FR2838514A1 - Motor vehicle parallactic telemetry system error correction method in which a distance measured using the system is compared with one independently determined using a relative velocity method so that corrections can be applied - Google Patents

Abstract

Method for detecting a systematic error in a parallactic telemetry system (12, 14) used in a motor vehicle (10). Accordingly a distance (Dp) is measured using the parallactic measurement system and using an independent relative velocity method. From the difference a systematic error can be corrected.

Description

light source (14) is a laser.

  The present invention relates to a systematic error detection method in parallactic telemetry systems in

motor vehicles.

  STATE OF THE ART Motor vehicles are increasingly equipped with driver assistance systems (FAS systems) which assist the driver in his driving or facilitate special driving maneuvers. Such systems are also known as ADAS systems. In general, such systems require sensors providing the o necessary information concerning the environment of the vehicle, for example the layout of the boot, obstacles on the shoe or next to it. In some known systems, the sensors include a radar system which allows a direct measurement of the distance and relative speed of the objects reflecting the radar waves. In other cases, a Lidar system is used which allows telemetry on the basis of optical signal travel time and generally has higher angular resolution than radar systems. But as existing traffic information and driving systems are optimized primarily by visual perceptibility by a driver, video sensors seem to be particularly suitable

  for driver assistance systems.

  In principle, the video sensors consist of at least two optical sensors, for example two video cameras installed at a known distance, that is to say the basic width in the dirty transverse direction of the vehicle to allow to determine no only directional information but also the removal of objects by

  apparent parallactic shift of the objects detected by the two sensors.

  These optical sensors thus provide at least the function of a parallactic telemetry system. The results of distance measurements can, however, be distorted by systematic errors

  of an adjustment error of one or both optical sensors.

  OBJECT OF THE INVENTION The object of the present invention is to develop a method enabling the automatic detection and, where appropriate, the correction of such

s systematic errors.

  Solution and advantage of the invention For this purpose, the invention relates to a method of the type defined above, characterized in that it differentiates the distance measured Dp parallactically from at least one object as a function of time and Independently of this, the relative speed V of this object is determined, and a systematic error is known if the results differ in a different way.

significant from each other.

  The basic idea of the invention consists in forming, by time derivation of the distance measured in a parallactic manner, a value of the relative speed of the object, and for motor vehicles it is relatively simple to determine the The relative velocity of many objects also depends on the parallactic measurement, so that, by comparison of the results, we can conclude that there is a systemic error. For the independent determination of the relative speed, immovable objects are used, for example circulation panels, directional panels or the like, placed at the edge of the sill. For those objects that can be identified relatively easily by electronic image processing or by statistical analysis of the frequency distribution of the relative velocities, the relative velocity corresponds to the amplitude of the velocity of the velocity. vehicle that can

  measure in a relatively accurate way with the usual rangefinders.

  The invention is not limited to the exploitation of immovable objects. If, for example, in addition to the video sensors there is a radar system, it is also possible to use as a comparison value the relative speed measured by the radar system and the moving objects. Such a radar system allows direct telemetry, but the method according to the invention has the advantage of allowing a comparative measurement independent of the measurement errors of the radar telemetry and of having a greater precision when detecting the radar. 'fitting errors that get pro

  especially in the near area.

  According to a preferred characteristic of the method not only systematic errors are detected according to the method of the invention but they are automatically corrected by calculating relative speeds measured in various ways, a correction for distances measured by the parallax. In particular, this makes it possible to correct systematic errors based on a false value of the base width and / or a misalignment

  s gulaire of the two optical sensors.

  For example, the correction consists of multiplying the measured distance Dp by a correction coefficient 1 / F and calculating a correction value AB of the basic width B of the television system.

  parallactic metrics, where a correction value Ap of an error

  the angle of adjustment of the sensors of the telemetry system.

  According to one embodiment of the method, by electronic processing of the image, objects considered to be immobile are identified, for example traffic signs, steering poles, shoe markings or similar panels. In this case, the exploitation can be done separately for each object identified, so that the corrections calculated for the different objects will be checked.

  as for their consistency and accuracy will increase by the average value.

  o For example, if the relative speed of moving objects is determined from the vehicle's own speed V, the objects are detected.

  as still objects by electronic image processing.

  According to another variant of the method, it is used that immobile objects occur statistically less frequently than moving objects, and this is why in a histogram which gives the frequencies of the relative velocities measured parallatically for a relatively large set of objects. objects, they become perceptible in a very limited maximum for a zero relative speed. The adjustment errors which lead to distorting the measured distances of a constant coefficient over time, are then observed in that this maximum along the axis of velocities is shifted with respect to the zero value without the width of the maximum does not change. Adjustment errors which distort a time-dependent coefficient are advantageously developed so that the measured relative velocities in an extended angular range are loaded with noise so that the amplitude of the maximum decreases and its

width increases.

  Thus, we establish a histogram of the frequency (n) at which the relative velocities Vp obtained by differentiation of the measured distances Dp as a function of time occur and the immobile objects are recognized by their frequency in the histogram. Moreover, in order to calculate the correction coefficient 1 / F, the quotient of the relative velocity Vp measured parallactically and the actual speed V of the vehicle are calculated, if the velocity distribution of the immobile objects in the histogram is not greater than the statistical measurement error, and a

  correction of the correction coefficient 1 / F as a function of time if the

  velocity of immobile objects in the histogram is greater than

than the statistical error.

  Drawings The present invention will be described hereinafter with the aid of exemplary embodiments shown in the accompanying drawings in which: FIG. 1 is a schematic diagram of a paternactic telemetry system of a motor vehicle FIG. 2 is a diagram for explaining a method of determining the correction parameter; FIGS. 3 and 4 are histograms showing the frequency distribution of measured parallactic relative velocities, with and without

adjustment error.

  Description of embodiments

  Figure 1 shows the front part of a motor vehicle shown schematically in plan view. The vehicle 10 is equipped with a driving assistance system, known per se and not presented in detail for this reason. The video sensor system is composed of, among other things, two video cameras 12, 14 installed at a certain lateral spacing with respect to each other in the front area of the vehicle. By an electronic image processing by the two cameras vio deo 12, 14, objects 16, 18 are detected and identified. In the example presented, the objects 16, 18 are fixed objects situated at the edge of the eye sce. The optical axes 20, 22 of the video cameras 12, 14 are parallel to the longitudinal axis of the vehicle if the cameras are adjusted correctly; their spacing defines the width B of the base of a telemetry system formed by the two video cameras 12, 14. The exploitation of the image information of each camera defines for each object, for example for the object 18, the azimuth angle pL and pR according to which the object is seen by the corresponding camera. The difference pL - pR is the parallax p of the object 18 making it possible to determine the distance D of the object in the direction parallel to the optical axes 20, 22 in conjunction with the width B of

the base.

  By approximation, for very large distances (and thus small azimuth angles), the distance D is calculated as follows: BL / D = tanpL (pL BR / D = tanpR pR The quantities BL and BR give, in each case, the shift

  side of the object 18 with respect to the optical axis of the camera concerned.

  Thus, we have the following relation: B = BL-BR = D * ((pL - (pR) = D * (p D = B / (p

  Defective adjustment of video cameras 12, 14 may result in

  on the one hand, to an error AB of the basic width and, on the other hand, to an error Ap of the parallax. As an example, FIG. 1 shows, in phantom, the optical axis 20 'in the case where the video camera 12 is incorrectly adjusted by the angle Δ p, the measured azimuth angle ρ is then increased by Ap value, that is, the video system "sees" object 18 at location 18 '. The distance Dp measured by parallax of the object 18 is then too small compared to the value D and the error AD. The error calculation makes it possible to establish the following relations: AD = (5D / 5B) AB + (0D / (p) A (p AD = (l / (p) AB- (B / (p2) A (p Dp = D + AD = (B / p) + (AB / (p) - (B / (p2) A (p Dp = (B / (p) * (1 + (AB / B) - A (p / (p) Dp = D * (1 + - ((p / B) D) If the error quotient Dp / D is called F. then for the approximation we have the following relation: F = 1 + - (( p / B) * D, with = AB / B In general, we have: F = 1 + AD / D = 1 + ((5D / 0B) / D) AB + ((0D / (p) / D) A (p Dp = D * F If we differentiate the equations as a function of time, we obtain a parallel value Vp for the relative velocity of the object 18: Vp = 0 / 0t (Dp) = 0 / ôt (D * F) = V * F + D * F / at in this formula, V represents the effective relative velocity of the object to be known to carry out the process described here.

by another source.

  Subsequently, it is assumed that the object 18 has been recognized as being an immobile object by electronic image processing, for example a steering column. In this case, the actual relative velocity V is equal to the actual speed of the vehicle measured directly by the tachometer of the vehicle. For the quotient Vp / V we obtain: Vp / V = F + ((5F / 0t) / V) * D In the special case where the error quotient F does not depend on the time, this error quotient F is directly given by the ratio VpV and a more accurate corrected value Dc is obtained for the distance D of the object 18 by multiplying the distance Dp measured parallactically with the correction coefficient 1 / F:

Dc = (V / Vp) Dp.

  In a general way we have: Vp / V = F + A * D, with A = (0F / ôt) / V If we admit that the coefficient A is in first approximation independent of the distance D, then we determine A from the quotient Vp / V measured for several objects by linear regression. We obtain thus: Dc = Dp / F = Dp / ((Vp / V) -A * D) This hypothesis is for example fulfilled if the quotient of error F is given as above by the following relation F = 1 + - ((p / B) D We thus obtain: Vp / V = F + ((0F / ôt) / V) * D 1 + - ((p / B) D + (/ 8t [1 + - ((p / B) * D]) / V) * D s = 1 + - ((p / B) * D - ((p / B) * V) / V) D = 1 + - ((p / B) * D- (<p / B) D Vp / V = 1 + -2 * (A (p / B) * D (1) Vp / V = F + A * D, with F = 1 + and A = -2 * ((p / B) In Figure 2, for several objects which are each represented by a measuring point X, the coefficient Vp / V is represented in relation to the measured distance Dp. on a line that intersects the ordinates at the point (Dp = 0, 1 + A) The slope is equal to 2 * (p / B) With the point of intersection of the y-axis and the slope of this right we set the correction value for both the width B of the base and for the error of paral

lax A <p.

  The calculation means to be used for the method described above are relatively small and in large part it is a matter of separating stationary objects from their classification as signposts, steering poles or the like. The use of statistical procedures makes it possible to further reduce the calculations to be made.

  as will be described hereinafter with reference to FIGS. 3 and 4.

  2s. FIG. 3 shows a histogram giving the frequency (n) representing the relative speed for the parallax measurement with respect to the measured speed Vp. Measured speeds are normalized

  here on the own speed V of the vehicle.

  Curve 26 drawn in full line shows the frequency distribution in the case where video cameras 12, 14 are adjusted correctly. Moving objects form a relatively flat and wide background signal because the velocities of moving objects are in general very highly dispersed. All stationary objects thus have the same speed Vp = V, so that the curve 26 has a very high maximum and very limited

for the value 1.

  The curve 28 shown in dotted lines shows the frequency distribution for the case of a time-dependent fault F (for example F = 1 + A). In this case, the measured speeds of immob

  they are all offset by the same value so that the maximum

  also appears offset from this value without its shape being modified

importantly.

  FIG. 4 shows on the contrary the curve 30 represented in dashed lines which corresponds to the frequency distribution in the case of a time-dependent error F (for example F = 1 + - 2 * (<p / B) * D). In this case, the maximum generated by moving objects is less offset and is flatter and wider. Nevertheless, it is also possible to apply an error correction method similar to that of FIG. 2. If the histograms are taken at different times, not only will the measured distances Dp of the different objects be distinguished, but also their value.

  respective average. Thus, as in Figure 2, it will be possible to record

  measurement points and for Vp the maximum value will be

  of the curve 30 and for Dp each time the average value of the objects which forms the maximum accentuated curve 30. In a modified embodiment, which requires even less calculation, it is represented, according to FIG. histograms of objects that are between two relatively narrow distance windows. The two distance windows then provide two measurement points in the diagram of FIG. 2 and make it possible to determine the straight line 24. As the adjustment errors are generally constant in FIG.

  time, this is acceptable without difficulty if the recording of the histo-

  grams takes a relatively long time.

  According to a development of the process, an even more precise error correction is made if not only the speeds

  relative parallax measurements but also in addition to the

  its angular, that is to say the drifts in time of the azimuth angles

  L and pR objects in front of which passes the vehicle 10. These speeds

  gules are also calculably influenced by and Ap so that in combination with equation (1) a system is obtained

  of equations from which we can calculate the two unknowns and Ap.

Claims (7)

  1) Systematic error detection method in telecommunication systems   parallactic metric (12, 14) in motor vehicles (10), characterized in that the measured distance Dp parallax of at least   an object (16, 18) as a function of time and independently of this we de-   terminates the relative speed V of this object, and a system error is recognized.   if the results differ significantly from each other. 2) Method according to claim 1, characterized in that, in case of detection of a systematic error, an error correction is automatically carried out by calculating a correction for   measured distance Dp, from the difference between the relative speed Vp ob-   held parallactically by deriving Dp, and the relative speed Vp me- independently.   3) Process according to claim 1, characterized in that   the correction consists of multiplying the measured distance Dp by a coefficient correction 1 / F.   4) Method according to claim 2, 2s, characterized in that a correction value AB of the basic width B of the system is calculated.   telemetry device (12, 14).    Method according to Claim 2 or Claim 4, characterized in that a correction value Ap of an error of the adjustment angle is calculated.   sensors (12, 14) of the telemetry system.   6) Method according to claim 1, 3s, characterized in that the relative speed of immobile objects (16, 18) is determined from the own speed V of the vehicle (10).   7) Method according to claim 6, characterized in that the objects (16, 18) are detected as immobile objects by treatment electronic image.   8) A method according to claim 7, characterized in that one establishes a hi stogram of the frequency (n) at which occur the relative velocities Vp obtained by differentiation of the measured distances o Dp as a function of time and recognize the immobile objects (16, 18) by their frequency in the histogram.   9) Process according to claims 3 and 8,   characterized in that for calculating the correction coefficient 1 / F, the quotient of the relative velocity Vp measured in a parallactic manner and the actual velocity V of the vehicle are calculated, if the velocity distribution is not stable in   the histogram is not greater than the statistical measurement error.   O) Method according to claim 9, characterized in that a correction of the correction coefficient 1 / F as a function of time is carried out if the dispersion of the speeds of the immobile objects in   the histogram is greater than the statistical error.