WO2003050565A1

WO2003050565A1 - Device for measuring position, orientation and trajectory of a solid object

Info

Publication number: WO2003050565A1
Application number: PCT/FR2002/004221
Authority: WO
Inventors: Jean-Claude Lehureau; Renaud Binet
Original assignee: Thales
Priority date: 2001-12-11
Filing date: 2002-12-06
Publication date: 2003-06-19
Also published as: FR2833342B1; FR2833342A1

Abstract

The invention concerns a measuring device comprising two speckle generators (16, 17) generating fields measured by two CCD matrices (14, 15) fixed on the mobile object (11), Said device combines high measuring precision and high dynamics.

Description

DEVICE FOR MEASURING THE POSITION, ORIENTATION AND TRAJECTORY OF A SOLID OBJECT

The present invention relates to a device for measuring the position, orientation and trajectory of a solid object.

Position measurement systems play a key role in a large number of industrial applications (metrology, industrial control, robotics, etc.). Their precision, their dynamics, their robustness, their cost, their conditions of use, are all determining factors. The cost of such a system depends on the precision and the dynamic desired.

Traditional position measurement systems are very diverse. The two main classes are systems:

Mechanicals

* Probes and rules: This is the most classic solution. It is potentially very precise but costly for large dynamics, and difficult to use.

* Accelerometers: The accuracy of acceleration measurement being 1 μg, it appears problems of drift with the time of measurement (approximately 1 cm / minute) which can be troublesome.

optical

* Stereoscopic triangulation with several cameras: The measurement accuracy is linked to the aperture of the cameras. This solution is therefore expensive in good quality optical equipment.

* Laser telemetry: Unclear for the distances considered (approximately 1 meter). Optical encoder wheels: Used widely in robotics, it is a precise, reliable, but expensive solution. * Interferometry: All these processes are very precise but are suitable for displacement measurements of the order of the wavelength. The ambiguity problems of the fringing sequences are still not resolved. Projection of a light network (fringes or patterns) on the object: Potentially precise solution but a problem of ambiguity remains for a displacement greater than the pitch of the fringes. With a displacement of 10cm per second and a step of 100μm, the acquisition rate for counting the fringes must be greater than 200 kHz, which requires the use of an ultra-fast and therefore expensive camera.

Of all these systems, none allows to combine precision, great dynamics, and low cost.

The subject of the invention is a device for measuring the position, orientation and trajectory of a mobile object, a device which is of low cost price and makes it possible to make measurements with great dynamics and high precision.

The measuring device according to the invention comprises at least two devices for generating optical random fields illuminating the entire area in which the moving object can move, at least one device for measuring the local field secured to the moving object, connected to a digital correlator device to which is connected a device for memorizing the characteristics of said random fields, the correlator device being connected to a device for determining the coordinates of said measurement devices.

The present invention will be better understood on reading the detailed description of an embodiment, taken by way of nonlimiting example and illustrated by the appended drawing, in which:

• Figure 1 is a diagram explaining how to produce a field of scab and showing characteristics of these scab; • Figure 2 is a block diagram of a simplified embodiment of the device of the invention; and

• Figure 3 is a simplified diagram of a measuring device according to the invention using an "optical pencil".

The object of the invention is a process making it possible first of all to reference the absolute position and orientation of a mobile object. To this end, the invention provides for moving this object in at least two random light fields. Preferably, these fields are of the “speckle” type (scab fields). These fields are referenced and saved beforehand, in electronic form. Indeed, it is easy today, thanks to the low cost of digital storage, to record the data relating to a light field on a Qace surface of approximately 1 m ² with an accuracy of 100 μm (which corresponds to 10 8-bit samples, i.e. 100 M bytes of data). The measurable dynamics of the positioning of the mobile object is then directly linked to the storage capacity used. In addition, thanks to the granularity properties of the speckles, a relative positioning of the moving object carried out between the acquisitions of its successive absolute positions significantly increases the measurement accuracy. Positioning relative to the reference speckle field provides poorer accuracy than relative positioning, but corrects the drifts of the relative measurements. The object to be positioned is equipped with 1, 2 or 3 light detectors (for example CCD detectors) allowing respectively to position in space one, two or three points of this object.

We have schematically represented in FIG. 1 the various parameters of a “speckle” field. This field is created by an illumination beam in coherent lighting 1 of a diffusing surface 2 (for example a frosted glass). This surface is assumed to be circular, of diameter d. The axis of the beam 1 is substantially perpendicular to the surface 2, and passes through its center. A grain of medium speckle 3, assumed to be elliptical in shape, is formed at the distance L from the diffusing surface 2. The characteristic dimensions of this grain 3 are statistically well known according to the laws of diffraction in free space; λ being the wavelength of beam 1, these grain dimensions are λL ² / d ² along the major axis of the ellipse (aligned with beam 1) and 2λLJd for the minor axis. An example of a speckle figure is shown in 4 (as can be observed downstream of the surface 2 on a screen), the characteristics of the speckle fields are known for example from “Topics in Applied Physics, Vol 9, Laser Speckle and Related Phenomena "by JC Dainty, from publisher Springer Verlag. The reference field is created by separately illuminating two temporally stable diffusers (such as surface 2), each by a coherent light source such as a laser.

We then obtain a reference field with two components. The mobile object in question can be illuminated in reflection or in transmission (the latter possibility is preferable for questions of stability of the optical characteristics). The speckle field thus created constitutes a reference for positioning in space due to the statistical heterogeneity of the shapes and dimensions of its grains. It is therefore possible to determine the positioning of a moving object in this field if we have previously mapped the two components of this reference field. The mapping of these two components is established once and for all at the desired resolution (by drawing the coordinates of the contours of the speckle grains) and is stored digitally on an appropriate computer medium (CD-R or hard disk, for example). The speckle field can thus be known in a given volume, but the measurement of the complex field of the speckle in one plane is sufficient, the values of the field in the other planes being determined by digital propagation of the speckle.

A speckle field is characterized by the size and characteristic orientation of its three-dimensional autocorrelation function, in other words by the "size" of a speckle grain. This size is given by the law of diffraction in space, as specified with reference to Figure 1. The mapping of an orthogonal plane to the diffusing surface (2) allows referencing in this plane with better precision than the transverse dimension of a speckle grain.

To measure the position and orientation of an object, two or three arrays of photodetectors (for example CCD or CMOS arrays) are fixed on it. These matrices are arranged from. so as to constantly receive the speckles field and are sufficiently distant from each other to obtain good measurement accuracy. The maximum of the 2D intercorrelation function of the field detected by each of these matrices from a referenced speckle field gives the position of the object in an XOY plane parallel to the diffusing surface, which therefore provides two parameters of position of this object. This intercorrelation is calculated in a manner known per se, by a digital computer. This measurement process is illustrated in FIG. 2. A laser source 5 orthogonally illuminates a circular diffusing surface 6 at its center. A matrix 7 of CCD type photodetectors is fixed on the object in question (not shown) and the latter is manipulated so that the sensitive surface of the matrix is always substantially parallel to the diffusing surface 6 and turned towards the latter. A digital correlator 8 receives the information from the matrix 7 on the one hand, and from a memory 9 in which the data relating to the speckle field thus created are stored (grain topology and corresponding coordinates). When the correlator 8 determines an intercorrelation peak between the information it receives from the matrix and data from the memory 9, it transmits the coordinates of a characteristic point of the matrix (for example its center) such that it was previously registered for the same speckles configuration. Two speckle fields are produced by placing two sources of coherent light illuminating two diffusers whose optical axes are convergent and form an angle between them whose minimum value is fixed by the desired measurement precision in the direction of the distance from the two sources. . Thus, for a substantially identical accuracy in all directions, this angle is preferably between 40 and 60 °, while a value of 20 ° would result in an imprecision three times greater in the direction of the distance than in the other two directions. There are several detector arrays on the object, each detector matrix provides four measurement parameters for a given position of these matrices. One can then easily determine with redundancy the position in space of each of these matrices, in a fixed reference with respect to the diffusing surfaces. The calculation times of current computers (PC type) today do not yet allow measurements to be made in real time (in particular even for a relatively slow movement of a few cm. S), the measurement is made a posteriori. To facilitate the software processing of the data supplied by the detectors, they are preferably oriented in the same direction, which makes it possible to have the same angle of rotation for these detectors.

In order for the determination of the intercorrelation peak to be unambiguous, the amplitude of this peak must be approximately seven times greater than the amplitude (in rms value) of the intercorrelation noise, ie a signal / noise ratio of 17 dB in power. However, this ratio is linked to the number of independent modes in the field intercepted by a matrix, that is to say to the number N _s of speckles grains thus intercepted. The signal / noise ratio is substantially equal to

in amplitude, if we neglect the random measurement noise. It is therefore necessary that the array of detectors intercepts at least 50 speckle grains so that the position measurement can be made without ambiguity. The position of the intercorrelation peak is interpolated in order to determine its value most precisely. It is commonly accepted that the positioning accuracy of this peak is equal to 1/100 ^th of a pixel if the signal / noise ratio is good (17 dB as specified above).

The amplitude and the shape of the intercorrelation peak depend on the orientation of the detector array. A rotation of an angle θ of this matrix with respect to its ideal angular position (parallel to the diffusing surface) induces an affine distortion of the measurement compared to the referenced field. This distortion in turn induces a peak distortion of the same amount. Fortunately, this distortion is only a function of cos θ, which means that if θ is less than 30 °, the distortion is negligible.

The intercorrelation peak is also sensitive to the distance between the matrix and the diffusing surface. The tolerance on the position of the matrix on the Z axis (see Figure 2) is a function of the longitudinal size of the speckle grains. For example, for a speckle field created by a laser source at the wavelength of 633 nm, with grains of transverse dimension of approximately 100 μm, the longitudinal dimension is approximately 60 mm. The influence of this parameter is therefore not very significant on the characteristics of the peak. In addition, other speckle formation plans can be calculated from the data of a measured complex field, which makes it possible to choose a plane appropriate to the position of the matrices in space.

In conclusion, the uncertainty on the position of the object carrying the detector arrays depends on the observation base, i.e. on the angle formed by the normals at the two diffusing surfaces, and on the orientation of the detector arrays in relation to this observation base. According to an advantageous aspect of the invention, provision is made to provide the user with several possible bases for orientation and distance to the diffusers, in order to adapt the system to the desired measurement precision. The orientation of the two diffusing surfaces need not be known precisely if a calibration is carried out before the measurement, which is made possible by said redundancies of the measurements. A rigorous mathematical analysis shows that by moving a sensor on a random trajectory, it is possible to exactly position the diffusers with respect to each other, to the nearest homothetic factor. In the application to a “pencil”, described below with reference to FIG. 3, the precise knowledge of the distance separating two sensors from this “pencil” makes it possible to remove this uncertainty linked to the factor of homothety.

In the case of the measurement of a trajectory of the object in question, if the measurement of position and orientation of this object using a reference speckle field offers a precision limited by several parameters (l absolute orientation of the detectors, their distance from the reference planes and other factors mentioned above), it is possible to additionally determine the relative position of a detector linked to the object in question, of an acquisition measurement to the next, if there is an overlap between the images intercepted by this detector for these successive acquisitions. In this case, the measurement accuracy is much better, as there may be fewer disparities between the relative measurements than between absolute measurements that are not linked to each other. Such a complementary measurement can be used to refine the absolute measurement made from the reference field.

FIG. 3 shows a diagram of an application of the device of the invention, for measuring the trajectory of an optical "pencil". An optical "pencil" 11 is equipped with a tip 11 A of optical telemetry producing a beam 12 focused at a point F. This pencil 11 is manipulated so as to make point F follow the outline of the part 13 (or any part of this room whose topography we want to raise).

Two dies of CCD detectors 14,15 which are illuminated by two fields of speckles produced by two generators 16,17 are fixed on the pencil 11. These generators 16, 17 are produced in the manner described above. Their axes (normal to their diffusing surfaces) 16 A, 17 A respectively, converge at a point P and form an angle A of about 20 ° between them. The relative arrangement of the generators 16,17 and of the pencil 11 is such that when this pencil moves, the detectors 14,15 are always as close as possible to point P. The pencil trajectory, that is to say its relative displacement, is obtained from its successive coordinates (calculated by a calculation device such as the device 10 in FIG. 2) which are processed by appropriate software. This gives information on the part 13 analyzed, for example its outline, which can be displayed on a screen.

According to an application example, the acquisition of the successive coordinates of the pencil is done over approximately 10 cm. The position and orientation of the pencil must be known with an accuracy between 1 and 10 μm over the entire length of its trajectory. The distance between the pencil and the object 13 must be kept fixed to the nearest 15 mm (so that the point of convergence F of its detection beam is well focused on the surface to be explored).

In an exemplary embodiment of the measuring device of the invention, the laser source of the speckle generator had a power of 30 mW at the wavelength of 650 nm. The diffusing surfaces were in frosted glass, lit by transmission. It should be noted here that the angular diffusion function of these diffusers is not isotropic. The scattering is angularly concentrated around the zero order of the lighting laser beam, so that approximately 90% of the light energy is concentrated in the speckle field analyzed. The diffuser and the laser are rigidly linked. The angle A between the two diffusing surfaces was 20 °, in order to avoid excessive distortion of the correlation peaks. The distance between the diffusers and the pencil was about 1m. The average dimension of speckle grains at the pencil level was 100 μm transversely and 60 mm longitudinally, that is to say an illumination of the diffuser on a surface of approximately 6 mm ² . The analysis area of the speckles field was 0.25 m ² . Each of the pencil detectors had the following characteristics: square matrix of 256 X 256 pixels, pixel size: 10 μm, quantum efficiency: 0.8, sampling of the video signal on 8 bits, electronic shutter with integration time of 0, 1 ms (to allow movements of the pencil up to 1 m / s), video rate of 1 kHz, reading noise of 1000 photoelectrons. An interference filter suppressed the surrounding white light. The constraints imposed on the handling of the pencil to perform optimal measurements were as follows: initially, the pencil was oriented in such a way that the detectors were facing the diffusing surfaces (to avoid distortions as much as possible), and during the measurement, the surfaces of the detectors should not rotate by an angle of more than 30 ° relative to their initial orientation. The preliminary measurement of the reference field is carried out at once by electronic holography in the laboratory, using a high-performance camera and precise means of translation to ensure the movements of this camera. One can thus raise the complex field of speckle, and thus obtain by a computation of beam propagation the knowledge in volume of the speckle. Thus, for example, for a measurement surface of 0.25 m ² and a speckle grain size of 100 μm, the complex field is stored in the form of complex 64 × 64 pixel images, in 16 bits, the pixel dimension being 50 μm. Sampling then takes place without loss of information. The memory size of the recording medium required is 50 Mbytes per broadcaster, or 100 Mbytes for two broadcasters.

For the same example of embodiment as above, we find that the number of photo-electrons involved in the calculation of a correlation peak is approximately 2400, the reading noise being 1000 photo-electrons, which means that the signal / noise ratio for each pixel is 7 dB in power.

With the detection matrix of 256X256 pixels and speckle grains with a diameter of 100 μm, we intercept on average 25X25 speckle gains, and the signal-to-noise ratio of the cross-correlation peak, without distortion, is therefore 28 dB , which ensures unambiguous peak detection. The signal-to-noise ratio of the correlation peak is more than 60 dB, which ensures that the position of the peak can be obtained at 1% of its width, that is to within 1 μm. The relative position measurements between each video frame do not need to be corrected as a function of the rotation of the pencil, since this rotation can be considered negligible in the space of 1 ms.

The pencil trajectory is calculated a posteriori. Given the relative dimensions of the speckle grains and the pixels, a real-time pre-processing of the images is carried out in order to reduce them to images of 64 × 64 pixels, without loss of information. For one minute of video acquisition, the number of images to be stored from each CCD matrix is 60,000, unless there are high-performance calculation means enabling the relative movements of the images to be calculated in real time. In this case, it would suffice to store only one image on M (with 10 <M <100), that is to say only the images making it possible to correct the biases by positioning them with respect to the reference speckle. If the movements printed by the user in pencil are at low frequency, the trajectory can be interpolated from non-successive frames. If this real-time calculation or interpolation is not possible, 64 ² X 60000 X2 = 500 Mbytes / minute of data must be stored for their subsequent analysis.

The following program can be implemented to analyze the data provided by each detector:

For the two detectors, carry out: Search for the first image in the reference speckle field. We give ourselves a research area of a few centimeters.

Calculation time: 40 correlations (5 ms each) + 20 rotations (1 ms each) <300 ms

A) For each new image of each detector, calculate the position of the intercorrelation peaks with the previous image.

Calculation time: 10 ms by software. This time could be reduced by an order of magnitude by calculating this cross-correlation using dedicated FFT processors.

An image on M, perform: B) For each angle θ = [-1 °, 0, + 1 °]

Calculation of the new rotation matrix compared to the reference.

C) For each detector

- Rotation of the image of θ compared to the previous one. - Search for the 4 correlation peaks in a given interval

End of loop C End of loop B

Sub-pixel interpolation of the two peaks found.

Estimation of the new trajectory and the new area of research End of loop A

The trajectory is calculated by rough registration (50 μm) of the positions on the reference fields, then by refinement with the relative position data.

Calculation time by increment:

- 3 rotations of the image to the nearest neighbor = 3 ms

- calculation of 12 correlations = 60 ms

- A "sub-pixel" interpolation would be negligible if we proceeded by simple interpolation (polynomial or barycenter)

Calculation time per increment is less than 100 ms

If M> 100, the calculation is therefore feasible during the acquisition. The total calculation time is therefore limited by the processing of the relative displacements (10 ms each). It is therefore important to be able to do this calculation by dedicated processor to reduce this calculation time. In this case, the treatment would last less than a minute. Otherwise, it would take 10 times more processing time than acquisition time. Another solution if the movements are slow consists in spacing the relative measures and therefore reducing the number of measures.

Claims

1 - Device for measuring the position, orientation and trajectory of a solid object, characterized in that it comprises at least two devices for generating optical random fields (5,16,17) illuminating the entire area in which can move the mobile object (11), at least one local field measurement device (7,14,15) integral with the mobile object, connected to a digital correlating device (8) to which is connected a storage device (9) characteristics of said random fields, this correlating device being connected to a device (10) for determining the coordinates of said (said) measuring devices.

2- A measuring device according to claim 1, characterized in that the random fields are fields produced by diffusers illuminated by a coherent beam (1 -2, 5-6, 16-17).

3 - Device according to claim 2, characterized in that the optical axes of the diffusers are convergent (P) and form between them an angle of about 20 °.

4 - Device according to one of claims 2 or 3, characterized in that the "speckle" generators comprise a coherent light source (1, 5) illuminating a diffusing surface (2,6)

5 - Device according to one of the preceding claims, characterized in that the local field measurement devices are arrays of photodetectors (7,14,15).

6 - Device according to one of claims 3 to 5, characterized in that the mobile object is an optical "pencil" (11) on which are fixed two local optical field measurement devices (14, 15).

7- Device according to claim 6, characterized in that during movements of the "pencil", it is maintained so that the point of convergence (P) of the optical axes of the two generators of "speckle" is maintained substantially equidistant from the two optical field measuring devices (14,15).