DE102018104668B4 - Photoflash camera and method of operating such - Google Patents

Abstract

Method for operating a light transit time camera according to the phase measurement principle using at least two modulation frequencies, wherein for determining a distance value, a determined measured value is transformed into a base orthogonal to the space diagonal of a rauschnormierten vector space and a probable distance value is determined in this orthogonal basis.

Description

The invention relates to a method for operating a light transit time camera according to the phase measurement principle according to the preamble of the independent claims.

From the DE 10 2013 207 647 A1 For example, a method using several frequencies for determining a distance value within an extended unambiguity range is already known, in which the most probable range values are determined with the aid of spatial diagonals in a so-called modulo space. In this case, two frequency pairings are preferably always considered, with incorrect measurements being recognized, for example, when two consecutive measurements with different frequency pairings indicate different distance values.

From the DE 10 2016 213 217 A1 a time of flight camera system is known which determines distances from the difference of the phase shift of a transmitted and received light, wherein in a memory, a physical noise model of all light transit time pixels of the light transit time sensor is stored, with the aid of an evaluation unit depending on the deposited noise model limits and / or weightings of a temporal and / or local noise filters.

The US 2018/0014003 A1 shows a distance measuring system in which the performance of the system is tested by means of standardized test objects.

In particular, the limits of disparity are determined in a multi-camera system. To determine these limits, normalized intensity histograms are used and compared with a reference histogram. The local binary patterns are primarily sensitive to the noise. It is therefore proposed to match the noise of the images to the noise characteristics of the reference camera.

The object of the invention is to improve the accuracy and the measuring range of the distance determination.

The object is solved by the independent claims. Advantageous embodiments of the invention are specified in the subclaims.

Advantageously, a method for operating a light transit time camera according to the phase measurement principle using at least two modulation frequencies provided in the distance values and allowed space diagonals vectorially described and normalized using a noise term and converted into rauschnormiert distance values or in allowed, rauschnormierte room diagonals.

This approach has the advantage that all further calculations are simplified and improved in their accuracy by the isotropic properties of the space auschnomierten.

In a further embodiment, it is intended to determine with a most probable, cut-out distance value on the allowed, normalized space diagonal which comes closest to a measured, cut-down distance value, with a back transformation of the most probable, cut-out distance value to a most probable range value.

It is particularly useful to determine an area that is orthogonal to the extracted space diagonals and intersects the cut-out distance value, forming an intersection coordinate system from intersections of the space diagonals with the area, such that the intersections are represented by integer multiples of the unit vectors of the intersection coordinate system are described
Transforming the measured uncommitted distance value into the intercept coordinate system into a transformed range value, determining a most probable transformed range value, the closest intersection point to the transformed range value, backtransforming the most probable transformed range value to a most probable range value.

As a result, the calculations can be advantageously simplified, in which they are performed in a reduced by one dimension computing space.

In a further embodiment, it is provided to determine the most probable, transformed distance value, by first determining the closest intersection points by rounding up and down the vector factors of the transformed distance value, and determining the distance only to these closest intersection points,
wherein the intersection with the smallest distance to the transformed distance value represents the most probable transformed distance value.

It is also helpful to limit the distance measurements and the calculations of the distance values to a maximum distance value, wherein a most probable range value assignment occurs only for range values smaller than the maximum range value.

In this case, the maximum distance value is preferably smaller than the maximum uniqueness range, whereby the calculation effort for determining a distance value is again reduced by this procedure.

It is also useful if a confidence value is formed starting from the distance of the most probable distance value from the measured distance value.

On the basis of this trustworthy example, further decisions can be made in further processing steps.

In a further embodiment, it is provided that the assignment of at least one preceding or adjacent distance measurement is taken into account for the assignment of a most probable distance value.

In particular, it is advantageous to design a light cycle camera for carrying out one of the aforementioned methods.

The invention will be explained in more detail by means of embodiments with reference to the drawings.

• 1 Erlaubte, rauschfreie Messungen mit verschiedenen ModulationsFrequenzen,
• 2 Raumdiagonalen nach Normierung mit der Rausch-Stärke,
• 3 eine Ebene senkrecht zu den Raumdiagonalen, die einen Messpunkt s schneidet,
• 4 ein Gitter von Raumdiagonalen auf der senkrechten Ebene gemäß 3.

They show schematically:

• 1 Allowed, noise-free measurements with different modulation frequencies,
• 2 Room diagonals after normalization with the noise power,
• 3 a plane perpendicular to the spatial diagonal, which is a measuring point s cuts,
• 4 a grid of room diagonals on the vertical plane according to 3 ,

In the following description of the preferred embodiments, like reference characters designate like or similar components.

In the case of continuous wave (CW) light time measurements, the problem typically arises that the distance can only be measured modulo integer multiples of a uniqueness range UR: $$d_{\text{w}}=\overline{d}+n\cdot UR,$$

Here is d _w the true object distance, d the measured distance and n an (initially unknown) integer. The uniqueness range UR = c / (2 f) can be easily determined from the modulation frequency f.

wobei die unbekannten n_i ∈ ℕ „möglichst plausibel“ gewählt werden, d.h. so dass jede Messung in derselben Distanz d_w resultiert.This problem can be solved by measurements with several different modulation frequencies f ₁ , ..., f _D : $$\begin{array}{c}{d}_{\text{w}}={\overline{d}}_1+{\mathrm{<figure-callout\; id="1"\; label="n"\; filenames="DE102018104668B4\_0015.png"\; state="\{\{state\}\}">n</figure-callout>}}_1\cdot \mathrm{<figure-callout\; id="1"\; label="U\; R"\; filenames="DE102018104668B4\_0015.png"\; state="\{\{state\}\}">U\; </figure-callout>}{R}_{}{}_1\\ ⋮\\ d_{\text{w}}={\overline{d}}_D+n_D\cdot U{R}_D.\end{array}$$

where the unknown n _i ∈ ℕ are chosen as "as plausible as possible", ie such that each measurement is at the same distance d _w results.

To find the true distance d _w the following procedure is proposed according to the invention. In this case, it is helpful to represent the uniqueness regions UR ₁ ,..., UR _D as multiples of a basic unambiguity range UR _m , it is particularly preferred if the UR _{i can be} represented here as an integral multiple of UR _m . For example, the largest common divisor UR _m : = gcd (UR ₁ ,..., UR _D ) of the uniqueness ranges can be determined for this purpose.

Equations 1 can then be described vectorially as follows: $$d_{0\text{m}}=\left(\begin{array}{c}{\overline{d}}_1\\ {\overline{d}}_2\\ {\overline{d}}_3\end{array}\right)=t\cdot U{R}_{\text{m}}\left(\begin{array}{c}1\\ 1\\ 1\end{array}\right)+U{R}_{\text{m}}\left(\begin{array}{c}{m}_1\\ m_2\\ m_3\end{array}\right),$$

This equation 2 represents noiseless measurements with D = 3 frequencies in a 3-dimensional space. 1 shows a corresponding plot. A first section of the room diagonal d _{0, m} starts with an object distance d _w = 0, and thus in the origin 120 of the diagonal plot, and travels at small distances along the original diagonal. This can be described (parameterized by t) with an offset m = (m ₁ , m ₂ , m ₃ ) ^T = (0,0,0) ^T. This continues until, for the first time for a measurement i, a uniqueness range jump is achieved, first at the smallest of the three UR _i . In the example of 1 for i = 3. From there jumps d _i back to zero while getting the other components d _{j ≠ i} do not change. This results in a new diagonal with offset m _i = 0, while m _{j ≠ i} from the current values of d _j / UR _m .

With the new offset, Eq. (2) until the next uniqueness jump is reached and analogously a new offset m = (m ₁ , m ₂ , m ₃ ) ^T must be determined. This procedure can Continue until you reach the maximum object distance dmax that can be measured with this system, the smallest common multiple of the uniqueness ranges dmax = lcm (UR ₁ , ..., UR _D ). Here you meet again the original line with m = (0,0,0) ^T.

Since uniqueness jumps always occur at integer multiples of UR _m , the m _{i are} always integer. However, not all integer combinations may be allowed.

It should be noted that all diagonals have the spatial direction (1,1,1) ^T.

It is helpful to document when performing the above procedure which object distances d _w Uniqueness jumps happen, and these together with offset m = (m ₁ , ..., m _D ) ^T store. If one wishes to determine an object distance later from a straight line with a certain offset m, this information is advantageous. It should be noted here that the displacement of a straight line by UR _i in the i-direction to a solution according to Eq. 1 leads. Therefore, offsets m = (m ₁ , ..., m _D ) ^T + (k ₁ UR ₁ , ..., k ₁ UR ₁ ) ^T / UR _{m are} equivalent to integer k ₁ , ..., k _D. An algorithm proposed in the sense of the invention makes it possible to choose this straight line with k _i = ± 1, therefore it is also advantageously possible to store offsets for k _i = ± 1 for each offset found. Alternatively, the Chinese Remainder Theorem can be used to exclude (noise-free) measurements d ₁ , ..., d _D to determine the object distance.

angenommen werden, wobei t_exp,i und f_i jeweils Belichtungszeit und Modulationsfrequenz der i-ten Messungen sind. In nullter Näherung können die σ_i auch als identisch angenommen werden. Aber selbstverständlich können im Rauschterm σ_i auch weitere Effekte berücksichtigt werden.Gln. 1 and 2 are only valid in a noise-free case. In the general case, the measurements d _i is corrupted by measurement inaccuracies that are unique for each modulation frequency. These inaccuracies of measurement can have different causes and are referred to below as "noise". We assume that for every measurement i there exists an estimate σ _i about the strength of the noise. In first approximation can ${\sigma }_i\alpha \sqrt{t_{\text{exp}.i}}/f_i$

where t _{exp, i} and f _{i are} respectively the exposure time and the modulation frequency of the ith measurements. In the zeroth approximation, the σ _{i can} also be assumed to be identical. But of course, in the noise term σ _i also other effects can be considered.

The description of the basic idea is done with three measurements (D = 3), whereby the generalization to D measurements is obvious or described by a note.

According to Eq. (2) and the above comments can be a noisy 3-frequency measurement with the three results d ₁ , d ₂ , d ₃ represent about the context: $$d\equiv \left(\begin{array}{c}{\overline{d}}_1\\ {\overline{d}}_2\\ {\overline{d}}_3\end{array}\right)=t\cdot U{R}_{\text{m}}\left(\begin{array}{c}1\\ 1\\ 1\end{array}\right)+U{R}_{\text{m}}\left(\begin{array}{c}{m}_1\\ m_2\\ m_3\end{array}\right)+\left(\begin{array}{c}{\sigma }_1\cdot {\eta }_1\\ {\sigma }_2\cdot {\eta }_2\\ {\sigma }_3\cdot {\eta }_3\end{array}\right),$$

Here, m ₁ , m ₂ , m _{3 are} integers that result as listed above. The η _i in the rightmost term are random numbers from a (not necessarily Gaussian) probability distribution with mean 0 and width 1 , This describes the noise behavior of the individual measurements. By way of example, such a noisy measuring point d in 1 located. The task is now such a noisy measuring point d a true or at least most likely distance d assigned.

und mit η = 0 bestimmen sich die Raumdiagonalen s_0,m im rauschnomierten Raum: $$s_{\mathrm{0,}\text{m}}=t\cdot U{R}_{\text{m}}\left(\begin{array}{c}1\text{/}{\sigma }_1\\ 1\text{/}{\sigma }_2\\ 1\text{/}{\sigma }_3\end{array}\right)+U{R}_{\text{m}}\left(\begin{array}{c}{m}_1/{\sigma }_1\\ m_2/{\sigma }_2\\ m_3/{\sigma }_3\end{array}\right)$$

To find the most likely object distance d from noisy measurements it is helpful to normalize the measurements with the corresponding noise strengths σ _i : $$s\equiv \left(\begin{array}{c}{s}_1\\ s_2\\ s_3\end{array}\right):=\left(\begin{array}{c}{\overline{d}}_1/{\sigma }_1\\ {\overline{d}}_2/{\sigma }_2\\ {\overline{d}}_3/{\sigma }_3\end{array}\right)=t\cdot U{R}_{\text{m}}\left(\begin{array}{c}1\text{/}{\mathrm{<figure-callout\; id="1"\; label="\sigma "\; filenames="DE102018104668B4\_0015.png"\; state="\{\{state\}\}">\sigma </figure-callout>}}_1\\ 1\text{/}{\mathrm{<figure-callout\; id="2"\; label="\sigma "\; filenames="DE102018104668B4\_0015.png"\; state="\{\{state\}\}">\sigma </figure-callout>}}_2\\ 1\text{/}{\sigma }_3\end{array}\right)+U{R}_{\text{m}}\left(\begin{array}{c}{\mathrm{<figure-callout\; id="1"\; label="m"\; filenames="DE102018104668B4\_0015.png"\; state="\{\{state\}\}">m</figure-callout>}}_1/{\sigma }_1\\ {\mathrm{<figure-callout\; id="2"\; label="m"\; filenames="DE102018104668B4\_0015.png"\; state="\{\{state\}\}">m</figure-callout>}}_2/{\sigma }_2\\ {\mathrm{<figure-callout\; id="3"\; label="m"\; filenames="DE102018104668B4\_0015.png"\; state="\{\{state\}\}">m</figure-callout>}}_3/{\sigma }_3\end{array}\right)+\left(\begin{array}{c}{\mathrm{<figure-callout\; id="1"\; label="\eta "\; filenames="DE102018104668B4\_0015.png"\; state="\{\{state\}\}">\eta </figure-callout>}}_1\\ {\mathrm{<figure-callout\; id="2"\; label="\eta "\; filenames="DE102018104668B4\_0015.png"\; state="\{\{state\}\}">\eta </figure-callout>}}_2\\ {\eta }_3\end{array}\right).$$

and with η = 0, the spatial diagonals are determined s _{0, m} in the room: $$s_{0\text{m}}=t\cdot U{R}_{\text{m}}\left(\begin{array}{c}1\text{/}{\mathrm{<figure-callout\; id="1"\; label="\sigma "\; filenames="DE102018104668B4\_0015.png"\; state="\{\{state\}\}">\sigma </figure-callout>}}_1\\ 1\text{/}{\mathrm{<figure-callout\; id="2"\; label="\sigma "\; filenames="DE102018104668B4\_0015.png"\; state="\{\{state\}\}">\sigma </figure-callout>}}_2\\ 1\text{/}{\sigma }_3\end{array}\right)+U{R}_{\text{m}}\left(\begin{array}{c}{\mathrm{<figure-callout\; id="1"\; label="m"\; filenames="DE102018104668B4\_0015.png"\; state="\{\{state\}\}">m</figure-callout>}}_1/{\sigma }_1\\ {\mathrm{<figure-callout\; id="2"\; label="m"\; filenames="DE102018104668B4\_0015.png"\; state="\{\{state\}\}">m</figure-callout>}}_2/{\sigma }_2\\ {\mathrm{<figure-callout\; id="3"\; label="m"\; filenames="DE102018104668B4\_0015.png"\; state="\{\{state\}\}">m</figure-callout>}}_3/{\sigma }_3\end{array}\right)$$

This gives, as in 2 shown a to 1 similar diagram, in which the diagonals are now oriented in the spatial direction (1 / σ ₁ , 1 / σ ₂ , 1 / σ ₃ ). The index m specifies by the offset which of the diagonals it is.

The noise term (η ₁ , η ₂ , η ₃ ,) ^T now turns out to be a spherically symmetric deviation of the measurement from the idealized value s _{0, m} indicated by the cloud 110 ,

For a given measurement s is the most likely object distance, or the most likely, cut-down distance value s ∈ s _{0, m} one point s on a room diagonal s _{0, m} , the measuring point or the measured distance value s is closest.

In a distance space according to Eq. (3) this would only be the case if the noise were the same in all spatial directions, which is not typically the case.

To find the most probable object distance, various approaches, e.g. a brute-force approach, possible.

entlang der Raumdiagonalen gerichtet ist und n₁, n₂ das orthonormale Komplement zu n₃ sind, sodass n₁, n₂, n₃ eine Orthonormal-Basis bilden. Im allgemeinen Fall mit D > 3 Messungen ist n_D analog zu Gl. (6) zu wählen und dann D - 1 orthonormale Basis-Vektoren n₁, ...,n_D-1 (z.B. durch das Gram-Schmidt'sche Orthogonalisierungsverfahren) zu finden.According to the invention, however, the following approach is preferred. As in 3 shown becomes the rauschnomierten room diagonals s _{0, m} an area 330 determined perpendicular to the space diagonals and the rauschnomierte space in the coordinate system of this surface with the base vectors n ₁ , n ₂ , n ₃ transformed, in the example $$n_{}$$$${}_3=\frac{1}{{\displaystyle {\Sigma }_i{\sigma }_i^{-2}}}\left(\begin{array}{c}{\sigma }_1^{-1}\\ {\sigma }_2^{-1}\\ {\sigma }_3^{-1}\end{array}\right)$$

is directed along the spatial diagonal and n ₁ , n _{2 are} the orthonormal complement to n ₃ , so that n ₁ , n ₂ , n _{3 form} an orthonormal basis. In the general case with D> 3 measurements, n _{D is} analogous to Eq. (6) and then find D - 1 orthonormal basis vectors n ₁ , ..., n _D-1 (eg by the Gram-Schmidt orthogonalisation method).

The area 330 is determined so that the rauschnormierte measuring point s on the surface 330 lies. The intersections of the spatial diagonals s _{0, m} with the area 330 are marked with p _{0, m} .

Now only the components in n ₁ and n ₂ direction (or in D measurements in n ₁ , ..., n _D-1 direction) have to be considered for the distance to the diagonals. It is therefore possible to calculate a distance of a measuring point s to the next diagonal s _{0, m} from three (or D) dimensions to two (or D - 1) dimensions. The corresponding procedure is in 4 shown.

On the to the room diagonals s _{0, m} orthogonal surface 330 the flock of the rauschnormierten space diagonals form s _{0, m} as points p _{0, m} or point grid p _{0, m} from. In order to simplify the evaluation, it is provided according to the invention to switch off the further calculations on the grid of points using grid vectors v ₁ and v ₂ (or v ₁ ,..., V _D-1 ). By this procedure, each spatial diagonal point p _{0, m can be represented} as an integer multiple of the grating vectors v ₁ , v ₂ . $$p_{0\text{m}}=G_1{\text{<figure-callout id="1" label="v" filenames="DE102018104668B4\_0015.png" state="{{state}}">v</figure-callout>}}_1+G_2{\text{<figure-callout id="2" label="v" filenames="DE102018104668B4\_0015.png" state="{{state}}">v</figure-callout>}}_2\text{With}{G}_i\in \mathbb{N}$$

The rauschnormierten distance values s can also be in the coordinate system of the grid vectors as transformed, rauschnormierte distance values p represent s → p: $$p=G_1{\text{<figure-callout id="1" label="v" filenames="DE102018104668B4\_0015.png" state="{{state}}">v</figure-callout>}}_1+G_2{\text{<figure-callout id="2" label="v" filenames="DE102018104668B4\_0015.png" state="{{state}}">v</figure-callout>}}_2$$

und floor-Wert $\lfloor g_i\rfloor$

der Faktoren zu berechnen, um anhand dessen den nächstgelegenen vier (bzw. 2^D-1) Raumdiagonal-Punkte p̃ ∈ p_0,m zu bestimmen. Durch die verwendeten floor- und ceiling-Operationen können (wie oben beschrieben) auch Offsets erreicht werden, die um ganzzahlige Vielfache von UR_i/UR_m in der i-ten Dimension verschoben sind. $$\begin{array}{c}{p}_{\mathrm{0,00}}=\lfloor g_1\rfloor {\text{v}}_1+\lfloor g_2\rfloor {\text{v}}_2\\ p_{\mathrm{0,01}}=\lfloor g_1\rfloor {\text{v}}_1+\lfloor g_2\rfloor {\text{v}}_2\\ p_{\mathrm{0,10}}=\lfloor g_1\rfloor {\text{v}}_1+\lfloor g_2\rfloor {\text{v}}_2\\ p_{\mathrm{0,11}}=\lfloor g_1\rfloor {\text{v}}_1+\lfloor g_2\rfloor {\text{v}}_2\end{array}$$

Since the measurements are generally not at a spatial diagonal point p _{0, m} , the factors g _{i are} typically not integer. To determine the most probable object distance or the most probable transformed distance value p ∈ p _{0, m} it is now planned to round up the factors or the so-called ceiling $\lceil G_i\rceil$

and floor value $\lfloor G_i\rfloor$

of the factors to calculate the nearest four (or 2 ^D-1 ) room diagonal points p ∈ p _{0, m} . The floor and ceiling operations used (as described above) can also achieve offsets that are shifted by integral multiples of UR _i / UR _m in the ith dimension. $$\begin{array}{c}{p}_{0.00}=\lfloor {\mathrm{<figure-callout\; id="1"\; label="G"\; filenames="DE102018104668B4\_0015.png"\; state="\{\{state\}\}">G</figure-callout>}}_1\rfloor {\text{<figure-callout id="1" label="v" filenames="DE102018104668B4\_0015.png" state="{{state}}">v</figure-callout>}}_1+\lfloor {\mathrm{<figure-callout\; id="2"\; label="G"\; filenames="DE102018104668B4\_0015.png"\; state="\{\{state\}\}">G</figure-callout>}}_2\rfloor {\text{v}}_2\\ p_{0.01}=\lfloor {\mathrm{<figure-callout\; id="1"\; label="G"\; filenames="DE102018104668B4\_0015.png"\; state="\{\{state\}\}">G</figure-callout>}}_1\rfloor {\text{<figure-callout id="1" label="v" filenames="DE102018104668B4\_0015.png" state="{{state}}">v</figure-callout>}}_1+\lfloor {\mathrm{<figure-callout\; id="2"\; label="G"\; filenames="DE102018104668B4\_0015.png"\; state="\{\{state\}\}">G</figure-callout>}}_2\rfloor {\text{v}}_2\\ p_{0.10}=\lfloor {\mathrm{<figure-callout\; id="1"\; label="G"\; filenames="DE102018104668B4\_0015.png"\; state="\{\{state\}\}">G</figure-callout>}}_1\rfloor {\text{<figure-callout id="1" label="v" filenames="DE102018104668B4\_0015.png" state="{{state}}">v</figure-callout>}}_1+\lfloor {\mathrm{<figure-callout\; id="2"\; label="G"\; filenames="DE102018104668B4\_0015.png"\; state="\{\{state\}\}">G</figure-callout>}}_2\rfloor {\text{v}}_2\\ p_{0.11}=\lfloor {\mathrm{<figure-callout\; id="1"\; label="G"\; filenames="DE102018104668B4\_0015.png"\; state="\{\{state\}\}">G</figure-callout>}}_1\rfloor {\text{<figure-callout id="1" label="v" filenames="DE102018104668B4\_0015.png" state="{{state}}">v</figure-callout>}}_1+\lfloor {\mathrm{<figure-callout\; id="2"\; label="G"\; filenames="DE102018104668B4\_0015.png"\; state="\{\{state\}\}">G</figure-callout>}}_2\rfloor {\text{<figure-callout id="2" label="v" filenames="DE102018104668B4\_0015.png" state="{{state}}">v</figure-callout>}}_2\end{array}$$

The most likely, transformed distance value p ∈ p _{0, ij} is the one with the smallest distance $$\tilde{p}={\text{arg min}}_{ij}\left|p-p_{0ij}\right|$$

In this case, ij corresponds to an offset m.

By inverse transformation of the determined, most likely, transformed distance value p ⇔ s ⇔ d can be the most probable distance value d determine.

The procedure can basically be summarized as follows:

• - Finden eines Vektors s̃, gekennzeichnet dadurch, dass er eine erlaubte (unverrauschte) Messung beschreibt und möglichst nahe an dem Vektor der mit Rauschstärken skalierten Messungen s = (s₁, ..., s_D)^T = (d ₁/σ₁, ...,d _D/σ_D)^T liegt.
• - Bestimmen der optimalen Objekt-Distanz aus dem Vektor s̃.
• - Bei der die erlaubten Messungen durch Geraden gemäß Gleichung 5 beschrieben sind
• - Bei der der näheste erlaubte Vektor s̃ dadurch gefunden wird, dass die Messung s auf eine Hyper-Ebene projiziert wird, die senkrecht auf den erlaubten Geraden steht, wodurch das D-dimensionale Problem auf ein (D - 1)-dimensionales reduziert wird.
• - Bei der das Bestimmen der nähesten Geraden darauf beruht, dass die Geraden auf einem regelmäßigen Gitter liegen und
  • o die skalierten Messungen s durch Linearkombination der Gittervektoren dieses Gitters beschrieben werden,
  • o die 2^D-1 benachbarten Gitterpunkte/Geraden aus den Koeffizienten der Linearkombination mit Hilfe der floor- und ceiling-Operation bestimmt werden und
  • o die Suche der nähesten Gerade / des nähesten Gitterpunkts auf die derart gefundenen Möglichkeiten eingeschränkt wird.
• - Bei der der Messbereich vorzugsweise auf einen Wert d_r eingeschränkt wird, der geringer ist, als der sich aus dem kleinsten gemeinsamen Vielfachen der Eindeutigkeitsbereiche gegebenen maximalen Messbereich. Hierbei werden Messwerte nur Punkten auf Geraden zugeordnet, die Objekt-Distanzen d ≤ d_r entsprechen.
• - Methode nach einem der vorherigen Ansprüche, bei der außerdem aus dem Abstand zwischen dem Vektor der skalierten Messung s und dem Vektor der nähesten erlaubten (unverrauschten) Messung s̃ ein Konfidenzwert für das Resultat bestimmt wird.
  • • Vorzugsweise werden mehrere Messungen gemacht, bei denen ein ähnliches Ergebnis wahrscheinlich ist (z.B. bei benachbarten Pixeln in einem Sensor-Array, oder zeitlich kurz aufeinanderfolgenden Messungen), bestehend aus
    • o Vergleich aller 2^D-1 benachbarter gefundener Geraden/Gitterpunkte für eine gegebene Messung mit denjenigen der anderen (ähnlichen) Messungen.
    • o Entscheidung für die zu wählende Gerade / den zu wählenden Gitterpunkt durch Mehrheitsvotum aus vorigem Vergleich,
  • • wobei nach Bedarf das Votum mit einem Konfidenzwert gewichtet wird.

It is a method for distance measurement with D ≥ 2 different uniqueness ranges UR ₁ , ..., UR _D , so that at true object distance d _w for each measurement only d _i = mod ( d _w , UR _i ) and the d _i continue to be corrupted by noise of strength σ _i .

• - Find a vector s , characterized by the fact that it describes a permitted (unobstructed) measurement and as close as possible to the vector of measurements scaled with noise power s = (s ₁ , ..., s _D ) ^T = ( d ₁ / σ ₁ , ..., d _D / σ _D ) ^T lies.
• Determine the optimal object distance from the vector s ,
• - In which the allowable measurements are described by straight lines according to Equation 5
• - At the closest allowed vector s is found by the fact that the measurement s is projected onto a hyperplane that is perpendicular to the allowed line, reducing the D-dimensional problem to a (D-1) -dimensional.
• - In which the determination of the closest line is based on the fact that the lines lie on a regular grid and
  • o the scaled measurements s be described by linear combination of the lattice vectors of this lattice,
  • o the 2 ^D-1 neighboring grid points / lines are determined from the coefficients of the linear combination by means of the floor and ceiling operation, and
  • o the search for the closest straight line / the nearest grid point is limited to the possibilities thus found.
• In which the measuring range is preferably restricted to a value d _r which is lower than the maximum measuring range given by the smallest common multiple of the unambiguous ranges. Here, measured values are only assigned to points on straight lines, the object distances d ≤ d _r .
• Method according to one of the preceding claims, further comprising the distance between the vector of the scaled measurement s and the vector of the closest allowed (undistorted) measurement s a confidence value is determined for the result.
  • Preferably, several measurements are made where a similar result is likely (eg adjacent pixels in a sensor array, or temporal short-term measurements) consisting of
    • o Comparison of all 2 ^D-1 adjacent found straight lines / grid points for a given measurement with those of the other (similar) measurements.
    • o decision for the straight line / lattice point to be selected by majority vote from previous comparison,
  • • where necessary, the vote is weighted with a confidence value.

LIST OF REFERENCE NUMBERS

d

Distance value

d _w

true distance value

d _{0, m}

allowed room diagonal

d

most likely distance value

s

rauschnormierter distance value

s _w

true rauschnormierter distance value

s _{0, m}

allowed auschnormierte space diagonal

s

most likely rauschnormierter distance value

p

transformed distance value

p

most likely, transformed distance value

Claims (9)

Method for operating a light transit time camera according to the phase measurement principle using at least two modulation frequencies, wherein the distance values (d) and allowed spatial diagonals (d _{0, m} ) vectorially described and normalized by means of a noise term (σ _i ) and in rauschnormiert distance values (s ) or in allowed, rauschnormierte room diagonals (s _{0, m} ) are converted. Method according to Claim 1 , with a determination of a most probable, rauschnormierten distance value s ∈ s _{0, m} on the allowed, auschnormierten space diagonal (s _{0, m} ), which comes closest to a measured, rauschnormierten distance value (s), back transformation of the most probable, rauschnormierten distance value s ∈ s _{0, m} in a most probable distance value (d). Method according to one of the preceding claims, in which a surface (330) which is orthogonal to the extracted space diagonals s _{0, m} and intersects the cut-out distance value (s) is determined, formation of an intersection coordinate system (v ₁ ,. v _D-1 ) starting from points of intersection (p _{0, m} ) of the spatial diagonals (s _{0, m} ) with the surface (330), so that the points of intersection (p _{0, m} ) are denoted by integer multiples of the unit vectors (v ₁ , .. ., v _D-1 ) of the intersection coordinate system, transformation of the measured, rauschnormierten distance value (s) in the intersection coordinate system (v ₁ , ..., v _D-1 ) in a transformed distance value (p), determination a most probable, transformed distance value (p ∈ p _{0, m} ), that the transformed distance value (p) nearest intersection point (p _{0, m} ), inverse transformation of the most probable transformed distance value (p) into a most probable distance value (d). Method according to Claim 3 in which the most probable, transformed distance value (p ∈ p _{0, m} ) is determined by first rounding up the vector factors (g ₁ , ..., g _D-1 ) of the transformed distance value (p) lying intersections (p _{0, m} ) are determined, and the distance is determined only to these closest intersection points (p _{0, m} ), wherein the intersection point (p _{0, m} ) with the smallest distance to the transformed distance value (p) the most likely transformed distance value (p) represents. Method according to one of the preceding claims, in which the distance measurements and the calculations of the distance values (d) are restricted to a maximum distance value (dmax). Method according to Claim 5 in which an assignment of a most probable range value (d, s, p) takes place only for distance values (d, s, p) which are smaller than the maximum distance value (d _max , S _max , p _max ). Method according to one of the preceding claims, in which, starting from the distance of the most probable range value (d, s, p) to the measured distance value (d, s, p), a confidence value is formed. Method according to one of the preceding claims, in which the assignment of at least one preceding or adjacent distance measurement is taken into account for the assignment of a most probable distance value (d, s, p). Time of flight camera designed to carry out one of the aforementioned methods.

Images