Fig. 5. Identical technical features are provided with the same reference numerals and fulfill the same or at least very similar functions. The embodiment according to Fig. 6 differs in that no shutter 110 is arranged in front of the measuring device 101. In this respect, the sequence control device 130 in this embodiment only controls the measuring device 101. The same applies to the variant in which a noise analyzer is present, which, however, can also be completely omitted in this embodiment and whose function can be taken over by the sequence control device 130. In this embodiment, the measuring device is in any case designed such that even without the presence of the shutter, some of the detection elements 102 are not exposed to radiation from the object or the scene. These areas of the array 103 or the matrix of measuring elements 102 serve to estimate the noise characteristic, which is then transferred to the remaining measuring elements 102a exposed to radiation. Otherwise, this system is similar to Fig. 6 that of Fig. 5.

[0093] The sequence control device 130 and / or the noise characteristic prediction device 140 and / or the noise reduction device 150 and / or the sequence selection device 160 and / or the noise analyzer 190 can be implemented entirely or partially in dedicated electronic circuits or by means of program-controlled processors. The noise reduction device 150 and other dedicated circuits of the remaining devices can be implemented entirely or partially as a component of the evaluation electronics 106 or one of the other devices, or can be considered as components. The various devices can be implemented jointly in a control device. A subdivision made here is based on the functions of the individual devices for the purpose of explanation, rather than on implementation in hardware components.Parts can be outsourced to an external program-controlled computer or a computer that can be used for several different facilities and measurement systems. Particularly in a spacecraft, it can be advantageous to run parts of the calculation that are only performed infrequently, such as measurement sequence optimization, on hardware that can also perform other tasks. List of reference symbols 100 measuring system 101 Measuring device 102 measuring elements 102a Measuring elements exposed to radiation when the shutter is open 102b Measuring elements that are not exposed to radiation when the shutter is open 103 Array (Matrix) 104 column 105 row 106 Evaluation electronics 110 shutter aperture 120 Object (Scene) 130 Sequence control device 140 Noise characteristic prediction device 150 Noise reduction device 160 sequence selection device 170 boundary conditions 171 Interface 180 output signal 190 Noise Analyzer

Claims

[1] Method for minimizing noise during data acquisition with a measuring system (100) comprising the steps: Acquiring data of a reference measurement with the measuring system (100) under defined conditions, wherein the conditions include at least a total measurement time, a rate of individual measurements per time interval, an integration time per measurement, and a temperature of an array of measuring elements (102) of a measuring device (101) of the measuring system (100), wherein the data acquisition of the reference measurement comprises acquiring a plurality of individual measurements according to a reference measurement sequence; Deriving an estimator for a measure of the noise of the measuring system (100) based on measured values ​​of the reference measurement; Determining an optimized measurement sequence taking into account boundary conditions (170), which include at least one possible total measurement time and a rate of individual measurements per time interval, for intended measurements with a view to minimizing a measure of the noise, wherein the data acquisition is carried out with an imaging measuring system (100) with an array of infrared-sensitive measuring elements (102), in front of which a controllable shutter (110) is arranged such that, in a closed state of the shutter (110), none of the measuring elements (102) is exposed to radiation from a measured scene or a measured object (120), and in an open state, at least some of the measuring elements (102) are exposed to radiation from the measured scene or the measured object (120), wherein the measuring process is carried out via a A shutter sequence function is defined which, for a sequence of individual measurements acquired one after the other in time, defines which individual measurements are acquired with the shutter (110) open and which with the shutter (110) closed, wherein the estimation function for the measure of the noise is derived from a spatial-temporal autocovariance function of the measuring system (100), which is acquired on the basis of the reference measurement with the reference measurement sequence in which a sequence of individual measurements acquired one after the other in time with the shutter (110) closed. [2] Method according to claim 1, characterized bythat the shutter (110) is operated in such a way that, even when the shutter (110) is open, measuring elements (102) of the array are not exposed to radiation from the measured scene or the measured object (120), and row offsets and / or column offsets and / or a total offset when carrying out the intended measurements are each determined for this individual measurement on the basis of a spatial expected value of the measured values ​​of the measuring elements (102) of a respective column (104), a respective row or the array of an individual measurement which are not exposed to radiation from the measured scene or the measured object (120), and are used to correct measured values ​​for the measuring elements (102) which are exposed to radiation when the shutter (110) is open. [3] Method according to one of the preceding claims, characterized bythat the measure of the noise is calculated by a root of the spatial variance of a difference measurement formed from the individual measurements, which is time-weighted with respect to the shutter sequence function determining the measurement process. [4] Method according to one of the preceding claims, characterized by that the following function is used as a measure of the noise: σΔ≈∑t*=1Tfseq0(t*)γ¯(0)+2∑t*=1T−1∑u*=t*+1Tfseq1(t*,u*)γ¯(u*−t*), where t* and u* time points and a time difference by Δt = u* - t* is specified, where γ¯(Δt):=1MN∑ijγij(Δt), where M indicates a number of columns (104) and N a number of rows of the array of measuring elements (102) of the measuring system (100), and γij(Δt):=1T−Δt−1∑t*=1T−Δt(S^ijt*''−S^¯ij'')(Sijt*+Δt''−S^¯ij'') with S^¯ij''=1T∑t*S^ijt*'' and, S^ijt'':=S^ijt−Ei[S^ijt]−Ej[S^ijt]+Eij[S^ijt], where Ŝ ijt a signal value of a measuring element of the i-th column (104) and the j-th row of the individual measurement at time t of the reference measurement, and E i [Ŝ ijt ] is an expected value of a row j at time t considered over all columns (104), E j [Ŝ ijt ] is an expected value of a column i (104) at time t considered over all rows and E ij [Ŝ ijt ] is an expected value at time t considered over all rows and columns (104), where seq t and seqt− Factors of a shutter sequence function are: seqt:={0:Shutter closed at time t1:Shutter opened at time t and seqt−=1−seqt, wherein the shutter sequence function represents the position of the shutter (110) during a measurement sequence, T the total number of individual measurements, T Dthe number of individual measurements with the shutter open (110) and T S indicate the number of individual measurements with the shutter closed (110), and where: fseq0(t*):=(seqt*(TS)2+seqt*−(TD)2) fseq1(t*,u*):=(seqt*sequ*(TS)2+seqt*−sequ*−(TD)2−seqt*−sequ*−+seqt*−sequ*−TSTD). [5] Measuring system (100) with improved noise minimization comprising: a measuring device (101) with an array of infrared-sensitive measuring elements (102); a control device which controls the acquisition of a reference measurement under defined conditions according to a reference measurement sequence, wherein the acquisition of the reference measurement comprises the acquisition of a plurality of individual measurements according to the reference measurement sequence, wherein the conditions comprise at least a total measurement time, a rate of individual measurements per time interval, an integration time per measurement, and a temperature of the array (103) of measuring elements (102) of the measuring device (101) of the measuring system (100); a noise characteristic prediction device (140) which is designed to derive an estimation function for a measure of noise of the measuring system (100) based on measured values ​​of the reference measurement; a sequence selection device (160) which is designed to determine an optimized measurement sequence for intended measurements, taking into account boundary conditions (170), which include at least one possible total measurement time and a rate of individual measurements per time interval, for the intended measurements with a view to minimizing the measure of noise, wherein a controllable shutter (110) is arranged in front of the measuring elements (102) of the measuring device (101), so that in a closed state of the shutter (110) none of the measuring elements (102) is exposed to radiation from a measured scene or a measured object (120), and in an open state, at least some of the measuring elements (102) are exposed to radiation from the measured scene or the measured object (120), and the control device comprises a sequence control device which controls the opening and closing of the shutter (110) and the acquisition of measured values ​​by the measuring elements (102) of the measuring device (101), wherein the noise characteristic prediction device (140) is designed to derive the estimation function for the measure of the noise from a spatial-temporal autocovariance function of the measuring system (100), which is acquired on the basis of the reference measurement with the reference measurement sequence, in which a sequence of temporally successive individual measurements with a closed shutter diaphragm (110) is acquired, and wherein the sequence selection device (160) is designed to determine a measure of a predicted noise for different shutter diaphragm sequences or shutter diaphragm sequence functions on the basis of the estimation function for the measure of the noise of the measuring system (100) and Selecting a shutter sequence function for which minimal noise is predicted for the intended measurement under the given boundary conditions (170), wherein the measurement sequence of the intended measurement is defined via the shutter sequence function, which defines, for a sequence of individual measurements of the measurement sequence acquired one after the other in time, which individual measurements are acquired with the shutter (110) open and which with the shutter (110) closed. [6] Measuring system (100) according to claim 5, characterized by that the infrared-sensitive measuring elements (102) of the array of the measuring device (101) are arranged in columns (104) and rows. [7] Measuring system (100) according to claim 6, characterized bythat the shutter (110) is designed such that even when the shutter (110) is open, measuring elements (102) of the array are not exposed to radiation from the measured scene or the measured object (120) and a noise reduction device is designed to correct systematic errors due to a row offset and / or a column offset and / or a total offset, wherein the correction is carried out according to the equation S^ijt'':=S^ijt−Em[S^mjt]−En[S^int]+Emn[S^mnt] is executed, where Ŝ xyt indicates a measured value of a measuring element in the column (104) with an index x and a row with the index y at time t, E k [Ŝ kyt ] an expected value of the measured values ​​Ŝ kyt of a row y at time t considered over all columns (104), E l [Ŝ xlt ] is an expected value of a column (104) x at time t considered over all rows and E xy [Ŝ xyt] is an expected value at time t considered over all rows and columns (104), where n and m are the column (104) and row indices for measuring elements (102) which are taken into account for estimating the row and / or column and / or total offset, where, when correcting the measured values ​​of the respective individual measurement, n and m indicate the measuring elements (102) which are not exposed to radiation from the measured scene or the measured object (120), and i and j indicate measuring elements (102) which are exposed to radiation from the measured scene. [8] Measuring system according to one of the preceding claims, characterized by that the noise characteristic prediction device (140) is designed to estimate the measure of the noise σ Δ of the measuring system (100) using the following function: σΔ≈∑t*=1Tfseq0(t*)γ¯(0)+2∑t*=1T−1∑u*=t*+1Tfseq1(t*,u*)γ¯(u*−t*), where t* and u* time points and a time difference by Δt = u* - t* is specified, where γ¯(Δt):=1MN∑ijγij(Δt), where M indicates a number of columns (104) and N a number of rows of the measuring elements (102) of the measuring system, and γij(Δt):=1T−Δt−1∑t*=1T−Δt(S^ijt*''−S^¯ij'')(S^ijt*+Δt''−S^¯ij'') with S^¯ij''=1T∑t*S^ijt*'' and, S^ijt'':=S^ijt−Ei[S^ijt]−Ej[S^ijt]+Eij[S^ijt], where Ŝ ijt is a signal value of a measuring element of the i-th column (104) and the j-th row of the array of the individual measurement at time t of the reference measurement, and E i [Ŝ ijt ] is an expected value of a row j at time t considered over all columns (104), E j [Ŝ ijt ] is an expected value of a column i (104) at time t considered over all rows and E ij [Ŝ ijt] is an expected value at time t considered over all rows and columns (104), where seq t and seqt− Factors of a shutter sequence function are: seqt:={0: shutter closed at time t1: shutter opened at time t and seqt−=1−seqt, wherein the shutter sequence function represents the position of the shutter (110) during a measurement sequence, T the total number of individual measurements, T D the number of individual measurements with the shutter open (110) and T S indicate the number of individual measurements with the shutter closed (110), and where: fseq0(t*):=(seqt*(TS)2+seqt*−(TD)2) fseq1(t*,u*):=(seqt*sequ*(TS)2+seqt*−sequ*−(TD)2−seqt*sequ*−+seqt*−sequ*TSTD).

Images