IE913793A1

IE913793A1 - Optical inspection of moving objects

Info

Publication number: IE913793A1
Application number: IE379391A
Authority: IE
Original assignee: Saint Gobain Cinematique
Priority date: 1990-11-20
Filing date: 1991-10-30
Publication date: 1992-05-20
Also published as: EP0487402A1; JPH04291140A; NZ240443A; FR2669429B1; CA2055640A1; AU648354B2; AU8691891A; MX173902B; ZA918739B; BR9105019A; NO914522D0; NO914522L; FR2669429A1

Abstract

The subject of the invention is a machine for inspection of objects in movement which are equipped with an angular reference mark, associating, along a conveyor, an optoelectronic device comprising an emitter and a receiver working in rectangular cartesian coordinates followed by an image analyser, slaved to a microcomputer and equipped with an electronic converter (10) able to restore into fixed position in front of the cartesian analysis grid of the analyser (9) any instantaneous image obtained on the receiver (4). This machine, usable in particular in the hollow glass industry, makes it possible in particular to inspect the bottoms of receptacles in real time, at a very high rate, on production lines.

Description

OPTICAL. .INgRECTIoN PL.MQ21Hfi^SBJ£gIS The optical inspection of objects made in large numbers 5 both with a view to deciding whether they are acceptable or not and to monitor the general quality level of a production line involves certain dimensional examinations and inspections, but particularly with respect to the appearance. The latter reveals the surface state and more particularly its abnormalities and, in the case of transparent or translucent objects, possible internal faults.

This is why such an inspection is very widely used in the glass and in particular hollow glass industry, either for the fine inspection of de-limited zones able to have specific defects, particularly checks and then varyingly wide or narrow oriented light beams are used, or on a larger part of the wall for the detection of more varied defects and faults, such as inclusions, paste defects, thickness, adhesion, etc. In this case it is necessary to study an image of the moving object with or without rotation, usually formed by transparency in diffuse lighting.

It is then necessary to analyze in real time i.e. at a very high speed, the signal supplied by an optoelectronic device located along a conveyor and associating with an adequate emitter a receiver operating in rectangular Cartesian coordinates, i.e. constructed around a matrix camera (CCD: Charged Coupled Device) equipped with an array of cells capturing a virtually instantaneous overall image. A priori this image is not uniform, even on a perfect object and, no matter what analysis method is chosen, must, if it is wished to obtain adequate information, be compared by image analysis members and then decision members with a model constituted by an empirically determined table of criteria. The complete means is controlled by a microcomputer.

One of the difficulties is that the higher the inspection speed, the more difficult it becomes to manipulate the objects. A widely adopted solution consists of conveying them upright on generally rectilinear, horizontal belt conveyors using lateral belts for placing them and, if need be, for rotating them in the case of circular bottles. However, it remains difficult to observe the image in a uniform position, even by synchronizing its acquisition with the passage of the object.

The first objective of the invention is to provide optical inspection machines equipped with a device able to restore in a fixed position in front of a Cartesian analysis grid, each instantaneous image of an object provided with an adequate angular marker and placed on the conveyor obtained in a variable position on their receiver. Thus, it proposes for said machines a generally applicable electronic converter means able, by insertion between the camera and the analyzer, to transpose in real time at speeds of 36,000 articles per hour and higher the digitized electronic signal from said camera. This device must at the minimum be able to reset each image in a fraction of the video reading time, not only in translation in the direction of the conveyor, but at least also in perpendicular translation and in rotation starting from the detection of its silhouette and the position of the marker.

Receiving its data from the camera to transmit them to the image analyzer, the converter according to the invention groups, in conjunction with the central processing unit ' (' I -r W controlling the machine and the circuits of the analyzer, several specialized units namely a location unit, a graphic unit and a transposing unit, which will he described in greater detail hereinafter in connection with the inspection or examination of the bottom of a glass container. 0 This embodiment chosen for reasons of clarity will be accompanied by information on certain transpositions which can be easily made to other cases. In the drawings shown: Fig. 1 A possible structure of an inspection station. Fig. 2 The image received on the camera.

Fig. 3 The shape Cf the row signals.

Fig. 4 A flow chart.

Fig. 5 The image stored by the initial memory.

Fig. 6 The standard image of a perfect bottle. is placed a gap The image of the As shown in Fig. 1, usually, but not necessarily, between two successive sections la, lb of a horizontal belt conveyor 1, beneath a light emitter 2, covered by two opposing lateral belts 3. bottom, obtained by observing in accordance with the vertical axis of each of the containers D in the form of bottles driven in translation at minimum intervals by these two belts, is formed on the sensitive surface of a camera 4 e.g. constituted by an array 5 of 384 x 288 square meshes, whose useful field E is arranged lengthwise in the direction of the conveyor.

No matter whether it is circular or not, direct or indirect and even supplied by the flat bottom (in reality slightly concave) of a perfect bottle, said image [S] shown in Fig. does not have a uniform brightness. There can not only be thickness or lighting differences, but also systematic disturbances, namely ribbed areas S, for cooling the shoulder, S2, S3, coded or uncoded writing: factory, mould number, etc., orientation marker M, e.g. for sticking labels, etc. and said different zones involve different severity criteria and treatments.

As shown by Fig. 3, the video signal 8 supplied linewise by the camera is firstly digitized by a converter (ADC = analogic-digital converter) in a series of digits Z to 2® « 2S6 grey levels, representing the average illumination of each mesh, usually addressed in a target memory for supplying an adequate analyzer comparing the image obtained with the table of criteria defined by the inspection protocol, whereby the latter can be likened to a standard image.

In principle, the image is of a constant size, because it is taken at fixed distances on the optical device, but will be captured by the array 5 in a slightly variable position within the available field E. Moreover, even if it is not circular, a priori the bottle has a not known orientation.

It is necessary to overcome this double difficulty for comparing its image with the analysis grid formed by the model, if it is e.g. wished to avoid any local marking neutralizing a complete ring on the bottom.

So as not to overburden reading, Fig. 4 provides with respect to the operation provided by the invention a diagram reduced to the circulation of information obtained on the basis of the signal Z from the converter 6 in conjunction with the central processing unit or CPU 7 of the inspection machine.

Here, the signal Z is not directly stored in the target memory 8 for supplying the analyzer 9, whose principle ie known per se and which is therefore not specifically described, but is firstly converted into a signal W by an electronic conversion device 10 also controlled by the CPU 7. Thus, it is firstly stored as a quadruple image (Z) on a matric [XY] corresponding to the design of the sensitive surface, in the four parallel memory planes 11a, lib, lie, lid of an initial memory ll (1 M-byte random access memory or RAM).

Summated on the complete image, a histogram of grey levels will bring about the systematic appearance of two populations, that of the very light colours corresponding to the bright bottom and that of the darker colours which may carry certain significant peaks of systematic or accidental abnormalities, but which certainly indicate the presence of the object. This makes it possible to find a threshold s from the signal Z to a black (0) and white (1) binary signal z, which will supply the silhouette [zj of the object, namely that of the bottle bottom or more precisely its shadow. The presence of belts causes no problems.

Working in a data flow and able to firstly carry out the aforementioned processing on the signal received at Zv it is the location unit 12 which will then summate row j by row j and column i by column i the matrix {XY]» in accordance with the diagram of Fig. 5, on the basis of the signal Z2 reextracted from its video memory 12a the respective quantities j.nj and i.n,, in which nj and ni are respective black numbers of the signal ζ in the corresponding row or column. To within the factors of the numbers of rows of the matrix, the thus obtained sums I and - ΖΛ »· J are the coordinates of the centre of gravity G of the shadow, which is independent of the chosen system of coordinates, i.e. in the present case the orientation of the bottle on its conveyor. When entered in a matrix (L], they will conversely supply the translation L'1 to be imposed on the image in order to bring said centre to the chosen origin 0 of the target matrix [UV], with reference to which will then be made the image analysis calculations for characterizing the bottle with respect to the standard image [T]. During this operating phase, it should be noted that it is possible to use the still free memory β as the memory 12a.

It should also be noted that for fixing a centre it is possible, on a double symmetry object, to replaae the investigation of the centre of gravity G by that of the median point. The corresponding calculation is simpler, but this gain poorly compensates the loss of flexibility resulting from such a choice.

The aforementioned information is also transmitted by the central processing unit 7 to the graphic unit 13, which uses it for authorizing in the memory li, along the circle C, centred on G and whose radius r is known beforehand, namely parallel in the four planes lla to lid on four quadrants C# to CQ (monotonic functions), the reading Z3 by the analyzer 9 of the light profile of the stored image [Z]. The analyzer will detect a specific intensity sequence or variation in it, characterizing an angular marker M, either previously existing or which is specifically provided for this purpose on the bottom and in particular a bead forming a lump in front of the emitter.

As soon as the coordinates P and Q of this marker are identified in the matrix [XY], the CPU 7 will supply the directional coefficients, cosine and sine, p - (P-l)/r and q = (Q-J) /r of the direction of said marker point (M) . When inscribed in a matrix [R], they will conversely determine the necessary rotation R*1 for bringing its image on to one of the axes of the target plane, e.g. the U axis.

The translation L*1 and then the rotation R'1 of the complete image [Z] will bring the latter in front of the stendard image [T] of a perfect bottom in order to permit the analysis thereof by any appropriate method. In practice, the desired algorithm, in the present case the moving matrix (D), which is the opposite of the matrix product [L][R] will make it possible, according to Fig. 6, to pass from the matrix [UV] to the matrix tXY] and consequently to the image obtained on the camera, in order to determine with which elements of the latter corresponds the signal intensity W which will characterize a pixel (U,V,W) of the target matrix, or better conversely to which pixels (Χ,Υ,Ζ) of the initial matrix to refer for calculating W and in what way. Use will be made of homogeneous coordinate· with 3x3 matrixes.

The CPU 7 carries out the calculation of the matrix (Dj and transmits it to the transposition unit 14 for that of W. If it is accepted that W is the direct representation of the illumination of the image, a possible method for determining the value to be attributed to it is the transfer of the value Z of the closest pixel, or even that corresponding to the integral values of X' and Yz. Indeed the coordinates X' and Y' of the initial point corresponding to said target pixel will only exceptionally be integers. The circuits used are suitable for using more accurate methods, but these methods will be more complex. In view of the time constraints, the preferred, relatively simple method is that of bilinear interpolation between four adjacent pixels, pro rata to the inverse of their coordinate differences.

If it is assumed that to a certain pixel A(U,V) of the target matrix corresponds an initial point A' (X' - Χ+χ, Y' = Y+y) , in which X and Y represent the integral parts of the coordinates and x and y their fractional parts, said point is located within the square formed by the centres of the four boxes (X,Y), (X,Y+1), (X+1,Y), (X+l, Y+l). If the respective brightnesses thereof are ZQ0, ZQ1, Zlo, Ζυ to A' and therefore to A will be attributed a brightness Z' = (1-x) (l-y) Zoo + (l-x)y Z01 + x(l-y) Zw + xy Z„ = W. As an arbitrary example with decimal notation and assuming that to the pixel A of the target matrix corresponds, within the square formed by the centres of the four boxes: * (Xo = 8, Yo - -3) of brightness 137, - (Xo « 8, Y, = -4) of brightness 145, “ (X, “ 9, Yo “ “3) of brightness 123, ’ (Xn = 5, Y, - -4) of brightness 111, an initial point A* of the coordinates X' » 8.37 and Y' = -3.48, to it will be attributed respective multiplying coefficients: kx » 0.63 and 0.37, ky = 0.52 and 0.48, i.e. a brightness W equal to: Z' = 0.63x0.52x137 + 0.63x0.48x145 + 0.37x0.52x123 + 0.37x0.48x111 - 132.

The calculation to be carried out will therefore consist of 35 investigating boxwise with respect to the useful area of the target matrix the coordinates of the initial point, on the basis of their integral parts, the brightnesses of the four initial pixels surrounding the same and of their fractional parts the four multiplying coefficients k, in order to carry out the above summation.

The transposition unit 14 essentially has two data transposing circuits 14x and i4y (irs ·» image Resampling Sequencer) carrying out the calculation in question by a series of linear iterations, and a multiplier-accumulator 15 (MAC).

Object after object, the CPU 7 calculates the matrix (D), whose coefficients are stored in buffer stores of each of the two transposers I4x and 14y. On the basis of these coefficients and for each pair U,V of the useful field, said transposers process the values X' and Y'. They select their integral parts X and Y in order to sample in parallel the four corresponding values Z in the four planes of the initial memory 11 and their fractional parts x and y for storing the corresponding interpolation coefficients k in a buffer store 16, The multiplier-accumulator 15 performs the calculation of Z' = W, which it transfers into the target memory 8, whose data will then be used only for supplying the analyzer 9. It should be noted that the latter could directly extract certain of its data from the values Z00 ... Z1V but once again this would lead to a possible loss of use flexibility.

The main members used for constructing the device are as follows: - central processing unit 7 : Philips RTC microprocessor 68070 - location unit 12 : asics HISTO of Imapply International, - graphic unit 13 : QPDM AM95C60 of Advanced Micro-Devices, - transpiosers 14x and I4y - multipllier-accumulator 15 - memories 8 and 11 : TMC 2301 of TRW, : TMC 2210 of TRW, : 2*x29 box RAM (1 M-byte). 10 Within an inspection or approximately loo ms, such examination period a device is now able of e.g. to record an image in less than 15 ms, i.e. less than 20% of the cycle time, with an adequate rendition, on the raster of a 512x512 box matrix memory and even increase or decrease the size of the image. It should also be noted that under certain speed conditions said members, which can be completed by auxiliary circuits, make it possible to make the treatment of the image [Z] more or less complicated, as is explained by their use instructions. Thus, in the case of an inspection of sidewalls, it is possible to give the wall a developed image [W] by modifying the scale of the single coordinate u. For example, for the inspection of a pitted bottom, it would be possible through adding a conversion matrix, to expand in a variable manner around the point G the two coordinates U and V, Finally, optionally certain of the above-described units could during the remainder of the cycle fulfil different functions in conjunction with the CPU and the image analysis members. Thus, it is possible to use as the 0 graphic unit 13 that intended for controlling the processing operations to be carried out on the final image [W] in the different zones of the standard image [T], dead zones Tv Τ?, N, reading an alphanumeric code T3, investigating local defects F with various sensitivities T4, T5, thickness measurements, etc.

Claims

1. 5 1. Machine for the inspection of moving objects on a conveyor controlled by a microcomputer, associating along the conveyor an opto-electronic device incorporating an emitter and a receiver supplying and transcribing to rectangular Cartesian coordinates one instantaneous image 10 ([fr]) of each of the objects followed by an image analyzer, characterized in that, the objects being provided with an angular marker (Μ), it is also equipped with an electronic converter (10) able to restore in a fixed position in front of the Cartesian analysis grid ((TJ) of the analyzer (9), 15 its image ([Z]) obtained on the receiver (4) on the basis of the identification of the centre (G) of the image and the position of the marker.

2. Optical inspection machine according to claim 1, 20 characterized in that said device is able to transform the image both by deforming it and by similarity transformation.

3. , Optical inspection machine according to claim l, 25 characterized in that the transformation merely consists of a bidimehsional displacement.

4. Machine according to any one of claims 1 to 3 for inspecting the bottom of containers, comprising a light 30 emitter (2) associated with a matrix camera (4) on the same vertical axis.

5. Machine according to any one of claims 1 to 4*> placed between two successive sections (la, lb) of a horizontal 35 belt conveyor (1), facing a gap covered by two opposing lateral belts (3) driving the containers with a mininun spacing.

6. Electronic conversion device for a machine according to any one of claims 1 to 5 able, by insertion between the receiver and the target memory (8) supplying the analyzer, to transpose in real time the signal supplied by the 10 receiver (4) and then digitized, characterized in that it groups, in conjunction with the central processing unit (6) controlling the machine, a location unit (12) determining the centre (G) of the observed image, a graphic unit (13) authorizing the analyzer to seek on the light profile of 15 said image along a given marking line of the matrix (X,Y) on the basis of values (Z) of the signal stored in an initial auxiliary memory (ll) a characteristic sequence of the marker (M), the CPU (7) then calculating the transformation algorithm ((D)), which will bring the grid 20 (T) on to the image ([S]), a transposition unit (14) carrying out pointwise the transformation and calculating by interpolation, on the basis of the initial value (Z), the converted signal value (W) to be conversely allocated to the successive boxes (U,V) of the grid.

7. Conversion device according to claim 6, characterised in that the location unit (12) is able to supply the histogram of the intensities of the light signal (2) , to produce from a threshold found on said histogram a binary 30 signal (z) supplying the shadow ([z]) of the object, to calculate from the position of the centre thereof on the initial matrix ([X,Y]) the translation (L) to be imposed on the image ([S]) in order to bring said centre to the origin (0) chosen on the target memory.

8. Device according to claim 7, characterised in that the location unit determines the position of the centre of the 5 image by calculating the coordinates of the centre of gravity of the shadow or, optionally, on a double symmetry object those of the median point.

9. Device according to claim 7, characterized in that,

10. For determining the rotation (R) to be imposed on the image, the graphic unit authorizes the seeking along a marking line constituted by four quadrants of a circle (C) having as its centre that of the shadow ([«)). 15 10. Device according to claim 6, characterized in that the transposition unit (14) essentially comprises two data transposers (14x, I4y) supplying a multiplier-accumulator 15. 20 ll. Device according to claim 10, characterized in that the transposition unit (14) processes the value (W) of the converted signal by its multiplier-accumulator, boxwise of the useful area of the target matrix ([UV]), by bilinear interpolation between the intensities of the signal (Z) 25 appropriate for the four adjacent pixels of the corresponding initial point, on the basis of the integral parts (X,Y) and the fractional parts (x,y) of its coordinates. 30 12. Device according to claim 11, characterized in that the initial memory (11) has several memory planes scanned in parallel by various members.

11. 13, Device according to one of the claims 9 to 12, 35 characterized in that the transposition unit operates in homogeneous coordinates for recording the image ([S]) on the analysis grid (T) by iteration by means of a 3x3 matrix [D], inverse calculated by the central unit (7) of the product of the two matrixes of rotation <[RJ) and translation ([L]). io

12. 14. Device according to claim 5, characterized in that ite operating time does not exceed 20% of the duration of the examination cycle and that during the latter certain units (8, 13) are used in turn with different functions.

13. 15. A machine for the inspection of moving objects on a conveyor controlled by a microcomputer substantially as herein described with reference to the accompanying drawings.

14. 16. Electronic conversion device substantially as herein described with reference to the accompanying drawings.