NO753068L - - Google Patents

Description

ELECTRO-OPTICAL SCANNING SYSTEM FOR DIMENSIONAL MEASUREMENT

OF PARTS

The present invention relates to an electro-optical system

for measuring the dimensions of individual parts or assemblies of such.

Modern manufacturing technology makes it possible to manufacture and assemble parts at an extremely high speed, even up to thousands of units per minute. However, there is no commercially viable technique capable of automatically and accurately measuring the dimensions of manufactured parts at the same rapid rate to determine whether they meet all dimensional tolerances. Common practice is therefore not to individually measure each part or assemblies of such, but to inspect samples taken from the production line. However, available inspection rates are insufficient and do not support a meaningful inspection program when the unit production rate is very high. Even when the sampling speed is satisfactory, •it provides no assurance that each and every part satisfies all dimensional requirements. In situations where the manufactured part is intended to perform a critical function, individual inspection is an absolute requirement.

Common devices for measuring manufactured parts, such as micrometers, height gauges and dial indicators all involve physical contact with the part. The painting procedure requires a trained operator

and is often difficult and time-consuming. The widely used optical comparator technique in which a silhouette of the part is projected

on an enlarged scale for a reference screen also requires a trained operator and is prone to measurement errors due to .eye fatigue, poor judgment and other factors. The measurement speed of an optical comparator is limited by the operator's ability to see and judge the image

on the screen with respect to a reference contour.

To overcome the limitations of measurement procedures involving human operators, contactless electronic devices have been provided, many of which employ an electro-optical device such as a vidicon rudder or an image dissector rudder. These tubes include a sensitive cathode and are adapted to scan an image of the object projected onto it to produce a video output representing the dimensions of the object being scanned. Among the advantages achieved by using electronic scanning systems are the following:

A Dimensions can be obtained at a point that is far away

from the part being measured without the need for physical contact with it.

B Measurements can be made to determine part dimensions which

are not easily obtainable.

C Individual dimensions can be measured at very high speeds.

D Because there is no need for a physical shift to advance measurements at different coordinate positions on a single device, there is effectively no dead time between such measurements. The time it takes to switch from

one coordinate position to the next is limited only by the reaction time of the electronic detector system and is usually in the order of a few microseconds.

E The measurement precision, which is expressed as a percentage of full scale, is independent of the size of the part being inspected. Compensation for variations in the receptive object size is carried out simply by changing the magnification in the optical system that provides an image of the object tii. the detector.

Despite the advantages inherent in electronic scanning systems

for the measurement of machined parts and other manufactured objects, electronic scanning systems of the previously known type have not had significant commercial success, as such systems cannot

can be relied upon to make accurate measurements. A main reason for this lack of success is that existing systems are not able to compensate for distortion that takes place in the electronic circuits of the system and for non-linearities in the sweep and in the optics.

In an ideal scanning system in which there is zero distortion

in the optics and in the scanning parameters, and in which time is infinitely divisible without any ambiguity due to system noise and other variables, the system is theoretically capable of measuring the desired dimension with any desired degree of accuracy or resolution. But as a current system is inevitably exposed to various distortions and ambiguities, noise and quantizing errors, the measurement accuracy is worsened by these factors.

In US Patent No. 3,854,822 an electro-optical system is discussed

in which these distortions and ambiguities are substantially eliminated, whereby the system is able to measure the dimensions of parts with a high degree of accuracy. The present invention includes the features discussed in said patent, although it provides a somewhat different approach to the same problem. In the aforementioned patent, the measurement is carried out by means of an electronic gap measurement technique where the electronic scanning acts to determine the distance between opposite edges of the parts in a manner analogous to that of a two-jaw mechanical fit whose jaws engage with opposite edges. In the present invention, use is made of a coordinate-graphic technique whereby the detector examines the vicinity of the edges whose coordinates are to be determined.

In view of the foregoing, it is a main object of the present invention to provide an improved electro-optical scanning system for measuring by means of a coordinate-graphic technique the dimensions of parts or assemblies thereof.

More specifically, the purpose of the present invention is to provide a measurement system of the above-mentioned type which allows for a greater degree of statistical averaging of the video information that is collected without sacrificing the operating speed of the system. In a scanning technique where the scanning is used as a measuring fit, the scanning must take place over a large part of the image in which no metrological data exists. By eliminating the need to scan areas that lack information, it becomes possible to reduce the time required for the individual scan by one or two orders of magnitude without reducing the information content, thus a system that entails a scanning technique whose information density is high will lead to numerous repetitions of the scan and allows the integration of data generated by statistical averaging techniques to reduce system noise without increasing the operational time required for the measurements. In the coordinate graph solution according to the invention, the detector is directed to only see near the edges whose coordinates are to be. is determined. In a typical case, the edge tolerance is of the order of some

few thousandths of a centimeter over a dimension that is usually between 0.25 and 1.25 cm. With the previously known passer measurement solution, the scanning will be forced to extend over the entire width of the dimension of interest, i.e. the entire length 0.25 or 1.25 cm.

Thus, the scan in the coordinate graph technique only needs to proceed over the region where the edge is expected to be found, which region is normally no larger than<+>0.0125 cm. (Obviously, if the edge is not found within this region, the part may contain a significant error.) Instead of progressing .100-500 units of' every 0.0025 cm to make a passer measurement, it is done in the coordinate graph technique only an overlap of 10 of these same increments, which reduces the required sweep time by a factor ranging from 10 to 50.

It will be understood that as the ratio of the permitted tolerance is reduced with respect to the dimension being measured, the effective efficiency of the coordinate system will be improved accordingly. As a result, the coordinator-type system can repeat each measurement 10 or 50 or even more times (depending on the ratio of improved efficiency) and then use the integrated results of all these measurements to obtain a signal for noise improvement in the system operation and thereby significantly increase system accuracy. In practice, the coordinate graph solution, without increasing the measurement time, is able to improve the resolution from 3X to 10X compared to the passer solution and therefore provides a far greater degree of measurement accuracy.

It is a further object of the present invention

to provide a "permanent" method of calibration which in many cases will eliminate the need for a two-dimensional calibration network of the type discussed in said patent, as an operative element in the measurement procedure for the system. This greatly simplifies the optical design required for the system over that required for an in-use network calibration procedure, and it also accelerates the effective system measurement speed by a factor of two.

It is also a matter of the present invention to provide a system of the above-mentioned type where the measurements carried out for each measurement are individually controlled by a stored digital program whereby under the command of the program the measurements can be carried out anywhere within the field of view in the system.

Within this field of view, no limitation exists with respect to the length of the dimension to be measured or its slope with respect to a randomly determined set of coordinates.

In a system according to the invention, the digital program stores for each part the nominal coordinates for the dimension to be measured and the acceptable tolerance range on this dimension. The program determines whether the measurement system will operate on a straightforward "accept" or "no-accept" basis to accept or reject the drill, or whether the system will sort parts into separate categories that depend on tolerance bands within which the controlling dimensions fall.

The number of dimensions that can be measured on a given part is limited only by the capacity of the memory that holds the program. When. multiple dimensions are measured on a single piece, the machine can be programmed to accept the piece when all dimensions are within the predetermined tolerances, or by rejecting the piece for an out-of-tolerance dimension, to give a signal indicating which dimension was found to be unacceptable. This signal can be used to sort parts into separate bins according to the dimension found to be out of tolerance.

the robbery.

By using stored sub-routines, the measurement system can be made to perform standard multiple measurements of the type suitable for the inspection and measurement of machine or otherwise manufactured parts or assemblies. thus, with the help of a stored sub-routine, one is able to find the coordinates and diameter of all holes that are visible in the field of view of the system.'With another sub-routine one can determine the roundness of each of these holes. Agree the ice of the holes with regard to any other desired

shape whose perfect dimensions can be established as a reference

can be determined by yet another sub-routine. With additional sub-routines one can determine the concentricity of parts, the pitch diameter for gears or helical pieces as well as the tooth-to-tooth error of most hobbed parts.

Due to the flexibility of the electronic search can

the electro-optical measuring system is programmed to scan monsters that are optimized for special measuring functions, such as determining the radius of curvature and the center of curvature of curved surfaces. For this purpose, the system can establish the position of these curvature centers in space and make measurements from these positions to other points within the system's field of vision.

To establish the system, the system's basic measurement unit is necessary

to induce a stable high-precision relationship or transfer function between the voltage used to control the scanning generator and the relevant instantaneous position of the scanning with respect to the image plane. With this relationship established, the instantaneous position of the scan is then determined by basically just examining the value of the voltage fed to the scan generator. Using this technique, the following of a video edge (indicating that the scan crosses an image edge at the image plane) will command a readout from the circuit that generates the control voltages for the scan generator. With correct control of the transfer function as described above, these control voltages precisely determine the coordinates for the crossing point at the image edge in the system's image plane. These coordinates are then the basic measurement units for the coordinate graph detector.

The system can be used to perform several functions, and as an example, the following two functions can be performed simultaneously or independently: A Perform on-line, real time determinations for acceptance,

rejection and sorting of measured parts.

B Provide complete data showing the measured values of all values analyzed to a properly programmed digital computer.

Coordinated in this way, the measuring device and the computer can determine and control dimensional and tolerance trends that occur in the course of production where the parts are tested within the acceptable band of tolerances. Data obtained from such analysis can be used to reset machine tools, change or repair cutting tools, and generally to predict and thereby avoid the production of parts that have dimensions outside the tolerances.

In the case of assembly operations instead of individual parts, the measurement problem is more difficult to solve by conventional means because in many cases the assembly process adds parts that obscure the assembly area that needs to be measured or inspected to ensure that the entire assembly has been made correct. Often, because of this obfuscation, a completed assembly cannot be fully inspected to ensure that it has been properly assembled without first taking the unit apart. An automatic measuring device can be used to inspect partially completed assemblies to determine whether all assembly operations prior to the inspection point have been performed correctly. If such inspection stations are introduced just prior to each assembly step that will obscure an area requiring inspection, the entire assembly can be fully inspected without the necessity of taking the assembled unit apart to examine the obscured parts. It is therefore a further object of the present invention to provide means for measuring partially or fully completed assemblies to determine whether they have been correctly assembled up to the point where the measurement is made.

Briefly, these purposes are accomplished in an electro-optical system for measuring the dimensions of a machined part or other manufactured object or assembly thereof to determine whether the part meets predetermined standards. The system includes an electro-optical probe such as an image disector probe having a photocathode onto which is projected an image of the contour of the part being measured. Probing voltage for the probes is produced by a probe generator to create a probe whose shape, position and length are controllable to give a scanning path that crosses every edge of metrological significance.

Only one edge is scanned at a time, where scanning is for the purpose of determining its X-Y coordinates. The system program is controlled exclusively in the form of coordinates, where the accept/reject determinations are made on the basis of a comparison of the permissible tolerance for this coordinate measurement (or dimensionally determined from this) with the current value being measured.

Calibration means are provided which include a calibration network whose image is projected onto said photocathode instead of said partial contour, which means serve to modify the scanning voltages to correct for optical and electronic nonlinearities in the system. The network measurements can be used periodically to determine a long-term correction matrix or, when desired, the network can be projected onto the detector surface prior to each measurement to ensure very low operating conditions in a harsh environment.

In order to provide a better understanding of the invention as well as other purposes and features thereof, reference should be made to the following detailed description in connection with the attached drawings. Fig. 1 is a schematic view of the optical device for projecting an image of a part to be measured or of the calibration network onto the sensitive cathode of a detector unit in an electro-optical system according to the invention for measuring the dimensions of parts. Fig. 2 is a block diagram of the electronic system adapted to control the scanning shape and position and process the information present in the video signal provided by the scanner shown in Fig. 1. Fig. 3 graphically shows the image of a part which is projected onto the scanning surface and where a scanning is generated to examine an edge of the part.

Fig. 4A is an enlarged view of the scanning line.

Fig. 4B shows the voltage waveform required to generate the sweep as shown in Fig. 4A.

Fig. 5 shows the same part, 10, with a table showing where

then the individual dimensions are set up for measurement.

Fig. 6 shows a typical measurement subroutine.

Fig. 7 shows a calibration network for the system.

Figs. 8A to $G show the manner in which partial dimensions are examined by means of special sub-routines. Figs. 9A to 9C show the manner in which the radius of curvature and the center of curvature can be determined by other special sub-routines. Fig. 10 schematically shows the scanning device for following moving parts to be measured. Figures 11A to 11E show a collimated light technique for controlling the illumination of a part to be measured. Fig. 12 shows a preferred technique for measuring the thickness of a transparent object.

In fig. 1 schematically shows a device according to the invention for examining an object 3 in order to generate video information from which measurement data of interest can be obtained. The object 3 is shown in the form of a machined part which has a substantially rectangular shape and whose upper edge is provided with such an extension. It will be understood that this form is only intended to provide a simplified illustration and in practice the object can have a far more complicated design.

The object 3 is illuminated by an appropriate high-intensity light source 1 whose rays are focused on the object by means of a lens 2. The object 3 is inserted into the path that runs between the light source 1 and the sensitive photocathode in a scanning unit 11. Thus, only the profile or contour of the object is projected. The plurality of projected rays is, for the sake of simplicity, indicated by means of light rays 4 and 5 which are directed towards a lens 7 and emanate from this lens again as rays 8 and 9. These rays strike the light-sensitive surface of the detector unit 11, thereby to create an image on this, in a photomultiplier this surface has the form of a photocathode. The projected image 10 is a copy of the profile of the object 3 and therefore takes the form of a silhouette.

Interspersed between the object 3 and the lens 7 is an inclined, semi-reflective plate 6 which acts as a beam splitter which separates the rays from the object into two beams, one which passes successively through the plate 6 and the lens 7 and is directed towards the detector unit, and the other which is reflected by the plate and can be ignored. In practice and instead of a plate, one can use a beam splitter cube composed of two right-angled prisms that are joined, where the hypotenuse of a prism is coated with a semi-reflective layer. A thin semi-reflective plate is sometimes unsuitable due to ghosting or astigmatism unless parallel light beams are used.

Associated with the beam splitter 6 is also a second light source 2o whose light is focused by means of a lens 19 on the back of the beam splitter through a transparent calibration network 18, where the rays hitting the beam splitter are represented by rays 16 and 17. When the light source 20 is switched on and the light source 1 is switched on of the illuminated network formed by intersecting equally spaced vertical and horizontal lines is shown on the photocathode in the detector unit 11. Thus, the photocathode of the detector unit 11 sees either the silhouette of the object being examined or the calibration network, depending on which light source is active. The photomultiplier that is included in the detector unit is preferably of the image dissector type, as described in US Patent No. 3,593,286 where the photocathode, in response to incident light emits photoelectrons which are projected onto the plane of a delimiting opening. Operatively connected to this opening is an electron multiplier and a deØection system for deflecting the electron image over the delimiting opening in a way whereby different parts of the image are scrutinized at a desired frequency.

Thus the photomultiplier is provided with a set of deQection coils, one for vertical and the other for horizontal deflection,

and a focusing coil. A detector generator 12 supplies deflection currents to the coil set and a focus generator 13 supplies current to the focusing coil. A high voltage supply 14 which is connected to the photomultiplier supplies the necessary operating potentials to it. The output of the scanning unit 11 is fed to a preamplifier and waveform 15 whose video output contains the information to determine the crossing time of the scanning and the scanned part of the silhouette of the object 3 or the calibration network 18, depending on which light source is operative.

In fig. 2 shows the electronic system which is adapted to control the scanning shape and position and process the information present in the video signal provided by the scanning unit 11. The video output from the scanning unit is first processed in a video preamplifier 3 3 and a waveform shaper and video amplifier 3 4 to generate start and stop signals capable of controlling a high-precision clock 3 5 which provides periodic pulses at a predetermined constant repetition rate.

The output from the clock 35 is fed via the speed control circuit 28 (whose function will be covered later in this section) to a set of counters, namely an "X" counter 26 and a "Y" counter 27. These counters in turn each control a digital to analog converter (D/A) i.e. the counter 27 controls the D/A converter 25 and the counter 26 controls the D/A converter 24. The output from each D/A converter is directly proportional to the number fed into it by the associated counting. As the counter value is advanced by the input pulses from clock 17 through the speed control 28, each clock pulse advances the counter by one step and advances the associated D/A converter by a minimum increment (normally called the least significant bit or LSB = least significant bit).

The outputs from the D/A "X" converter 2 4 and from the d/a "Y" converter

25 are respectively fed into an "X" summing amplifier 38 and a

"Y" summing amplifier 21. Into the summing amplifier 38

is also fed another voltage obtained from a D/A converter 32 whose digital value comes from a computer-controlled storage register 31

to provide the voltage that determines the X scanning start position. Into the summing amplifier 21 is also fed another voltage obtained from a D/A converter 23 whose digital value comes from a computer-controlled storage register 22 to provide the voltage which determines the Y scanning start position.

The system function is as follows. When the computer program generates an order to measure the "next" dimension, the computer first reads the start coordinates of the survey for that dimension. These are read into the "X" register 31 and into the "Y" register 22 to set the associated D/A converters 32 and 23 to the desired value to establish the start of the scan. Then the clock 35 is opened to provide an input to the "X" counter 26 and the "Y" counter 27 via the speed control circuit 28 so that the sweep advances in a linear fashion by at least a significant increment

. at a time for each clock pulse into the counter.

Fig. 3 shows a typical sweeping of a part 10 in which only the horizontal sweeping movement is needed. As can best be seen from fig. 4a, an enlarged section of fig. 3, the scan moves from point XI.-Yl to point X2 - Yl to determine the coordinates of the point X2 - Yl which is the edge of interest. The X part of the survey is shown in fig. 4B where the scan drive voltage Vs starts at a value corresponding to XI and continues over a distance A X to drive the scan until the scan at X2 has cut off the edge of interest. This cutting will then stop the scanning process and order a readout from the X counter 26 to the computer. This reading corresponds to Ax

is the summary readout from the "X" counter 26 and is fed to the computer via the Ax coordinate data register 29. A corresponding readout for "Y" information is fed to the computer via A Y coordinate. the data register 30 which is connected to the "Y" counter 27. The search shown in fig. 3, 4A and 4B are in the horizontal direction and therefore do not require any change in the Y direction during the scanning period.

By simply feeding the output from the 3 o'clock 5 into either the "X" counter 26 or the "Y" counter 27, horizontal or vertical sweeps can be obtained. To obtain scans at any angle other than the simple horizontal and vertical direction, it is necessary to change the pulse rate ratio feeding the two counters. This is achieved by the speed control circuit 28. This circuit is set by the computer using input information which determines the direction (ie the angle) at which the scanning should take place.

From this, information is fed by the computer into the speed control circuit to change the pulse count ratio fed into the two counters.

After the scan's original position and motion requirement information has been sufficiently added together in the "X" summing amplifier 38 and the "Y" summing amplifier 21, the summed information is then used to generate a deflection current for the X coil in the scanning unit by means of an X - dehumidification drive circuit

36 and for the Y-coil by means of a Y-scan driver circuit 37. The scanner then generates the scan, where it is started at the correct point and the scan goes in the desired direction until it finds the edge that is being searched for. The video information containing this edge is processed by the video pre-amplifier 33 and the video amplifier and shaper 34, and the latter in turn generates a stop order to the clock 35 which stops the scanning with the correct count in the counters 26 and 27. Fig. 5 shows the way in which the image of the part ]0 can be scanned to find the coordinates of the eight corners that define the outline of the part. Each coordinate is given its corresponding cartographic number in figure 5. The survey used will be aimed at the regions marked A-B, C-D, E-F etc. up to and including 0-P. The communication between the person setting up the coordinate graph and the actual computer-controlled operation of this will take place by means of a table corresponding to the one shown in fig. 5. On this table, each of the scans used to find the edges associated with the corner coordinates is marked from A to. P as they are in the associated figure. The coordinates for each of the survey intersections are selected as shown. The size of the increment used to move the probe from the nominal value such as 0 to the probe position 0+ or 0- can be read in as a single number on - The table to be used for measurements of all coordinates for the same part, or a separate column can be created on the table for each of the instruments so that the scanning offset can be optimized for each coordinate. The table must also state the tolerance allowed for both X and Y, the calculation subroutine that may be required for the dimension being measured, and finally the layout subroutine that fits. The disposition routine is normally a provision with regard to how the part is to be sorted. In many cases, this is a simple accept/reject provision. In other cases, there will be a requirement to sort by type of dimensional error. In additional cases, there may be a printout that provides a permanently written information about certain types of errors. Fig. 6 shows a typical subroutine which will be listed under the calculation subroutine column. The one shown in fig. 6 requires determining the center, X-diameter, and Y-diameter of a circle. Since neither the exact position of the center nor the exact diameter of the circle is known in advance, it is necessary first to find the center of the circle and then the length of the X and Y intersections that pass through this center. The technique used is to draw a chord at an appropriate point on the circle. This is done by generating a chord such as the sweep line 39. A perpendicular bisecting line 40 is placed on this chord. This passes through the center of the circle and therefore, according to well-known geometric laws, the distance between its intersections represents the Y-diameter. A perpendicular bisector line 41 is deposited on the Y diameter. This also passes through the center of the circle and its intersections represent the X-diameter. Application of a sub-routine of this sort will

then determine the position and diameters of the unit circle 42 as also shown in fig. 5.

Fig. 7 shows a simplified calibration network for the system. It consists of 7 lines that run horizontally and 7 lines that run vertically. The current calibration network for the system consists of 64 equal lines, equally spaced. In practice, the calibration network can consist of black lines on a clear background or clear lines on an opaque background with any ratio between the width of the black lines and the width of the clear lines, or even a checkerboard or a set of parallel lines that first would be used to calibrate the system for scanning in one direction and then used to calibrate it for scanning in the orthogonal direction.

The calibration procedure simply consists of ordering the system to

to scan each horizontal and each vertical line once per block, which conceptually requires only 56 scans in the horizontal direction and 56 scan segments in the vertical direction for the calibration network shown. The calibration program that requires the system readout from the relevant counter (X counter 26 when scanning in the X direction or Y counter 27 when scanning in the Y direction) whose number is in the counter at the time of crossing the network line in question in each block. The calibration program subtracts each of the numbers associated with a network line from the numbers associated with the adjacent network lines. The difference between these values (ie the other differences) is stored as calibration corrections. The calibration procedure is repeated by applying these calibration numbers as corrections to the scan-start information fed for the X scan to register 31 and for the Y scan to register 22. The process converges rapidly,

and the result is a series of correction numbers which are stored in the computer and fed to the scanning control circuit via registers 31 and 22 for linearization of the scanning during all subsequent measuring operations.

For the most accurate use of the coordinate graph, this degree of correction is often not sufficiently accurate. Non-linearities can occur between the correction points. These are corrected using an interpolation process that takes place after the individual measurement has taken place. The intersection of the scan with the edge of interest orders a readout from the counters via data registers 29 and 30. This readout contains correction made to the scan start position of the stored calibration matrix, but the scan is not corrected between separate calibration points. thus, second-order nonlinearities occurring between the calibration points will still be present. However, when the readings from the counters are obtained, a further interpolation correction is calculated. This results in a further increase in the effective system linearity by as much as 10X, providing with the 64 x 64 calibration network used in practice, a system linearity of the order of one part in 40,000.

It should be noted here that calibration information is generated

by determining first and second differences between the network line intersections when scanning the image of a very precise network. This data is used to establish a quantized calibration matrix (nominally 64 x 64) for the entire image plane in the embodiment of the invention described above.

However, the same first and second difference data can be used for

to provide calibration in a continuous form over the entire image plane. A well-known technique for doing this is based on generating a polynomial expression in X and . Y which is used to describe mathematical nonlinearities to be corrected. In cases where these nonlinearities can be described by a relatively short polynomial (which normally means no expression higher than fifth order), this method has been used to provide linearity corrections approaching one part in three thousand, where a higher order determination increases the achievable linearity up to it approaches the part in ten thousand available from the raw data. Where these conditions can be met, transformation of said first and second reference data generated by scanning the image of the high-precision network into the form where it will provide the values of the coefficients for the descriptive polynomial is a relatively straightforward problem in a numerical analysis and it can be carried out in the run of minutes on a commercial minicomputer.

A second method of providing relatively continuous corrections for the entire image plane based on data generated from first and second differences as described above requires a series expansion of the nonlinearity function in the vicinity of the points selected in the field to be linearized. This procedure is based on the well-known mathematical technique by which any function can be generated if its value and the value of all its derivatives are known at a point. In practical applications, it is not normally necessary to establish the value of more than the first few derivatives of the function. Ease of manipulation can often be achieved by establishing the value of the function and the determining derivative at a few points within the field of interest. This method has the disadvantage of requiring a calculation cycle for the correction of non-linearities for each measurement made. The time required to complete this will often reduce the overall operating speed of the system.

As stated above, the calibration network will in practice have far more vertical and horizontal lines than those shown in fig. 7. A suitable practical figure is 64 horizontal lines intersecting 64 vertical lines to provide a network which when projected onto the photocathode in the detector unit will generate an image whose dimensions are approx. 2.5 cm x 2.5 cm and where the network line is separated by a distance equal to 0.04 cm.

With each of these network lines generating an image that is approx. 0.075 cm wide, the deflectable photomultiplier used will determine the position of the leading and trailing edges of each line with an accuracy substantially greater than 0.00025 cm. By using a suction opening with a physical diameter of approx. 0.002 cm. This requires the application of integration and edge enhancement.

and both of these techniques are well known.

With a baseline detector plus optical unit linearity of 0.25%, the calibration performed using a 64 x 64 line network of the type described using edge enhancement techniques to indicate the edge position will result in a system that can is calibrated to measure the basic accuracy of the grid to ± 0.003%. To the extent that networks are commercially available with accuracies in excess of 0.001%, the accuracy of the present system with a network of this commercially available type will be better than 0.004%.

Figs. 8A to E show the scanning system's ability to probe the end of a cylindrical part 74 whose inner bore diameter and outer diameter represent the ID and OD of the part, respectively. We assume that it is desirable to determine the coordinates at the center of the circular bore and its ID. To find the ID, a series of horizontal soke scans such as those indicated by the reference numbers 75a to e in fig. 8A, above the nominal center by starting with offset 75a just below the lowest position allowed by an acceptable tolerance and ending with offset 75e just above the highest allowable position. The X-diameter of the bore is determined by the largest dimension measured by this probing procedure. The X-coordinate of the center of the bore is represented by the position of the bisector of the diametral distance thus determined.

In a similar way, the Y-coordinate and the Y-diameter of the bore are found by means of a group of surveys 76a to e as shown in fig. 8B. The coordinates for the center of the bore are then the intersection of the largest X diameter and the largest Y diameter just measured. The roundness of the bore can be determined by running a large number of diameters 77 through the bore center which has just been established, as shown in fig. 8e. Variations in cuts in the section above some of these diameters indicate a deviation from roundness. Similarly, a deviation from the symmetry of the outer diameter of the part and the total indicator reading (TIR) will be measured as shown in fig. 8C by running a set of diameters 78 through the center using the previously mentioned procedures and determining the length of these diameters as they intersect the outer surface of the part.

The concentricity for the inner diameter and the outer diameter can be determined by running a series of radial sweeps 79, as shown in FIG. 8C over the ID and OD to measure the wall thickness of the part. Variations in thickness are directly indicative of the concentricity of ID and OD.

In fig. 8F shows a wavy contour of a very coarse gear which is to be measured. The circle 81 represents the root diameter of the gear 80 and the circle 83 represents the OD of the gear and the circle 82 represents the nominal pitch diameter of the gear. A probe located on this pitch diameter can measure the tooth-to-tooth distance of the gear and the various imperfections associated with a gear configuration based on tooth-to-tooth errors.

In a similar way, measurements can be made along the division diameter for a screw-shaped part, such as a screw as shown in fig. 8G

in longitudinal cross section. The parallel lines 87 that touch the grooves of the threads represent the root diameter, while the parallel lines 86 that intersect the threads represent the pitch diameter.-

A survey that runs along the lines 86 will measure variations

in the helical contour along this division diameter. This measurement can be carried out while the part is held stationary, or while it rotates about its longitudinal axis. The latter measurement will provide a complete description of the variations in screw contour along the division diameter throughout the entire length of the screw-shaped part for 360° rotation.

Figures 9A, 9B and 9C show a further measurement technique made possible by the flexibility inherent in a deflector. photomultiplier detection. The curve 88 represents in these figures either ID or OD of a part whose radius of curvature and whose center of curvature is to be established. The sample waveform in fig.

9A is represented by the wave shape 89, the one in fig. 9B of the waveforms 90 and the one in fig. 9C of the waveforms 91. These sample waveforms are formed by a sweep that has a known radius of curvature and a known center of curvature. Superimposed on this curved sweep is a high-frequency modulation component.

The sample survey and the unknown curve 88 in fig. 9A will have the same radius of curvature and center of curvature only when all the high-frequency cycles of the modulation component are symmetrically cut by the curve being measured. However, the curve 89 in the sample scan is in disagreement with the curve 88 being measured. Curve 89 has a larger radius of curvature and the intersection with curve 88 therefore starts and stops near the bottom of the high frequency modulation component, as indicated by points and Mj.

From fig. 9B, it will be seen that the sample curve 90 has a smaller radius of curvature than the curve 88 for the part being measured and that its intersection with the part's curve starts and stops at the top of the high-frequency modulation component, as indicated by the points M3 and M4.

In fig. 9C, the test curve 91j has the correct radius of curvature relative to the curve 88 of the part being measured, but its center is displaced in space from the center of the part curve 88. Under these circumstances, the intersections with the high-frequency modulation component will start at the bottom, as indicated by point M5 and stop at point Mg.

In all the examples shown in fig. 9A, 9B and 9C where the sample curve deviates in terms of radius of curvature or center of curvature from the curve for the part being measured, the magnitude and direction are determined

of this deviation by techniques which are well known for the generation of

an error signal with corresponding magnitude and direction. This error signal acts to control a closed servo valve to change the radius of curvature and center of curvature of the sample probe to match those of the unknown type. The fast-acting measurement system described here is particularly useful in the following situations.

A When it is necessary to make a large number of measurements on a complicated machined part such as an eccentrically shaped cam or machine stop.

B When multiple measurements of the same dimension are required on a high-speed production line. An example of this situation is a line that manufactures roller bearings at a speed of 5 per second,

and where inspection requirements dictate measurement at multiple positions along the length of the roll.

C When measurements of implied dimensions are required, such as the pitch diameter of a screw thread or gear. In this situation, a reference surface must be established. This reference surface must be placed carefully with regard to the unknown object and measurements must then be made with regard to specified parts of the reference surface. Under current conditions, physical wires are used to generate the reference surface. This whole procedure can be carried out by means of a system according to the invention under suitable computer program control which includes the generation and use of "electronic wires" as references.

The surface of the part presented to the system must represent

the dimensions to be measured and it must therefore be free of chips or other machining residues. Apart from the requirement of cleanliness

no other special surface preparation is necessary and the system will operate with equal ease on parts whose surfaces are either polished or have a matt or black anodised finish.

It is possible to use the detector 11 and its associated optics

for monster registration purposes in accordance with the technique described in US Patent No. 3,593,286. In this patent, a deflectable photomultiplier rudder is used to provide a scan of an optical image of a monster to establish the degree of similarity between the monster and reference monster. thus, the scanner 11 acts both to carry out dimensional measurement of a part and for pattern recognition. This dual application of the probe requires only a relatively small increase in the electronics associated with the system and yet provides the following additional features which are of great value in a device for measuring parts: A The part being measured does not need to be oriented precisely in order to

perform the desired measurements. By allowing the scanner to function as an input device to a correlator in the manner described in the latter patent (see figs. 4, 5 and 6), the video information generated thereby can be used to align the scanning to the shifted or rotated image of the part to be measured.

There is no mechanical movement during the operation as the scanning device takes place electronically as stated in the aforementioned patent. With the scan arranged in this way, the coordinates for the measurement scan are shifted and rotated by adding to them the same displacement and rotation correction voltages that are produced by using the scan in a correlation mode. As a result, the measurement probe is rotated and translated to the degree appropriate for the shifted image in its field of view.

B When using image - motion compensation techniques as described in connection with fig. 11 and 12 of the said patent

can one follow an object that moves through the field of view. This is of particular value when measurements are to be made on

parts that are transported on a moving belt, as measurement can then be made in the part during the entire period that the part remains

within the field of view of the system. It is also practical when measuring parts that are allowed to rotate by rolling over this field of view, thereby enabling the measurement of the diameters of balls or cylinders and making it possible to carry out corresponding measurements which must be identical regardless of the orientation of the part. By this corollary

nod, one can measure out-of-round conditions and other deviations from symmetry without the necessity of determining the total indicator reading or corresponding readings. The simplest techniques for making this measurement are shown in fig. 10 where the part 69, which moves over a flat surface 66 is observed by two separate detectors 67 and 71.

A detector 71 which responds to an image generated by light rays 73 projected onto it by means of a lens 72 follows the part as it moves across the field of view in the manner described in the aforementioned patent. The second detector 67, which uses a lens 68 and light rays 70 for measurement, is subordinated to the tracking detector by means of inputs supplied from the tracking detector to the analog summing amplifier 51 shown in fig. 6 in US Patent No. 3,854,822. A typical input to this amplifier is shown in fig. 6 of this patent as "DC OFFSET CONTROL 54".

When subordinated to follow in this manner, the measuring scanner 67 sees the image of the moving part stabilized within its scanning, and it makes measurements just as if there were no practical movement of the image. It should be noted that the nature of the calibration technique discussed here allows the probe to be calibrated at each new position of the probe essentially by using a standard sample and hold circuit between the follower/correlator and the control inputs designated by DC OFFSET CONTROL

54 in fig. 6 of US Patent 3,854,822.

When calibrating, the sample and hold circuits are put into "hold" mode so that calibration can take place with a static sweep at approximately the position at which it will be for measurements. The sample and holding circuit are then set in "sample" mode whereby the measurement scan "catch up" with the subsequent scan, thereby stabilizing the moving image with respect to the measurement scan. With circuits of a well-known type, the time constants are such that the "hold" mode of calibration and the "rotate" mode of tracking can be accomplished with at- V

■sufficient speed to allow total system operation to continue at its normal rate of measurement.

It is often desirable to be able to choose different plans

in object space to generate the image to be measured on the surface of the detector unit. These techniques have previously been used to do this to a limited extent. More specifically, collimated lights can be used to provide a limited degree of control, as will be explained in connection with fig. 11. Fig. 11A shows a hollow cylinder 92 of tubular design and shown in cross-section. Fig. 11B

is the same cylinder shown in end view. The inner diameter is not a perfect cylinder, as the upper section is slightly thicker at point B and the lower section is slightly thicker at point E.

As shown in fig. 11C will collimated light 93 which is used to measure

the inner diameter of this cylinder will be cut off at points B and E. The image generated by passing the collimated beam through the inner diameter of this cylinder will indicate an inner diameter smaller than the left diameter (AD) or the right the diameter (CF).

It is often necessary to obtain either the diameter AD or the diameter

pure CF while illuminating the inner diameter of the cylinder, with a light from the left of the cylinder. The image must be viewed with a detector placed to the right of the cylinder, with appropriate optics between the right side of the cylinder and the detector to generate the desired images. In fig. 11D, in order to see the diameter CF while illuminating the cylinder from the left, it is necessary to cause the light beams 94 to diverge in a controlled manner. The divergence angle can be determined geometrically. The divergence must be large enough that the rays emerging from the right side of the ID of the cylinder have been cut off by the diameter of interest, namely the diameter CE

and not of the diameter that is slightly to the left of the diameter of interest.

In a similar way it is, to choose the diameter AD which is on

the illuminated side of the cylinder, it is necessary to apply illumination with a controlled degree of convergence as shown in fig. 11E. The degree of convergence can be determined geometrically. The requirement is that the rays 95 coming out on the right side of ID of the cylinder are intercepted only by the diameter AD and not by any surface to the right of AD.

Another measurement problem that arises in practice for which no practical solution exists today is measuring the thickness of a transparent material such as a glass plate 97 as shown in fig. 12. If the physical design of the material to be measured permits, techniques similar to those used for metal may be used. These include micrometers, height gauges etc. However, it is common to generate glass forms with highly twisted and often closed wall designs. The problem is particularly aggravated by the fact that the generation of these shapes is normally carried out by a blowing or sinking process during which it is extremely difficult to control the wall thickness of the glass being processed.

It is necessary, to ensure the overall strength of the final object, to determine the wall thickness of the finished object. A method for making such measurements in a non-destructive manner under practical production conditions in accordance with the invention includes directing an incoming light beam 96 towards the glass plate 97, the thickness of which is to be measured, so that two reflections are generated. One reflected ray designated by reference number 99 originates from a first surface reflection and the second reflected ray designated by reference number 98 is a second surface (often called ghost) reflection.

It is quite clear from studying the geometry shown

in fig. 12 that for a given angle between the incoming beam 96 and the first surface of the glass plate 97, the distance between the two reflected beams 98 and 99 is a direct measure of the thickness of the plate. The scanning technique for measurement discussed above is directly applicable to measuring the distance between these reflected rays to provide a possible conversion method for measuring piece of glass. Although a preferred embodiment of a system according to the invention has been discussed, it will be understood that several changes can be made in this embodiment without deviating from the essential inventive idea and the scope of the invention as defined in the following claims.

Claims (14)

1. Electro-optical system arranged and adapted for inspection of manufactured objects by determining coordinates of critical parts of the objects, the system comprising an electro-optical detector having a light-sensitive surface, and an optical projector arranged to project a contour of the object which is to be inspected on said surface whereby an image of the object is created on it, characterized in that a scanning generator (12) is operatively connected to the scanner (11) and provides scanning voltages which cause the scanner to follow a scanning path that crosses such edges of the image for which the coordinates are to be determined, whereby video signals generated by the scanner enable the determination of said coordinates. 2. Electro-optical system as stated in claim 1, adapted to establish whether the objects meet predetermined standards or not, characterized by an information store containing a corresponding nominal coordinate value for each coordinate to be determined and a comparator that compares the coordinates that are determined from the video signals with the nominal coordinate values that correspond to these and generates an output signal that indicates any difference between the two values. 3. Electro-optical system as set forth in claim 2, characterized by an input information matrix connected to the seeker generator and to the information store for supplying information to the seeker generator relative to each coordinate to be determined in order to perform scanning along a scanning path suitable for each such coordinate and adding to the information store the corresponding nominal coordinate values, a computer computer network and adapted to calculate the distance between two sets of determined coordinates to thereby obtain the dimension between said coordinates, a reference register containing the nominal dimension between said determined coordinates and a further comparator arranged to compare the calculated dimension with the nominal dimension and provide an output signal indicating any difference between them. 4. Electro-optical system as specified in claim 1, characterized in that instead of the object, a reference sample (18) can be arranged which can have its image projected onto the surface in order to modify the scanning voltages thereby to correct optical and electronic irregularities in the system with respect to each coordinate that is determined. 5. Electro-optical system as stated in claim 4, characterized in that the reference sample comprises a calibration network (Fig. 7) whose projected image covers the entire surface and defines a reference coordinate for each coordinate to be determined and that a correction comparator compares the coordinate to be determined with its reference coordinate to thereby generate an error signal adapted to modify the scanning voltage. 6. Electro-optical system as stated in claim 4 or 5, characterized in that a beam splitter (6) is arranged to project the reference sample image instead of the object image and includes a first projector (1,2) arranged to direct an illuminated contour of the object on one side of the beam splitter so as to project one of the resultant beams onto the surface' and a second projector (20, 19) arranged to direct an illuminated image of the reference sample onto the other side of the beam splitter. so as to project one of the resultant rays onto the surface. 7. Electro-optical system as stated in claim 1, characterized in that the detector is formed by an image disector rudder whose surface comprises a photocathode, where said rudder includes vertical and horizontal deflectors which are sensitive to the scanning voltages. 8. Electro-optical system as stated in claim 3, characterized in that the information matrix is provided by the computer, said computer containing a memory. 9. Electro-optical system as specified in claim 3 or 8, characterized in that the matrix supplies information to the information store regarding permissible tolerances for the coordinates to be determined, whereby the comparator output signal indicates whether the determined coordinates are to be accepted or rejected . 10. Electro-optical system as stated in claim 9, characterized in that the matrix further adds information to the information store with regard to sorting level criteria with regard to coordinate variation, where the comparator output indicates the sorting level to which the variation in the particular coordinate belongs. 11. Electro-optical system as stated in claim 5, characterized in that the scanning is modified by means of the error signal to the extent that the coordinates obtained by the scanning voltage agree with the actual coordinate position of the scanning opening which is projected onto the surface by means of a zero-error reference sample . 12. Electro-optical system as stated in claim 5, characterized in that the correction comparator includes a reset arranged to reset the origin for the coordinates that are electrically established by the detector and the detector generator to thus agree with the origin that is established by the reference sample. 13. Electro-optical system as stated in claim 12, characterized in that the computer is adapted to establish known distances from the origin on the surface and provide signals to displace the scanning so that it rests on a nominal position that is thus determined. 14. Electro-optical system as stated in claim 13, characterized in that the computer is further adapted to establish the actual position of the displaced scan relative to the reference position of the surface and provides signals to correct any difference between the positions as . is necessary for the calibration of such scanning displacement.