DE102008010965B4 - Robot web guide - Google Patents

Abstract

Method for three-dimensional path guidance of a robot head (26) with an optical test head (6) fastened thereto in a defined relative position on an object (1) along a desired path (24) in a defined working position (25) over an optically distinctive path (2) the following steps: optical scanning of the marked path 2 on the object 1 by the test head 6 by means of the triangulation light-section method, evaluation of the signals determined at the actual position in real time, constant readjustment of the actual path (34) of the robot head (26) in the direction of target path (24) to the test head (6) - both parallel to the longitudinal direction (10) and parallel to the transverse direction (11) of the target path (24) and thus the striking To keep web (2), - to keep at nominal distance (22) between the distinctive web (2) and the actual web (34), and - in a desired corridor (28) around the predetermined target path ( 24), characterized in that - that of the distinctive path ( 2) reflected, line-shaped images (4, 4 ') of two different optical sensors (12, 12') with the illumination direction (3) opposite viewing directions (5, 5 ') are recorded and - not sufficiently equal received light quantity of the two Sensors (12, 12 ') with the same viewing angle on a non-perpendicular irradiation direction (3) closed and this deviation of the irradiation direction (3) from the perpendicular (21) on the longitudinal direction (10) of the distinctive path (2) is automatically corrected.

Description

I. Field of application

The invention relates to a method for three-dimensional path guidance of a robot head along a desired path.

For the purposes of the present application, a robot head is to be understood as any controlled movable working head, regardless of whether it is actually fastened to a typical robot arm or is held by other guidance systems.

II. Technical background

In the industry, many operations are now done automatically by a machine. If the tool, for example in the form of a work head attached to the head of a robot arm, has to be moved along a stationary object to be machined, the robot head must be given the path of movement along which it is to move.

A typical example of this is the processing of an automobile bodywork in automobile production: If, for example, the soldered seam between the roof and the side part of a bodywork or just visually checked, the corresponding robot head must in a defined working position and distance along this Lötnaht be guided. In other applications, instead of a soldering seam another distinctive path or line of an object is traversed.

Under the working position of the robot head relative to the distinctive path or line is not only compliance with a given nominal distance understood, but beyond compliance with a certain angular position of the robot head relative to the longitudinal direction and transverse direction of this distinctive path z. B. exactly parallel to the nominal distance extending thereto on a target path on which the robot head or an attached optical test head to move.

For the determination of such a target path of the robot head or the attached optical test head, it was previously necessary to first determine the exact location of the traced striking line or orbit on the object.

In the case of the soldered seam on the automobile body, this means that at least the exact position of the soldered seam must first be determined at this workstation:
Either the location of this Lötnaht within the entire car body is known exactly enough. Then it is sufficient to determine their position in space by positioning or detecting a few control points on the body in order to know the exact position of the soldered seam.

In practice, however, the position of the solder seam within the body by production inaccuracies may still differ by at least a few tenths of a millimeter, so that the orientation of the body as a whole for the determination of the position of the solder seam at least for the post-processing or testing operations in question the soldering seam is not sufficient.

In this case, the actual position of the soldered seam itself must be directly determined, which so far happened as follows:
For example, the robot head with the optical test head attached thereto, which should determine its position by scanning the distinctive path in the form of the soldering seam, was manually guided once to each of the two end points of the soldering seam and scanned, possibly also at one or more points in between , approximately in the middle of the soldered seam. By means of these basic values, a rough position determination of the solder seam could be carried out automatically.

Subsequently, the robot head with the optical probe was once moved along the solder seam based on these coarse values in order to carry out a precise position determination of the soldered seam, also in the areas between the previously recorded individual points, which is generally referred to as "positioning run".

Apart from the small deviations of the position of the solder seam within the vehicle body was now also the location of the entire body in the room known with the location of the solder seam.

Now, the location of the soldering seam was exactly known, and in a subsequent operation, the desired operation (mechanical processing, painting, optical inspection or the like) could be carried out by the robot head equipped with a corresponding working head or tool again along the now exactly known Solder seam was performed.

However, the disadvantage of this approach becomes obvious when the actual work process can be handled relatively quickly:
Then the time required for first approaching a few points of the solder seam and the subsequent completion of the positioning run is very high Comparison to the time required for the actual work process.

In addition, the processing step to be performed by means of the robot head is often not mechanical processing, but merely the optical inspection, for example including the measurement of a three-dimensional contour, in order to subject this contour to quality control. The determination of the shape or other result data related to the surface shape, e.g. B. the volume is often carried out by means of the light section method.

In this case, a fan-shaped, so spread in only one plane, light beam is applied to the surface to be examined and observed at an angle, usually an acute angle to the irradiation direction, so that the course of the image of the fan-shaped beam on the surface there can recognize existing elevations in that the observed image also shows an increase when the line-shaped light beam fan passes over this elevation (unless the course of the light line is parallel to the direction of the elevation).

Such individual images of the light line on the object can be made many times at short time intervals while the optical test head moves relatively and transversely to the light line, so that the three-dimensional surface design can be determined by placing these individual height sections, ie scans, one behind the other / or related parameters such as height, volume, width, location of the distinctive track, etc ..

Also, solder seams between adjacent sheet metal parts in the automotive industry are examined in this way on their shape and thus quality, especially inadmissible recesses in the solder seam to be found before painting.

However, if the solder seam inspection is performed only at a single viewing angle, it may be that pits are not detected or, in turn, flawless areas are reported as pits, as one primarily seeks to detect a pit from that of the pit reflected amount of light is at least less, because no light is reflected any more than the amount of light reflected from the environment.

However, if in the direction of the camera, the depression is sealed off by an upstream bead, a collection in the form of a dust particle, etc., this is not recognized.

In contrast, the reduction behind a dust particle survey or even the worse reflection behavior of a flat, non-recessed point of the surface in the form of a color point, etc., lead to the display as a defect.

In this sense, the describes DE 44 05 865 A1 just viewing under a single viewing angle, so with a single sensor, and can not solve this problem either.

goal of DE 44 05 865 A1 It is, instead, by selectively reading only a few relevant portions of the sensor, such as a few rows or even only parts of lines, to greatly increase the read speed and thereby accelerate the calculation of the actual positions in real time with the movement of the robot arm, so that thereby the maximum movement speed of the robot arm can be de facto increased.

In addition, although the shows DE 102 46 449 A1 a sensor system with a lighting unit in two sensors, which are arranged with different viewing angles, however, this sensor system only serves to close by differences in the amount of light received at the two sensors on an irregularity on the irradiated surface.

Closer conclusions about the type and shape of the unevenness or even automatic measures with regard to the guidance of the sensor unit are not drawn from this.

Furthermore, the shows DE 10 2004 043 072 A1 a device for processing a workpiece with a guided relative to the workpiece laser head, which also includes a probe, are recorded from two different viewing directions image data of, for example, a pilot laser in front of the processing laser thrown onto the workpiece laser beam and depending on the machining head in all three spatial directions is guided to the distinctive path on the object in the desired target path.

III. Presentation of the invention

a) Technical task

Thus, the object according to the invention is to provide a method by means of which the setup times for the work can be reduced and in particular the positioning can be completely avoided without sacrificing quality in the work process.

b) Solution of the task

This object is solved by the features of claims 1 and 2. Advantageous embodiments will be apparent from the dependent claims.

The arrangement of an optical probe on the robot head, which from the actual position of the robot head is able to determine the corresponding part of the distinctive path in terms of its contour and position, in real time, can from a starting point - the manual or can be approached automatically coarse - the robot head immediately be guided in a work cycle along the distinctive path, without a prior positioning run must be performed.

Of course, if the machining of the salient web is not merely a visual check of that web, then in addition to the optical probe, a corresponding working head or tool must, of course, be attached to the robot head.

It goes without saying that both the relative position of the optical probe and possibly a working head or tool to the head of the robot arm (robot head) carrying it must be known.

However, if this is the case, the optical test head - and thus also the robot head, as well as a possibly existing working head - can travel along a nominal virtual path lying parallel to the marked path and, above all, reliably within a desired corridor defined around this theoretical path be guided by the starting point of the starting point from the current actual position, the position of the next piece of the target path and thus the target corridor is determined in real time and the probe along this target path and within the target corridor is tracked and is maintained or post-corrected in a given work position (with regard to the angular deviations about the longitudinal and transverse directions of the desired path).

It is assumed that the distinctive path or line within a certain distance in the longitudinal direction, starting from the current actual position, can only make a limited change in the longitudinal and transverse direction, so can not make abrupt changes in one of these directions. This is an acceptable condition due to the roughly known shape of the prominent path or line, especially if the prominent path or line is defined in terms of its start and end point in so far as such an abrupt change in the course between the start and end point is not given.

The approach of the starting point takes place transversely to the longitudinal direction of the target path, in such a transverse direction, in which no collision neither of the probe or the robot head with the robotic arm carrying him nor the working head possibly attached thereto can take place with the object.

The position of the desired corridor including the desired path and the working position at the actual position within the setpoint corridor for the test head is defined such that such collisions with the object can not occur.

The starting point can, in principle, be any point on the target path of the test head:
Either the test head is then guided from the starting point along the desired path to one end of the target path, and then moved back to the starting point and completed from there in the other direction to the other end of the work cycle.

Preferably, however, the starting point is selected at or near one end of the desired path, with the result that the test head is then guided from the starting point to the nearest end of the desired path as the first section of the work cycle - if the starting point is not the one anyway Endpoint of the target lane was - and then has only a short path back to the starting point, and then to complete the second part, the main part of the work process to the other endpoint of the target lane.

In this case, the distinctive path or line is scanned by the optical probe during the operation, preferably by means of the triangulation light-section method in which a transverse to the longitudinal direction of the web light line is irradiated on the distinctive path and the reflected images of two different optical sensors with different viewing directions are aligned to the surface of the distinctive path.

The different viewing angles of the two optical sensors are, in particular, different skew angles of the viewing direction of the two optical sensors either relative to the irradiation direction through the one common light source or also relative to the prominent path, in particular viewed in the side view on the longitudinal prominent path.

It essentially does not matter whether the two images are produced by projecting a single line of light onto the prominent path and recording it by two optical sensors aligned with different viewing angles and thus producing two images, or by means of only one optical sensor recorded two different optical images be radiated by two different directions of illumination the same light line on the distinctive path.

The following statements always assume the case of a single irradiation direction and the acquisition of two different images by means of optical sensors directed at the irradiation point with different viewing directions.

In this case, when moving the test head along the prominent path, the individual test positions in which images of the illuminated light lines are recorded preferably lie so close together that their spacing is less than the width of the illuminated light lines, so that the successively lined test positions and the thereby produced images of the light lines slightly overlap each other and thus give a continuous image of the distinctive trajectory. This will be especially the case when the distinctive track is to be subjected to a total visual inspection, z. B. a shape determination that can be carried out together in this way together with the orientation of the distinctive web in a single run. If, on the other hand, only a position determination is to take place, on the basis of which then along the distinctive path only at relatively long intervals, for. B. spot welds to be performed by means of the working head, such a close adjacent arrangement of the individual light lines, so test positions, usually not necessary, which increases the speed of the work process.

However, if a continuous optical inspection of the distinctive trajectory is to take place, the time interval between the individual recordings, ie test positions, in relation to the relative speed between the test head and target path is determined so that the longitudinal distance between the centers of two light lines radiated is greater than that Line width of this light line, in particular 1.0 to 3.0 times as large as the line width of the light line, and this longitudinal distance between two lines of light is chosen to be lower, the stronger the curvature of the distinctive path.

The tracking of the probe along the target path, ie within the target corridor, is carried out according to the following criteria:
In terms of the image of the light line radiated across it, the distinctive path with its distinctive cross-section also produces an image with a striking shape. The striking part of the image of the light line should lie within a predetermined target area on the optical sensor, which must be the case with correct setting of the working position of the probe at the starting point of the target path.

As soon as a movement of the probe along the target path, a point of the prominent part of the image generated on the optical sensor moves laterally out of the predetermined target area, a tracking of the test head in the transverse direction to the target path in such a direction and in such a direction Measured performed that the entire significant part of the image is within the specified target area on the optical sensor.

The height of the test head over the significant path to be scanned, ie the working height, is readjusted automatically when the highest or the lowest point of the marked part of the image on the sensor surfaces runs out of a predetermined target area in the vertical direction.

In this way, the test head is held within the desired corridor defined around the desired path.

In addition, the test head must be kept in a correct working position with respect to the angular position around the longitudinal and transverse axis of the target path around: For example, the transverse tilt angle of the probe, ie the tilt angle around the transverse direction, must be automatically corrected, if two Under the same viewing angle, the highest or lowest points of the striking part of the images of the imaged images in succession on the two sensors do not shift analogously on the two sensors.

If the illumination direction with which the image is irradiated is to be the perpendicular to the distinctive path, and the viewing directions of the two optical sensors are at an opposite equal angle to this illumination direction, this means that the illumination direction is no longer perpendicular to the distinctive path is directed and needs to be readjusted.

Another possibility is to compare the quantities of light received on the two sensors:
If the illumination direction is no longer directed exactly perpendicular to the striking path, different amounts of light will arrive on the two sensors because of the different long return path of the light, so that in turn can be carried out an automatic Nachkorrektur the transverse tilt angle until the received amount of light on both sensors is the same in each case, which suggests a cross-tilt angle of zero, ie a perpendicular irradiation on the distinctive path.

Such a light quantity comparison can also be carried out with separate light quantity sensors, for example photodiodes, PSDs or a split photodiode, instead of the existing optical sensors for recording the image.

A readjustment of the longitudinal tilt angle about the longitudinal direction then takes place on the basis of a height comparison of the ends of the image of a light line:
When the longitudinal tilt angle of the test head is properly adjusted, the two ends of the image of the light line on the optical sensor have a certain height difference corresponding to the target height difference that a reference image should leave on the sensor for that location of the significant web to be inspected.

If the actual height difference is different or lies outside a predetermined deviation range, the inclination of the test head is readjusted about the longitudinal axis until the actual height difference of the ends corresponds to the nominal height difference.

In addition, the actual distance between the test head and the marked track can be automatically determined based on the known triangulation angle (the angle between the viewing direction of the sensor and the irradiation direction), which causes a known position of the highest or lowest point of the image on the optical sensor. In the case of deviation of the position of the highest point from this known position, the difference between the actual distance of the test head from the marked track and the nominal distance can be calculated automatically from this.

Furthermore, it may be necessary to calibrate the measurement data, that is to say the images on the optical sensors, in order to enable the automatic conversion of these images into real values, ie from camera coordinates to world coordinates.

For this purpose, for example, with exactly maintained nominal distance of the probe to the distinctive path a Kalibrierbahn scanned in place of the distinctive trajectory, which has a precisely known contour. This results in a certain altitude of the image on the sensor, with the aid of a deviating position of the image of the distinctive trajectory to be scanned later allows the automatic calculation of the real values of this distinctive trajectory. For this it may already be sufficient to correlate only prominent points of the light line or the image of the light line using the known triangulation angle and the nominal distance between camera coordinates and world coordinates.

Another way to determine the deviation of the working position with respect to the longitudinal tilt angle and transverse tilt angle, is to radiate on the striking path two slightly spaced light lines and thus on each sensor resulting two images and their change in the relative position with progressive scans, thus progressive movement of the test head along the distinctive path, observe:
If the distance of the two images as a whole, for example, the distance between their highest and lowest points to each other, changes, this means that the illumination direction has deviated from the perpendicular to the longitudinal direction, ie the transverse tilt angle has deviated from the vertical, at least if Nominal distance was maintained correctly.

A deviation of the longitudinal tilt angle, that is, the angle about the longitudinal direction, on the other hand, exists when the relation of the distance of the left ends of the two images to the distance of the right ends of the two images changed or different on the two sensors

A different transverse offset of the two images relative to one another on the two sensors suggests a deviation from the desired longitudinal tilt angle and is used as a reason for a correction of the test head position with regard to the longitudinal tilt angle.

• – zunächst wird überprüft, ob die in den beiden Abbildern festgestellten Anomalien an der exakt gleichen Position liegen. Zu diesem Zweck werden die zugehörigen Erstreckungsvierecke bestimmt, nämlich die exakte Erstreckung in X- und Y-Richtung, in der das Abbild maximal diese Anomalie aufweist. Die Abweichung der Positionen der beiden Erstreckungs-Vierecke wird berechnet, insbesondere in Welt-Koordinaten, und wenn die Abweichung unter einem vorgegebenem Grenzwert liegt, wird von einer übereinstimmenden Position ausgegangen und damit von einer echten Anomalie.
• – Wenn das Abbild eine Lücke aufweist, wird der Höhenversatz von Beginn und Ende des linienförmigen Abbildes auf beiden Seiten der Lücke ermittelt und bei einem zu starken Höhenversatz oberhalb eines Grenzwertes daraus geschlossen, dass es sich um keine echte Unregelmäßigkeit handelt.
• – Es werden ferner der durchschnittliche Helligkeitswert des linienförmigen Abbildes, in dem die Anomalie festgestellt wurde mit den durchschnittlichen Helligkeitswerten der vom gleichen Sensor aufgenommenen, wenigstens einem vorangehenden und wenigstens einem nachfolgenden Abbild verglichen, oder mit mehreren vorangehenden/nachfolgenden Abbildern und bei Abweichung des Helligkeitswertes oberhalb eines Grenzwertes die Anomalie nicht als Unregelmäßigkeit der Prüfkontur gewertet.

If, in addition to the determination of the position of the prominent web, its shape is also to be optically checked for the presence of anomalies within the contour, for which the individual scans must then be connected without gaps, this check should be accelerated by additional measures:
In order to keep the evaluation effort low and thus the test speed, in particular a continuous relative movement between test head and test contour, high, is tested in several stages:
On the one hand, plausibility checks are carried out to determine whether the anomaly observed in the image is actually due to an irregularity of the test contour:

• - First, it checks whether the anomalies found in the two images are in exactly the same position. For this purpose, the associated quadrilaterals are determined, namely the exact extension in the X and Y directions, in which the image has at most this anomaly. The deviation of the positions of the two extension squares is calculated, in particular in world coordinates, and if the deviation is below a predetermined limit value, a corresponding position is assumed, and thus a genuine anomaly.
• - If the image has a gap, the height offset of the beginning and end of the line-shaped image is determined on both sides of the gap and if the height offset is too high above a limit, it is concluded that this is not a true irregularity.
• The average brightness value of the linear image in which the anomaly was detected is also compared with the average brightness values of the at least one preceding and at least one subsequent image recorded by the same sensor, or with a plurality of preceding / succeeding images and if the brightness value deviates above of a limit value, the anomaly is not considered an irregularity of the test contour.

In addition, and / or instead, the original scans of both optical sensors, or an averaging of these two scans can be saved as a continuous original image, which later facilitates the work, in particular the post-processing of the defects Finding the defects is facilitated, yes, for example, only have a diameter of 50 microns and thus are hardly visible to the human eye. If a comparison with predetermined absolute limit values is carried out during the first or second test step, these limit values must first be determined in a practical manner.

For this purpose, a manually positively tested reference contour is scanned along its entire length and its image data is stored as reference values, optionally subdivided into individual reference lengths, over which the reference values do not or only very slightly change.

• – Eine Maßnahme besteht darin, von den linienförmigen Abbildern nur den Bereich zu untersuchen, der der eigentlichen Prüfkontur in Querrichtung entspricht, was durch einen Vergleich mit den zuvor gescannten Referenzkonturen und deren Prüfdaten ermittelt wird.
• – Zu diesem Zweck wird ferner eine Nachführung des Prüfkopfes in Querrichtung durchgeführt, sobald der höchste bzw. tiefste Punkt des Abbildes der Prüfkontur seitlich aus dem vorgegebenen Zielbereich innerhalb des Sensors herausläuft.
• – Der Prüfkopf wird hinsichtlich der Höhe über der Prüfkontur auf eine Nennhöhe eingestellt und automatisch nachgeführt, wenn die höchsten bzw. tiefsten Punkte der Abbilder auf den Sensoren bzw. dem einzigen Sensor aus dem vorgegebenen Zielbereich herauslaufen.
• – Des Weiteren kann die Betrachtung jeder Prüfposition und unter jedem der beiden Beobachtungswinkel zweimal und zwar mit unterschiedlicher Lichtmenge durchgeführt werden, wobei die Lichtmenge entweder durch die Belichtungszeit oder die beaufschlagte Lichtmenge pro Zeiteinheit variiert werden kann. Wenn in den Abbildern eine Lücke oder andere Anomalie vorhanden ist, muss diese Anomalie in den mit den verschiedenen Lichtmengen angefertigten Abbildern jeweils vorhanden sein, wenn die Ursache eine Anomalie in der echten Prüfkontur darstellt. Ist dies nicht der Fall, wird nicht auf eine echte Fehlstelle geschlossen. Des Weiteren dienen die beiden Abbilder mit unterschiedlichen Lichtmengen dazu, daraus eine qualitativ hochwertiges Abbild zu erzeugen, beispielsweise dann, wenn bei dem Abbild mit niedrigerer Lichtmenge nur ein Teil des linienförmigen Abbildes gut sichtbar ist, die fehlenden Bereiche dagegen in dem Abbild mit der anderen Lichtmenge gut sichtbar sind. Die jeweils gut sichtbaren Bereiche können dann zusammen zu einem Abbild kombiniert werden.

Furthermore, measures are taken to minimize the evaluation effort and thus to be able to carry out a rapid evaluation, namely in real time, which in some applications means a speed of 250 mm test contour per second:

• One measure consists in examining only the region corresponding to the actual test contour in the transverse direction of the linear images, which is determined by a comparison with the previously scanned reference contours and their test data.
• - For this purpose, a tracking of the probe is further carried out in the transverse direction as soon as the highest or lowest point of the image of the test contour runs out laterally from the predetermined target area within the sensor.
• - The test head is set to a nominal height with respect to the height above the test contour and automatically tracked when the highest or lowest points of the images on the sensors or the single sensor run out of the specified target area.
• - Furthermore, the observation of each test position and under each of the two observation angles can be performed twice and with different amounts of light, the amount of light can be varied either by the exposure time or the amount of light applied per unit time. If a gap or other anomaly exists in the images, this anomaly must be present in each of the images produced with the different amounts of light if the cause is an anomaly in the true test contour. If this is not the case, it is not concluded that there is a real defect. Furthermore, the two images with different amounts of light serve to produce a high-quality image from it, for example, if only a part of the line-shaped image is clearly visible in the image with a lower amount of light, the missing areas in the image with the other amount of light are clearly visible. The clearly visible areas can then be combined together to form an image.

At least those test sites that gave a negative result in this way are stored in terms of the location of these test sites, both in terms of the original image data and any cross-sectional profiles generated.

Even after the end of the examination of an object, these data remain stored for the object, so that they are available to the post-processor during the subsequent manual post-processing of the object and facilitate its rework, in particular with regard to locating the flaw.

This procedure is carried out in real time during the continuous scanning of the test contour and within the test head, including both test steps and all exposure steps.

For this purpose, the test head contains not only the two (or preferably only one) light source, but also the one or the two required optical sensors, the image data memory and the electronic processing unit.

Preferably, the light beam of the light source passes at right angles from a side surface of the probe, while the two optical sensors in the direction of movement of the probe before and behind the position of the exiting Thus, in this case, in one case a positive, in the other case, a negative triangulation angle.

Preferably, the level spanned by the two observation directions defines precisely that plane in which the instantaneous direction of movement of the test head lies.

Furthermore, the method is insofar self-regulating, as a possible inclination of the irradiation direction, ie deviating from the exactly vertical irradiation to the test contour level of the test head or its control unit automatically based on the then non-analogous displacement of the highest and lowest points of the corresponding one Images of the two sensors or viewing angle is detected.

In this case, it is concluded from the magnitude of the unevenness of the displacement on the oblique position of the irradiation direction and these are taken into account mathematically in the evaluation or automatically corrected in an exactly vertical position.

c) embodiments

Embodiments according to the invention are described in more detail below by way of example. Show it:

1 : the entire object in the supervision,

1b -E: detailed representations of the distinctive path to be tested on the object,

2a C: the test head on the web,

3a -C: Reduction methods of computational effort

4 : an image of the light line on the sensor,

5a -C: another reduction method,

6 : the tracking of the test head,

7a d-: composing the image from several shots.

1a shows an object 1 , in this case, a car body, at the soldering seam 2 Namely, the soldering seam between body and roof of the bodywork, representing a striking trajectory, along which a robot head 26 to be moved by means of an optical probe 6 ,

The solder seam 2 is a three-dimensional path that is curved in at least two spatial directions.

Likewise, however, could also be the edge 32 on the body carcass as a distinctive line 2 serve.

To save the time for a separate positioning or other measures for the precise location determination of the body, the robot head 26 in which, in a defined assignment, at least the optical test head 6 is, without further prior measures, such as a positioning run or the like, in a work process along the solder seam 2 guided. The only measure before that is the automatic or manual approach of the robot head 26 with the test head 6 to a starting point 27 on the target path 24 for the robot head 6 ,

As the 1d and 2c show, the target path runs 24 along which the test head 6 , in particular with the exit point of its light beam 3 , to proceed, band-shaped and parallel to the distinctive path 2 in the form of the soldered seam and in a defined nominal distance 22 For this.

In addition, viewed in cross section ( 1d ) around the target path 24 around a target corridor enclosing annularly 28 defines the test head 6 - Or whose light exit point - may not leave, the location of both the target path 24 as well as the target corridor 28 as well as the additionally specified working position of the test head 6 and the whole robot head 26 including a possibly existing working head are set so that in compliance with these specifications no collision of the robot head 26 or the other heads attached to it, including the cantilever arm 29 with the object 1 can occur. This is possible because of the rough position of the object 1 relative to the robot head 26 basically known and given.

The essential difference to previous procedures is that the target path 24 and also the target corridor 28 not known from the beginning in their entire length, but successively with the progress of the probe 6 along its intended course 24 only in sections for the closer environment, in particular the nearest progress area, to the current actual position of the test head 6 to be recalculated:
For this purpose, by means of the optical probe 6 the striking railway 2 , in this case, the soldered seam 2 or the roof edge 32 , optically scanned and as a usable result not the location of the soldered seam 2 (relative to the test head 6 ) determines, but just the reverse of the relative position of the probe 6 in terms of this striking track 2 certainly:
As a longitudinal direction 10 should the course of the distinctive path 2 be defined as the transverse direction 11 their extension transverse to the longitudinal direction.

The striking line 2 is doing by means of z. B. the light-sliced triangulation method, as best seen in the 2a . 2 B and 1e understand:
2 shows a probe 6 in the side view, which also recognize how the well-known light-slicing triangulation method works in principle: This is a fan-shaped light beam 3 (please refer 2 B ) on the distinctive track 2 of the object 1 directed and generated there a direction transverse to the direction 10 this track 2 directed single light line 35 ,

The according to 2a from the surface of the object 1 and thus the distinctive railway 2 reflected light of the light beam 3 meets one of the two planar, optical sensors 12 . 12 ' in the test head 6 , and each creates an image there 4 . 4 ' on the surface of the sensor 12 . 12 ' , as in 3 and 4 shown.

Since the cross-sectional contour of the solder seam 2 as a distinctive train 2 is a hat-shaped, trough-shaped contour, also has the image 4 . 4 ' qualitatively in this form.

Thus, from these images 4 . 4 ' on the sensors 12 . 12 ' Conclusions on the position and possibly also the real contour of the surface of the object 1 at this test center 2 ' within the striking railway 2 to which the light beam 3 is currently directed, let go, allowed the direction of radiation 17 of the light beam 3 and the directions of view 5 . 5 ' the two optical detector units 6a , b, to which the sensors 12 . 12 ' belong, do not coincide, but must differ by a triangulation angle α ₁ , α ₂ .

In the present case these two angles α _1, α ₂ are chosen so that they are the same size, but oppositely directed from the emission 17 lie. That the reflected light rays still through mirrors 19 to be diverted is irrelevant.

The optical sensors 12 . 12 ' are to a computing unit 36 whose signals can be processed in real time and, above all, in real time, the further movement of the test head in real time 6 can correct:
2c shows the distinctive railway 2 , that - like from the 1a -E becomes clear - has a band-like shape, and the parallel running target path 24 along which the test head 6 with its viewing plane 20 , So for example, the underside of the probe 6 from which light emerges at right angles, as best in 2a apparent - should be led, and in a defined work situation 25 , in which the target path 24 tangential to the viewing plane 20 touched in the light exit point.

As the course of the distinctive train 2 is not yet known in detail, is at the beginning of the scan and the target path 24 not yet known in its course, nor the set corridor defined around it 28 ,

In 2c is the test head 6 positioned at any starting point 27 represented, for example, defined by the fact that he is opposite the distinctive path 2 the nominal distance 22 , measured along the perpendicular 21 on the train 2 , owns and the direction of radiation 17 identical to the vertical 21 on the train 2 is.

As well as you could the starting point 24a the target path 24 as the starting point, in the same way the beginning 2a the train 2 assigned.

Moving from this starting point 24a out of the test head 6 continue, for lack of another direction towards the viewing plane 20 and in the direction of the web 2 So this is the actual train 34 of the test head 6 more and more of the parallel to the distinctive railway 2 running target path 24 remove and eventually the target corridor 28 want to leave.

However, this is avoided by the movement of the probe 6 along the actual railway 34 successively at intervals on the train 2 projected light lines 35 - as in 1e represented - each images 4 . 4 ' on the sensors 12 . 12 ' whose changes are the departure of the actual train 34 from the target train 24 Show and a correction of the actual train 34 - For example, when reaching the upper limit of the target corridor 28 down to the lower limit of this target corridor 28 - cause:
6 shows in the left half of the picture, as the images 4 . 4 ' on the two with respect to the direction of irradiation 17 oppositely arranged sensors 12 . 12 ' look, namely on the one hand with a bulge 4a and on the other hand with a depression 4b in the middle area, resulting in the highest and lowest point corresponding top points 37 . 37 ' in the pictures 4 and 4 ' reveal.

Once one of these top points 37 . 37 ' a predetermined target area 38 . 38 ' on the respective sensor 12 . 12 ' on one of the next scans ( 6 right hand half), this means that - if the surface of the object was still illuminated perpendicularly - the distance of the probe 6 from the distinctive railway 2 has increased or decreased.

Would be in 4 the top of each sensor 12 . 12 ' also the upper edge of this sensor in the 2a That would be a shift of the top spot 37 in 6 left upper direction upper edge - as shown - according to 2a mean that the test head to the web 2 and in response to that, the arithmetic unit would have to 36 another removal of the probe 6 from the distinctive railway 2 in z. B. Defined increments, until the top point 37 again within the target area 38 located.

Since the increments of the distance steps performed are known at a known original distance of the probe - usually the nominal distance 22 at the starting point - also the current actual distance 33 known.

Furthermore, based on the images 4 . 4 ' in their sequence of successive scans Sx, Sx + 1, etc. are detected, whether the images 4 . 4 ' in an inadmissible way not analog, but unevenly on the two sensors 12 . 12 ' relocate.

For example, if the images 4 ' on the in 6 below shown sensor 12 ' move more towards the lower edge of the sensor than the images 4 of the sensor shown above 12 shift in the direction of the upper edge, so it benefits that the direction of radiation 17 from the light source 14 in the test head 6 no longer exactly perpendicular to the surface of the object 2 in the current checkpoint 2 ' radiates, and the probe 6 around the transverse direction 11 must be pivoted until this is given again, which is due to the uniform displacement of the images 4 . 4 ' can be recognized up or down.

Back to 2c this means that starting from any starting point, for example the endpoint 24a the target path 24 , out, the test head 6 in a defined work position within the target corridor 28 and along the distinctive path 2 is moved without this target corridor 28 or the target path 24 were previously known in their entire length, by starting from the starting point in the direction of the target path always only a next small longitudinal section of this target corridor 28 or from this or vice versa the target path 24 is determined.

However, once in this way, the test head over the entire length within the desired corridor 28 and thus along the distinctive path 2 has moved, is the location of the distinctive railway 2 accurately measured and known, although the probe 6 strictly speaking, not exactly to the distinctive path 2 parallel target path 24 followed, but a deviating actual train 34 but from every point of this actual train 34 from the relative position of the associated test site 2 ' on the train 2 is known.

From the known location of the distinctive railway 2 is thus indirectly also the overall position of the whole object 1 known, so in addition to the orientation of the train 2 with the help of one on the same robot head 26 fixed working head or with the help of another robot arm processing steps on the object 1 can be performed - depending on whether this is close to the distinctive railway 2 or further away from it.

Likewise, the processing to be performed, the contour measurement of the distinctive path 2 itself, for example, within this contour anomalies in the form of a depression 2a or a survey 2 B to detect how they are in the 1e such as 2a , b are shown. In this case, of course, this is done by means of the same probe 6 performing the orientation, and no additional working head is necessary.

In order to reduce the computational effort in the evaluation of the signals that are supplied by the optical sensors, are based on the 3 Reduction methods explained:
Many reduction methods are based on examining non-random contours, but expecting certain contours, as in the example of 3a a bulge 4a in an otherwise essentially linear straight image 4 in the form of a light line or a light band, the sensor 12a and a depression 4b at the sensor 12b according to 3b ,

The dimensions of the image of this bulge must 4a /Deepening 4b be determined and from this the actual dimensions of the recess 2a on the object 1 be calculated.

In 3a , b are two such images 4a , b on the surface of the respective optical sensor 12 . 12 ' shown.

The image 4 is that where the reflected line of light on the surface of the sensor 12 . 12 ' incident, the corresponding grid-like in rows R1, R2 .. and rows Z1, Z2 .. arranged pixels P1.1, P1.2 .. P2.1. P2.2 .. of the sensor 12 . 12 ' are more illuminated, so have a higher brightness value than those pixels on which the image, that of the object 1 reflected light beam 3 , does not hit. In the graphic black-and-white representation, this is shown inversely, namely the image 4 is shown dark in the form of the light line.

The pixels within the image 4 also have different brightness values, usually at the edge of the band-shaped image 4 with a lower and in the middle with a higher brightness value, which is not always the case and the reflection behavior of the test center 2 ' depends.

A first reduction method is that - because only the survey 4a /Deepening 4b interested - already when reading the image data from the image data memory of the sensor 12 only the records of those pixels are read out and processed further, in the interesting, mostly rectangular, area 40 which the bulge 4a shows, lie. This area of interest 40 is determined by the image data of all pixels of the sensor at the first scan S1 12 be read out and then on the basis of the determined course of the image 4 the location of the bulge 4a /Deepening 4b is determined according to width and height.

The next scan becomes just this area of interest 40 plus a surrounding security area determined by its size 41 as a new area of interest 40 read.

Also below is checked either at each scan or after a specified number of scans, as in this scan by the location of the bulge 4a /Deepening 4b defined area of interest 40 and the next or next scans will again become the newly specified area of interest 40 including a surrounding security area 41 considered, so read.

A second reduction method lies in the data reduction by computationally reducing the light band of the image, which in itself is over several pixels wide 4 ' , z. B. again only in the area of interest 40 , to the width of only one pixel.

So has the image 4 ( 3c ) in the form of a line of light, for example in series 17 a width of five pixels, in which the pixels P30.17 to P34.17 have an increased brightness value in this row.

Since primarily the course of the band-shaped image 4 and not interested in its breadth, it is sufficient the course of the image 4 in an extension direction 8th of the light band successive sequence of individual pixels, so to reduce to a line with a width of one pixel.

This can be done in several ways, eg. B. by at each longitudinal position along the extension direction 8th of the image, in this case in each row, of the pixels with the increased brightness value, the one with the highest brightness value is selected as the relevant pixel or the pixel in the middle of the illuminated pixels, in this case P32.17 from P30.17 to P34. 17th

Because the brightness distribution across the band-shaped image 4 not always even, z. B. not necessarily gaussian, must be, a weighting in terms of brightness value and therefore the focus of the width b of the image 4 applied brightness values are determined.

As a result, the band-shaped image is reduced to a sequence of individual profile points in the direction of extent 8th of the picture in a row.

A third reduction method is the addition of the survey 4a straight running areas of the image 4 parallel to the row or row orientation of the sensor 12 to bring and thus a position of the image as in the image 4 to achieve while in practice, the curve is at an angle. This will be without proper physical alignment of the sensor 12 to the base area 22 purely mathematically by "straightening", ie a normalization of the lying obliquely in the sensor surface image 4 to a straight lying image 4 , carried out.

For this purpose, the height offset of the beginning and end of the slanted image 4 on the sensor surface 12 - Preferably after a reduction of the image 4 on a sequence of individual profile points - proportionately deducted from the respective profile points:
While the z. B. low-lying left end point remains unchanged, all further right profile points are calculated lower by such a share of the height offset, the distance of the respective profile point of the unaltered endpoint in relation to the total distance from the left to the right endpoint, so the width of the sensor surface 12 , corresponds.

The better the extension direction 8th of the image 4 coincides with the direction of the lines or columns, the more so can be dispensed with this arithmetic straightening, if it only on the determination of geometric properties of the increase 4a arrives.

A further reduction of the computational effort can be achieved by deciding from the beginning whether it makes more sense, the extension direction 8th of the image 4 parallel to the row orientation or parallel to the row orientation of the sensor 12 to lay:
It plays a role that the internal structure of most optical sensors 12 are designed so that the data can be read out only in the form of a complete row of pixels and not just selected pixels from a line. An immediate column-wise, so row by row, readout of data is not possible with most sensors, but this is done indirectly by the row-wise read the corresponding pixels of the same row are sequentially read out.

If the expected contour, so for example the bulge 4a relative to the surface of the sensor 12 probably only a limited amount of z. B. is less than half, in particular less than a third of the length of the sensor surface will be an orientation of the extension direction 8th , ie the direction of the straight portions of the image 4 , are chosen parallel to the row direction, as in 3a shown.

On the other hand, it is expected that the bulge 4a anyway a larger part of the length of the sensor 12 will occupy or even more than the length of the usually in wide format, ie with row direction along the larger extent, present sensor surface 12 , so the extension direction 8th - as in 4 shown - parallel to the row direction and thus selected at right angles to the row orientation.

In the latter case, the data reduction is then preferably also carried out by the following method ( 4 ):
It tries to turn only the area of interest 40 ' of the image 4 evaluate. In this case, however, will not - as in 3 shown - a rectangular area evaluated in which the entire bulge 4a Place finds, but in this case are only the areas of interest 40 ' on pixels or their data records from each line further processed.

Also in this case, all pixels are read out of the image data memory of the sensor during the first scan, at least in the first line, the areas of interest 40 ' to determine this line. Analogous to the procedure of the two-dimensional region of interest, as based on the 3 then becomes the newly interesting area 40 ' the next line corresponding to that of the previous line, plus an adjoining security area 41 ' at both ends, which in turn is preferably fixed by a number of pixels.

The determination of the current area of interest is either carried out anew in each line or in a sequence of a few lines each.

The records of the areas of interest from the image data memory 9 whose records are organized in rows and columns like the pixels of the sensor 12 , are transferred in the computational further processing in a buffer 42 , This is usually mandatory because from the image data storage 9 - As mentioned above - in most sensor designs anyway only a readout of the data sets in whole lines and not only for some pixels of a line is possible.

Therefore, for all lines, the data records of the respective entire line are saved from the image data 9 - as in 4 and 5a displayed - read, but only the area of interest 40 ' in a cache 42 ( 5b ) transfer.

If it is a type of sensor that allows readout of given pixel positions within a line, such as a CMOS sensor, justify the following data reduction according to 5 nevertheless the transfer to the cache 42 :
The data set for a pixel consists of at least three data, namely on the one hand the pixel position (x and y value in the sensor) and on the other hand at least the brightness value (H (x; y) = gray value), that pixel on the sensor 12 possessed after the exposure.

As the area of interest in each line 40 ' each consists of an immediately successive sequence of pixels within this line, it is sufficient for the location information, in a first line of the cache 42 ( 5b ) each hold the position that had the first pixel with increased brightness value within a line. In the fields of the cache 42 Below this position specification, only the brightness values of this first and the following pixels with an increased brightness value within the respective region of interest are then still present, eg. B. the same line, in the cache 42 entered.

In contrast to the number of pixels and thus the number of data records of the image data memory 9 is thus the data volume of the buffer 42 on the one hand, because the number of data records has been drastically reduced (namely to the number of pixels with increased brightness value) and, on the other hand, the amount of data is reduced by virtue of each data record having at least one data item reduced by the position specification, for example only one third of the original data item, has.

From the information of a series R1, R2 .. from the cache 42 , which consists of the position specification (x; y) and the following brightness values H (x; y), can then be determined for the cross-section of the image, for example according to the method of determining the center of gravity 4 Calculate a single profile point in each line, so that this sequence of individual profile points the entire band-shaped image 4 represents for the further determination of the desired geometric data concerning the bulge 4a ,

7 shows how sketchy images 4 on a sensor 12 to complete images can be added:
Gap-like images 4 on a sensor 12 . 12 ' often result from reflections on the surface of the object 1 that cause the image 4 on the sensor 12 overexposed or underexposed, and thus pixels in a region of the image with too high or too low brightness value occur and thus a gap 39 in the image 4 form.

If, for example, according to 7a in the image 4 a gap 39 exists, then becomes the same light line 35 from the same position again from the sensor 12 sampled, but with a different exposure amount on the sensor, whether by changing the exposure time, changed aperture or lower irradiation power. In general, with low exposure of the sensor 12 started and gradually increased this exposure.

At some point, this will be exactly in the area of the gap 39 a continuous image 4 * which, however, may have a gap elsewhere in the course of the image, which, however, is insignificant.

At least in the area of the gap 39 will now be the area of interest according to 7b taken from the image and in the area of the gap 39 used, eliminating this gap at least on the size 39 ' is reduced, as in 7c shown.

This process can be repeated with additional exposure levels until the gap 39 completely in the image 4 is closed as in 7d shown.

This can be done in particular by means of two photodiodes, which - spaced in the direction of extension 10 the distinctive railway 2 at the test head 6 arranged - measure the amount of each incident light from the radiated line of light and then the transverse tilt angle 31 of the test head 6 around the transverse axis 11 is readjusted in the direction of alignment of these amounts of light.

LIST OF REFERENCE NUMBERS

1

object

2

Solder seam, distinctive web or line

2 '

checkpoint

2a

deepening

2 B

survey

3

Beam of light, direction of illumination

4

image

4a

upheaval

4b

deepening

5, 5 '

line of sight

6

probe

6a, b

detector unit

7

movement direction

8th

transverse axis

9

Image data storage

10

longitudinal direction

11

transversely

12, 12 '

optical sensor

13a, b

parallel output

14

light source

15

Internal memory

16

casing

17

radiation direction

18

observation width

19

mirror

20

viewing plane

21

vertical

22

Nominal distance

23

passage

24

Set path

25

working position

26

robot head

27

starting point

28

Target corridor

29

Auslegearm

30

Longitudinal tilt

31

Cross-Dump

32

edge

33

Actual distance

34

Actual path

35

light line

36

computer unit

37

Top point

38

target area

39

gap

40

interesting area

41

security area

42

cache

Claims (10)

Method for three-dimensional path guidance of a robot head ( 26 ) with an optical test head fastened thereto in a defined relative position ( 6 ) on an object ( 1 ) along a desired path ( 24 ) in a defined work situation ( 25 ) over a visually distinctive web ( 2 ) with the following steps: - optical scanning of the distinctive path ( 2 ) on the object ( 1 ) through the test head ( 6 ) by means of the triangulation light-slit method, - evaluation of the signals determined at the actual position in real time, - constant readjustment of the actual path ( 34 ) of the robot head ( 26 ) in the direction of the desired course ( 24 ) to the test head ( 6 ) - both parallel to the longitudinal direction ( 10 ) as well as parallel to the transverse direction ( 11 ) the target path ( 24 ) and thus the distinctive railway ( 2 ), - at nominal distance ( 22 ) between the distinctive path ( 2 ) and the actual lane ( 34 ) and - in a target corridor ( 28 ) around the predetermined target path ( 24 ), characterized in that - that of the distinctive path ( 2 ) reflected, line-shaped images ( 4 . 4 ' ) of two different optical sensors ( 12 . 12 ' ) with the direction of illumination ( 3 ) opposite in the same direction ( 5 . 5 ' ) and - if the quantity of light received by the two sensors is not sufficient ( 12 . 12 ' ) with the same viewing angle on a non-perpendicular irradiation direction ( 3 ) and this deviation of the direction of irradiation ( 3 ) of the perpendicular ( 21 ) to the longitudinal direction ( 10 ) of the distinctive railway ( 2 ) is automatically corrected. Method for three-dimensional path guidance of a robot head ( 26 ) with an optical test head fastened thereto in a defined relative position ( 6 ) on an object ( 1 ) along a desired path ( 24 ) in a defined work situation ( 25 ) over a visually distinctive web ( 2 ) comprising the following steps: - optically scanning the distinctive path ( 2 ) on the object ( 1 ) through the test head ( 6 ) by means of the triangulation light-slit method, - evaluation of the signals determined at the actual position in real time, - constant readjustment of the actual path ( 34 ) of the robot head ( 26 ) in the direction of the desired course ( 24 ) to the test head ( 6 ) - both parallel to the longitudinal direction ( 10 ) as well as parallel to the transverse direction ( 11 ) the target path ( 24 ) and thus the distinctive railway ( 2 ), - at nominal distance ( 22 ) between the distinctive path ( 2 ) and the actual lane ( 34 ), and - the test head ( 6 ) in a target corridor ( 28 ) around the predetermined target path ( 24 ), characterized in that the parallelism of the test head ( 6 ) to the longitudinal or transverse axis ( 10 or 11 ) the target path ( 24 ) is checked by two light quantity measuring devices, in particular two photodiodes, a PSD or a splintered photodiode - spaced in the direction of the axis to be tested - are arranged on the sensor unit and measure the amount of reflected light and uneven received light quantity on lack of parallelism to this axis closed and readjusted in the direction of equalizing the amount of light. Method according to Claim 2, characterized in that the quantities of light received by the two light quantity measuring devices, in particular photodiodes, are subtracted from each other and referenced to their sum, and the skew to the axis to be checked is adjusted so that the value obtained becomes minimal. Method according to one of the preceding claims, characterized in that the parallelism of the test head ( 6 ) to the transverse axis ( 11 ) and / or to the longitudinal axis ( 10 ) by two lines of light and thus two images per optical sensor ( 12 . 12 ' ) is checked by - from unequal distances between the two images on the two sensors ( 12 . 12 ' ) to a deviation of the transverse tilt angle about the transverse axis ( 11 ) is closed when the distance between the two images as a whole, in particular the distance of their highest and lowest points to each other on the two optical sensors is different or changed, - from a lateral offset of the two images on an optical sensor ( 12 ) to a deviation of the longitudinal tilt angle ( 30 ), or - from a change in the relation of the distance of the left ends of the two images ( 4 ) to the distance of the right ends of the two images ( 4 ) on the sensor ( 12 . 12 ' ) is closed to change the Längskippwinkels. Method according to one of the preceding claims, characterized in that the test head ( 6 ) above the distinctive path ( 2 ) is automatically readjusted at working height if the highest or lowest points of the corresponding images ( 4 . 4 ' ) on the sensor surface, ( 12 . 12 ' ) or the single sensor from a predetermined target area run out. Method according to one of the preceding claims, characterized in that the time interval between the individual recordings in relation to the relative speed between test head ( 6 ) and target path ( 24 ) is set so that the longitudinal distance between the centers of two images is greater than the line width of the image ( 4 . 4 ' ), in particular 1.0 to 3.0 times as large as the line width of the image ( 4 . 4 ' ), and the longitudinal distance between two receptacles is chosen to be lower, the stronger the curvature of the desired path (FIG. 24 ) and thus the distinctive railway ( 2 ). Method according to one of the preceding claims, characterized in that the actual distance ( 33 ) between test head ( 6 ) and distinctive railway ( 2 ) is calculated automatically on the basis of the known position of the highest or lowest point of the image ( 4 . 4 ' ) on the optical sensor ( 12 . 12 ' ) at a certain triangulation angle and at a different altitude above the actual distance ( 33 ) and calculated at the nominal distance ( 22 ) is corrected. Method according to one of the preceding claims, characterized in that a calibration of the measured data, in particular the contours of the images, by means of the nominal distance ( 22 ) between test head ( 6 ) and distinctive path ( 2 ) and scanning a calibration contour with exactly known contour and distance, so that the conversion of the determined contours of the images ( 4 . 4 ' ) is automatically possible in real contours. Method according to one of the preceding claims, characterized in that during calibration, the conversion of the camera coordinates into world coordinates takes place by striking points of the marked path ( 2 ) and their two images ( 4 . 4 ' ), using the known viewing angles and the nominal distance ( 22 ) are correlated with respect to the camera coordinates and the world coordinates. Method according to one of the preceding claims, characterized in that in the case of a deviation of the height differences of the two ends of the image ( 4 . 4 ' ) of a light line on the optical sensor ( 12 . 12 ' ) compared to the nominal height difference of a reference image of an impermissible pivoting of the test head ( 6 ) around the course direction ( 10 ) the target path ( 24 ) and this is corrected automatically

Images