GB2027963A - Automatic mark recognition - Google Patents

Abstract

Information is encoded manually on a document (2) by making marks (M) at appropriate positions of a marking matrix. The document has a row (TZ) of timing marks (TM) which are recognised (in preprocessor 41) in order to ascertain the column positions. When the document is read by being carried past a scanning unit (10, 11), level shifts and/or oblique positioning thereof can arise due to cutting and printing tolerances or misconveyance. To compensate such possible sources of error, the preprocessor (41) also recognises the vertical position of the timing row (TZ) to ascertain the row positioning relative to the scanning unit. In addition, a code column (CS) with code marks (CM1) is provided on the document and their position is analysed to determine the magnitude and direction of the oblique positioning (if any) of the document. Image data derived from marks (M) is stored in a store and the information derived from the timing row and code column is used to enable the data relative to each mark to be read out correctly by reading data from appropriate columns of the store which are selected in accordance with the amount of obliqueness detected. <IMAGE>

Description

SPECIFICATION Automatic mark recognition The invention relates to automatic recognition of marks on a machine-analysable document.

The advent of high-speed document readers has had a fundamental influence on data input in automatic data processing systems. They have facilitated direct data processing wherein the data to be input into the data processing system is directly read out from the original document. In order to achieve character recognition which is as accurate as possible, the mechanical read-out facilities have generally been limited, however, to the recognition of printed documents which are standarised for purposes of mechanical optical character recognition.

With this limitation on the possibilities of direct document processing, machine-readable data can only be entered onto a data carrier if so-called coder printers are available for purpose at each remote acquisition point. It requires less outlay, is more cost favourable, and nevertheless reliable, to enter handwritten marks in small boxes previously printed in a "drop-out" colour, e.g. by entering crosses, on a document printed in machine-readable script. The positions on the document defined by the marking boxes are assigned specific significances, e.g. the digits 0 to 9. The boxes which are to be marked are arranged in regular fashion in rows and columns and printed in the form of a marking matrix onto the document. Marks are read out centrally by means of an automatic mark recognition device.Devices of this kind for many purposes, replace conventional punched cards, the handwritten marks replacing the punched holes.

Frequently the device for the automatic recognition of such handwritten marks consists of an attachment to a document sorter. Then the marking matrix can be read out together with a OCR row during the same document run.

The information content of a mark is determined by the position of the mark on a specific blank document.

Therefore a device for the automatic recognition of handwritten marks must firstly precisely determine the position of the mark boxes previously printed on a document relative to its scanning pattern. The more precisely this can be effected, the less danger exists that a mark will be assigned to an incorrect position or in fact to a plurality of positions, i.e. mark boxes.

For this purpose, in mark documents, timing marks are printed in a row which is arranged in parallel to the rows of the marking matrix. Each of these timing marks is in alignment with a column of the scanning matrix.

As the marking matrix is normally read out in the direction of the rows, the positions of the columns in the scanning pattern is thus defined. In known mark readers a fixed row positioning is assumed relative to the lower edge of the document.

This rigid assignment involves difficulties when a document which is to be read out shifts in level due to misalignment in the document transport section or passes the optical scanner of the reading device at an oblique attitude. However, even when the document is transported satisfactorily in the scanning unit, the level of the marking matrix can nevertheless be subject to a shift e.g. when the document itself is not cut correctly. Oblique positioning can occur whenever the document is not correctly aligned at one of its edges during transport. All these deviations from a standardised positioning lead to inaccuracies in spite of the fact that the marking matrix is divided into columns by means of the row of timing marks, and therefore marks may occasionally be recognised as being assigned to the incorrect row position and column position.

According to the present invention there is provided an apparatus for reading a document upon which information has been encoded by the making of markings at one or more of a plurality of predetermined positions of a marking matrix formed by notional rows and columns, the document also having a row containing first reference marks and a column containing second reference marks, the apparatus comprising scanning means arranged in operation to scan the document and produce image data in respect of locations defined by the rows and columns of a scanning matrix, a store in which the image data can be stored, row identification means responsive in operation to image data obtained from the first reference marks to produce row assignment data indicative of the correspondence between rows of the marking matrix and rows of the scanning matrix, column identification means responsive in operation to image data obtained from the first reference marks to produce column assignment data indicative of the correspondence in the row of first reference marks between columns of the marking matrix and columns of the scanning matrix, alignment recognition means responsive in operation to image data obtained from the second reference marks to produce column correction data indicative of the magnitude and direction of any difference in the angular position of the columns of the marking matrix and columns of the scanning matrix and retrieval means arranged in operation to retrieve from the store image data stored therein relating to locations of the marking matrix using, for each location in respect of which image data is to be retrieved, said row assignment data for determining the relevant row(s) of the scanning matrix and said column assignment and column correction data for determining the relevant column(s) of the scanning matrix.

In this way, one can effect a purely electronic correction of the scanning result. The shift in level of a read-out document is non-critical since scanning means - e.g. the individual elements of a row of photo-diodes - which scans an entire row of the marking matrix, are not assigned a priori to specific rows of this matrix. In fact the position of the centroids of the individual first reference marks (timing marks) is used as reference value for the relevant row and column coordinate. In dependence upon this reference value the image data are assigned to the rows of the marking matrix. In order to correct an oblique position the second reference marks - for example in a so-called code column which precedes the marking matrix are analysed.

In addition to code marks characterising the nature of the document, this code column can also contain two position marks, whose mutual displacement in the two direction - expressed by the spacing between the columns in the scanning pattern - is determined and analysed in such a manner that the positions for the mark boxes of each column of the marking matrix are subsequently corrected in the scanning result.

In the following an exemplary embodiment of the invention will be described in detail with reference to the accompanying drawings, in which: Figure 1 is a block circuit diagram illustrating the basic construction of a device for the automatic recognition of marks; Figure 2 illustrates a portion of a typical document from which the structure of the marking matrix can be seen; Figure 3 is a block circuit diagram of a circuit arrangement for determining the vertical position of a timing mark; and Figure 4 is a block circuit diagram of a circuit arrangement for correcting oblique positions in respect of image data in each scanning column.

The recognition device shown in Figure 1 consists of an optical mark scanner and electronic circuit arrangements for the recognition of marks. The mark scanner contains scanning optics 10 and a photo-electric transducer 11 and is located at a position for read-out stations in the document transport route, for example accommodated in the document input section of a document sorter. The documents 2 which are to be scanned are conducted past the read-out stations in the direction of the arrow 21 by a suction drum (not shown).

The nature of these documents 2 is shown in detail in Figure 2. The document contains a marking matrix composed of 17 rows Z1 to Z17 and an arbitrary number of columns S. Mark boxes MK which are to define the size and position of marking M are pre-printed, in "drop-out" colours, at the intersection points of the rows Zn and the columns S. The drop-out colours which are used to print the mark boxes MK lie outside of the spectral range of sensitivity of the opto-electrical scanning unit 10, 11 and therefore do not influence the mark recognition. Commonly used drop-pout colours are light yellow, orange or red tones. The mark boxes MK are marked by entering a diagonal cross "X" in a soft pencil or ball point e.g. in black or blue. Other marks M, e.g. upright crosses "+" or vertical or diagonal lines are also recognised.However, in the present case markings comprising diagonal crosses offer the greatest accuracy of recognition.

Above the marking matrix is arranged a timing row marking TZ containing first reference marks, or timing marks, TM printed in black. The positioning of the timing marks TM defines the position of the columns S and rows Zn of the marking matrix. A recognition process takes place only when a timing mark TM is discovered. As a result of this provision the marking matrix can be interrupted by other items of information without affecting the recognition process, provided that the timing row does not contain any timing marks TM at these points.

The marking matrix is read out from right to left in a direction which is opposite to that of the transport direction indicated by the arrow 21. A code column CS is expected as the first column to be read out. By way of second reference marks this column contains two first code marks CM1 which extend over two rows, are printed in black and are contained in the first two rows Z13, Z14. As will be explained in the following, these position marks CM1 serve for recognition and electronic compensation of oblique positioning of the document 2. In addition the code column CS contains a pre-printed combination (e.g. of three out of seven elements) of second code marks CM2 which defines the nature of the marking matrix which is to be scanned and/or the nature of the document.

If the marking matrix is divided into sections by division lines printed in drop-out colours and running in straight lines throughout the entire matrix between columns and rows, and if it is arranged that an equal number of marks - which number is dependent upon the nature of the document - is always to be applied within each section, a plausibility check in respect of a specific number of marks can be carried out for all the sections.

To return to Figure 1, when passing the mark readout station an image of such a document 2 is focused by the scanning optics 10 onto an autoscan row, i.e. a row of 128 photo-diodes of the opto-electrical transducer 11. The row of photo-diodes is designed to extend in a direction parallel to the columns S of the marking matrix. A scanning circuit in the transducer 11 transforms the information from the photo-diodes into a serial analogue video signal VS.

In order than one can safely recognise both weakly written marks and marks on dirty documents, the brightness of the character background, i.e. the current maximum white value in the row of the image point which is to be scanned, is taken into account in a signal conversion unit 3 during the quantisation of the video signal VS to form an image pattern consisting of black-white values. From the established brightness of the character background there are taken three black-threshold values which are staggered in level and with the aid of which three different image patterns are obtained from the original video signal VS. A fourth special threshold is used to evaluate the high-contrast pre-printed timing marks TM contained in the rowTZ and the code marks CM1, CM2 contained in the code column CS. This process of deriving threshold values from the established background brightness primarily evaluates the contrast prevailing between character and background and simultaneously compensates any difference in sensitivity between the various photo-diodes of the opto-electrical transducer 11.

In the following stages the image patterns, quantised with the various threshold values, are further processed in parallel. In a preprocessing unit 41 the information concerning the position of rows Zn and columns S of the marking matrix is obtained from the position of the timing marks TM. In addition any oblique positioning of the document 2 is evaluated and compensated by means of a circuit arrangement for correcting oblique positioning which will be explained below. As the position of the rows Z1 to Z17 relative to the scan is then accurately determined, the scanning information can be reduced by the factor 1:2, i.e. to 64 image rows, without any noticeable losses in recognition accuracy. The data required for this reduction in information is taken from a read only memory which is referred to as parameter store 42.

Three image patterns formed by differing evaluation, and the information concerning the position of the image patterns relative to the scanning matrix now arrive in a recognition unit 5. This contains logic circuits which, in order accurately to determine whether or not a mark is present in a specific position forms, for each image point, in dependence upon its position, logical combinations of the image point and selected image points of its surroundings. The control data required for this purpose are likewise contained in the parameter store 42. In this way the arrangement of the marks is established for all the image patterns.

In a following analysis unit 61, a plausibility check is carried out in respect of each section of the marking matrix. For this purpose a previously known item of information concerning the marking sections is used as criterion. This is the known size of the marking sections and a number of permissible marks M in the marking section. This criterion is stored in the form of test programmes for various types of document in a further read-only memory, the plausibility store 62. The information of the code marks CM2 contained in the code column CS determines the test programme which is to be used. The plausibility check now indicates which of the three sets of image data, obtained using the three different threshold values, is fulfilling the plausibility criteria and can thus be output as a permissible result.If it is not possible to carry out a plausibility check because no criteria required for this purpose can be given, the image pattern digitalised with the middle threshold value is analysed and the mark result thereof is output.

Via an output interface 7, the determined marks are output as a bit pattern in a sequence of in each case three 6-bit words per column S, one bit position in the last word being redundant so that one bit position corresponds to each row. The output mark information is then interpreted by a data processing system.

The functional reliability of this device is checked by internal monitoring devices. Following each switch-on of the document sorter to which the device for automaticaliy recognising hardwritten marks is fitted, and following a fault standardisation, a process for the recognition of a simulated document with a full marking matrix takes place in the recognition device. All the data required for this simulation is taken from a further read only memory, a simulation store 8. All the described processes again take place during the processing of this simulation data. The established mark results are compared with theoretical values stored in the simulation store 8. In the absence of conformity a fault signal is emitted.

As has been briefly explained above, the vertical position of the timing marks TM is determined in the preprocessing unit 41 row identification by means responsive in operation to image data obtained from the first reference marks to produce row assignent data indicative of the correspondence between rows of the marking matrix and rows of the scanning matrix. Those parts of the preprocessing unit 41 required for this purpose are illustrated in the block diagram of Figure 3. The signal conversion unit 3 has again been shown to establish its relationship. In the above it has been briefly mentioned that the video signals VS are digitalised with three different threshold values and that a fourth threshold value is provided for the digitalisation of the timing marks TM as these pre-printed marks contrast strongly with the white background of the document.These items of scan data, digitalised with the fourth threshold value, are input into a first register block 310. This consists of six shift registers each comprising 128 bit positions. Thus it is able to store the complete scanning results of six consecutive scanning columns. In the following the terms "column" and "row" refer to the columns and rows of the scanning raster. These items of scanning data are supplied row by row to a first adder 320. This adder consists of a programmable read-only memory and provides, as a binary value, a row sum which indicates how many black scanning elements are contained in these six elements of each row. These three-digit row sums are conveyed in sequence to a second register block 330 which consists of five three-bit registers which are coupled to one another at inputs and outputs.Its stored contents thus corresponds to the row sums of five consecutive rows.

The row sums from the adder and the first two registers of this second register block 330 are output in parallel to a first area adder 340. This likewise consists of a programmable read only memory and supplies the number of items of black scan data in a section comprising six rows and three columns. The output data from the last three registers of the second register block 330 are fed to a second area adder 350 which corresponds to the first area adder 340 in terms of construction and function, but forms the area sum for the three rows lying immediately above.

The four-bit output signals of these two area adders 340 and 350 are fed in parallel to a comparator device 360. This comparator device determines the row position of the centrepoint of a timing mark TM by comparing the supplied area sums. This comparator device likewise consists of a programmable read-only memory and serves to analyse the data, supplied to the two area adders, relating to a section comprising six scanning columns times six scanning rows. The size of this 6 x 6 scanning section matches the size of a timing mark in the scanning pattern. By way of simplification the area sums with which it is supplied from the two area adders 340 and 350 have been referenced B and A respectively.

From its output 361 the comparator device 360 supplies a "1" signal to a first, six-staqe outout shift register 362 when the upper area sum A is smaller than a first constant K1. This condition indicates that a signal should occur at the signal output 361 for an upper white range when scanning elements evaluated virtually entirely as "white" have occurred in an area, comprising six scanning columns and three scanning rows, above a timing mark TM. In order to allow for faults, the first constant K1 can for example have a value of 3. The shift register is clocked at a rate such that its output is delayed relative to its input by six scanning elements.

A second output register 364 having three stages (and hence a three element delay) is connected to a second signal output 363 for a centre-of-gravity signal. A "1 " signal is to appear at this signal output 363 whenever two conditions are fulfilled simultaneously; namely that the upper area sum A is larger than the lower area sum B, and the sum of the two area sums A, B is greater than a second constant K2.

In the scanning pattern a timing mark TM generally covers a section of 4 to 5 columns and 4 to 5 rows. If one can picture the fact that this timing mark TM is "moved" over a selected zone comprising six columns and six rows of the scanning pattern, at some time the centroid of the timing mark, i.e. the largest number of black elements, will cross the horizontal central line of this observation window. At this time the first of the two conditions is fulfilled, i.e. A > B. In order however to avoid misevaluation, the second condition, i.e.

A+B > K2, which establishes a minimum value for the timing mark, e.g. K2=12, must be simultaneously fulfilled. Finally the comparator device 360 also possesses a third signal output 365 which carries a "1" whenever the lower area sum B is smaller than the first constant K1. Thus this indicates a white zone beneath atiming markTM.

The outputs of the two output registers 362 and 364 are connected to the J- and K- inputs of a release flip-flop 366. This flip-flop supplies a release signal FS whenever an upper white zone has been recognised in the comparator device 360. One clock pulse following the report of the centre of gravity of a timing mark TM at the output of the register 364, the release signal ceases. This release signal FS is fed to a coder 367.

In addition, the signal outputs of the two output registers 362 and 364 and the signal output 365 for the lower white zone lead to a conjunctive logic unit which is represented as an AND gate 368. A timing mark signal TS occurs at the output of this AND gate 368 whenever the corresponding conditions are fulfilled in the correct sequence of time in the comparator device 360, i.e. a timing mark TM has been correctly recognised and white zones have been reported above and below the timing mark. This output signal is also fed to the coder 367.

This coder - a programmable read-only memory - also has an input connected to a row counter 369 which tracks the absolute position of the scanning row currently being processed relative to the scanning pattern.

Thus it contains a seven-digit binary number for one of the 128 scanning rows. In respect of this absolute row statement, the coder 367 is to give the relative position of the timing mark TM and thus also the relative assignment of the individual scanning rows to the rows of the marking matrix. The expectation zone for the height position of the timing mark TM is limited to 16 rows, and therefore 4 bits will be adequate for this relative statement at the outputs of the coder 367.

This position information is fed to a parallel-to-serial converter 370 and output in serial form to a store 380 for position information which store can form part of the image data store. The stored data relating to the position information is fed to a data conversion state 390 in which the relative track format of the image data is produced on the basis of the established position of the timing mark TM. In the attached recognition unit 5, the whole of the image data is then further processed on the basis of its assignment to the mark rows.

In the above it has been explained, making reference to Figure 3, how the relative vertical position of the timing mark TM relative to the scanning position is ascertained and analysed in order to compensate a shift in level of a document being read. The relative position of each timing mark TM in the horizontal direction is also ascertained by means responsive in operation to image data obtained from the first reference marks to produce column assignment data indicative of the correspondence in the row of first reference marks between columns of the marking matrix and columns of the scanning matrix. This procedure is conventional and therefore not described in detail. Although in principle this can provide column assignment data for the entire marking matrix, in practice an oblique positioning of the document will give rise to errors.

With the aid of the circuit arrangement illustrated in Figure 4, we shall now describe the correction of the oblique positioning of the image data for each column. A brief reference will again be made to the expianations relating to Figure 2. As mentioned, a possible oblique positioning of the document is recognised using the position marks CM1 in order that the image data may be correctly assigned to the marking matrix.

The circuit arrangement illustrated in Figure 4 serves this purpose and includes circuitry responsive in operation to image data obtained from the second reference marks to produce column correction data indicative of the magnitude and direction of any difference in the angular position of the columns of the marking matrix and columns of the scanning matrix. The output data from the signal conversion unit 3 is again used. However, the number of scanning rows has already been reduced in the ratio of 2:1, i.e. further processing is based on only 64 image rows. Therefore the circuit arrangement comprises a row counter 401 which in an evaluation cycle calls up one of 64 image rows with a six-bit binary number. The output of this row counter 401 is connected to a coder 402 - a programmable read-only memory - which, in dependence upon the count of the row counter 401, emits row-dependent control signals from its outputs 403 to 408. The control signal output 403 is to carry a control signal Z9/Z45 whenver the row counter 401 specifies a zone from the 9th to the 20th scanning row and from the 45th to the 56th scanning row (counting in the direction of the column scan - i.e. downwards in Figure 2). The control signal output 404 emits a positive control signal Z32 when the row counter 401 allocates to the coder 402 the 32nd row in an evaluation cycle. Similar applies to the other control signal outputs 405, 406 and 407 which each carry positive control signals whenever the rows 64, 3 and 2 have been specified by the row counter 401 in the evaluation cycle.The last control signal output 408 itself comprises five signal lines for correcting the oblique positioning: the significance of these will be explained later. This control signal of the coder 402 serves to synchronise the circuit arrangement for correcting oblique positioning for image data in each individual case.

As in the establishment of the vertical position of the timing mark TM which was explained with reference to Figure 3, here again the density centre of gravity of the position marks CM1 is firstly established. This is effected by means of a shift register 409 which is connected to the signal conversion unit 3 and which accommodates the image data which have been digitalised with the threshold assigned to the pre-printed timing marks or code marks. The image data of vertically adjacent image elements are continuously input into this shift register 409.Outputs from four consecutive stages of the shift register 409 are fed to an AND-gate 410 which is enabled whenever the control signal Z9/Z45 is simultaneously emitted from the coder 402 via the control signal output 403, so that the data are allowed only in to the two zones extending from the scanning rows 9 to 20 and 45 to 56 in which the position marks CM1 are expected.

If four vertically adjacent image elements are black, this is recognised in the AND gate 410 and a first storage flip-flop 411 is set. This storage flip-flop 411 is reset via an OR gate 412 by means of the control signals Z32, Z64. The output of this first storage flip-flop 411 is connected to the input of each of two four-stage shift registers 413 and 414 which are enabled for receiving data serially from the storage flip-flop 411 by the control signals Z32 and Z64.

The parallel outputs of each shift register 413, 414 are fed to a respective AND gate 415,416. These are enabled by the control signal Z2 and evaluate the contents of the shift registers 413, 414. Further storage flip-flops 417,418 each connected to the output of a respective AND gate 415,416. The shift registers 413, 414 are clocked at a rate corresponding to the scanning columns so that the flip-flops 417,418 are set whenever a section of 4 x 4 image points has been evaluated as black in the respective group of scanning rows 9 to 20 or 45 to 56. Thus the output of the second storage flip-flop 417 indicates whether the upper of the position marks CMl has been found. Similar applies to the output of the third storage flip-flop 418 in respect of the lower of the position marks CM1.

An EXCLUSIVE-OR gate 419 is connected to the non-inverting outputs of the second and third storage flip-flops 417, 418. Thus this EXCLUSIVE-OR gate responds whenever only one of the two connected outputs of the storage flip-flops 417,418 carries a signal, i.e. only one of the two position marks CMI has been recognised. The output of this gate 419 is connected to a first binary counter 420 which is caused to count upwards during the time that this output signal from the EXCLUSIVE-OR gate 419 is present by means of a control signal Z3 with which it is simultaneously supplied. In terms of function this means that the binary counter 420 which comprises three bit positions counts up to a count which corresponds to the column shift between the two position marks CM1.As soon as both position marks have been established, the EXCLUSIVE-OR gate 419 disables the binary counter 420. The count reached by the first binary counter 420 remains unchanged during the document run and is reset at the start of the next document via a normalising signal 429.

The two storageflip-flops417 and 418 are also connected to a further AND gate 421 in such manner that the latter emits a setting signal to the setting input of a fourth storage flip-flop 422 whenever the upper of the two position marks CM 1 has been encountered first. This storage flip-flop 422 is also reset by the normalising signal 429. Its non-inverting output thus carries a positive signal whenever the upper of the position marks CM 1 has been established first, i.e. the top of the scanned document is inclined towards the right, whereas the reverse applies to the inverting output.Thus the direction of inclination of the scanned document can be derived from the condition of the fourth storage flip-flop 422, and the magnitude of the column shift of the position marks CM1 is defined by the count of the first binary counter 420.

The further functions of the circuit arrangement illustrated in Figure 4 consist of reading out the image data stored in an image store 425 in accordance with the established oblique positioning in offset column sections, i.e. in stepped fashion. Thus it includes circuitry arranged in operation to retrieve from the store image data stored therein relating to locations of the marking matrix using, for each location in respect of which image data is to be retrieved, said row assignment data for determining the relevant row(s) of the scanning matrix and said column assignment and column correction data for determining the relevant column(s) of the scanning matrix. Firstly a first multiplexer 423 is provided for this purpose. Its selector inputs are supplied with the current contents of the first binary counter 420.Accordingly the multiplexer switches through one of six data inputs to its output. Five of these inputs are connected to the control signal outputs 408 for correcting oblique positioning of the coder 402, whereas the sixth input is connected to the zero level. Thus in dependence upon the count of the first binary counter 420 the first multiplexer 423 switches through one of the five signals which are emitted from the coder 402 and are dependent upon the row counter 401 as counting pulses for a second binary counter 424. The upwards counting of the second binary counter 424 is repeated in each row counter cycle. The relevant count of the binary counter 424 indicates which column of the image store is to be output.

This image store 425 consists of twenty series connected shift registers each comprising 64 elements. It is thus able to accommodate the contents of 20 scanning columns. If, under ideal circumstances, no oblique position deviation has occurred in respect of a scanned document, the 64 items of image data of the same storage column, i.e. the contents of one shift register are read out from this image store in one cycle. In the case of an oblique positioning to the left orto the right having been recognised, however, after the elements of a specific number of rows in the relevant column have been read out, further read-out of the stored image data must be affected with the aid of two further multiplexers 426,427 which are each connected to one portion of the outputs of the image store 425.A selected one of these multiplexers 426 and 427 is activated according to the state of the fourth storage flip-flop 422 i.e. whether the required shift is to the left or the right. Which of the connected outputs of the image store 425 is selected in the relevant multiplexer is determined by the count of the second binary counter 424 whose output, which comprises three positions, s connected to the selection inputs of the multiplexers 426 and 427. Finally the two outputs of the multiplexers 426 and 427 are combined in an OR-gate 428 whose output is connected to the recognition unit 5 and via which the image data which have been corrected in respect of oblique positioning are output.

To summarise, the described circuit arrangement firstly establishes the occurrence of the two position marks CM1 and also determines the number of scanning columns (if any) by which these position marks are offset from one another. In dependence upon the degree of this column discrepancy which determines the count of the first binary counter, via the coder 402 and the multiplexer 423 in each row cycle the count of the second binary counter 424 is increased by one whenever, following a specific number of elements dependent upon the column discrepancy, the following image elements are to be read out from the image store 425 with a shift of one scanning column towards the left or the right, which is established by the row-dependent control signals for correcting oblique positioning of the coder 402.

Figure 4 also indicates that the signal conversion unit 3 possesses two outputs for items of image data which are evaluated in accordance with various threshold values and represent each image element in coded form. Accordingly, in addition to the illustrated image store 425, a further such image store with the corresponding output circuits has been indicated in chain-dot lines. As explained in the introduction, this is necessary because the items of image data are quantised with a plurality of thresholds in order to achieve an optimum scanning data pattern from marks differing in contrast.

Thus, in the apparatus described, a high degree of recognition accuracy can be obtained even in the event of a shift or oblique positioning of the pre-printed marking matrix relative to the pattern of the scanner unit in one or in both coordinate directions as it is possible to compensate the aforementioned mispositioning of the marking matrix relative to the scanning pattern.

Attention is drawn to our co-pending United Kingdom Patent Application No. 7926729 (Serial No.

(VPA 78 P 8032) which is also concerned with mark recognition devices, and in particular with such devices having means for plausibility checking.

To facilitate reading of the drawings, a list of reference symbols is given below.

List of References 10 scanning optics 11 opto-electric transducer 2 document 21 transport direction of document Z1 ... Z17 rows of the marking matrix S columns of the marking matrix MK mark boxes M mark TZ row of timing marks TM timing marks GS code column CM1 first code mark CM2 second code mark VS video signal 3 signal conversion unit 41 pre-processing unit 42 parameter store (ROM) 5 recognition unit 61 analysis unit 62 plausibility store (ROM) 7 interface unit 8 simulation store 310 first register block 320 row adder (PROM) 330 second register block 340 first area adder (PROM) 350 second area adder (PROM) 360 comparator device (PROM) for upper and lower area sum A and B 361 signal output for upper white zone 362 first output register 363 signal outputforcentroid 364 second output register 365 signal output four lower white zone 366 release flip-flop with release signal FS at the resetting output 367 coder 368 conjunctive logic unit with timing mark signal TS at the output 369 row counter 370 parallel-to-serial converter 380 store for position information 390 data converter stage 401 row counter 402 coder (PROM) 403 control signal output for zone release (Z9, Z45) 404 control signal output for row 32 (Z32) 405 " " ,, ,, ,, ,, (,,,) 406 " ,' 3 " 3 (Z3' 407 " ,, 2 " 2 (Z2) 408 Control signal outputs for oblique position correction 409 shift register 410 first conjunctive logic unit 411 first storage flip-flop 412 disjunctive logic element 413 shift register 414 shift register 415 conjunctive logic element 416 conjunctive logic element 417 second storageflip4lop 418 third storage flip-flop 419 EXCLUSIVE-OR logic element 420 first binary counter 421 second conjunctive logic element 422 fourth storage flip-flop 423 first multiplexer 424 second binary counter 425 image data store 426 second multiplexer 427 third multiplexer 428 OR-gate 429 normalising signal

Claims (12)

1. An apparatus for reading a document upon which information has been encoded by the making of markings at one or more of a plurality of predetermined positions of a marking matrix formed by notional rows and columns, the document also having a row containing first reference marks and a column containing second reference marks, the apparatus comprising scanning means arranged in operation to scan the document and produce image data in respect of locations defined by the rows and columns of a scanning matrix, a store in which the image data can be stored, row identification means responsive in operation to image data obtained from the first reference marks to produce row assignment data indicative of the correspondence between rows of the marking matrix and rows of the scanning matrix, column identification means responsive in operation to image data obtained from the first reference marks to produce column assignment data indicative of the correspondence in the row of first reference marks between columns of the marking matrix and columns of the scanning matrix, alignment recognition means responsive in operation to image data obtained from the second reference marks to produce column correction data indicative of the magnitude and direction of any difference in the angular position of the columns of the marking matrix and columns of the scanning matrix and retrieval means arranged in operation to retrieve from the store image data stored therein relating to locations of the marking matrix using, for each location in respect of which image data is to be retrieved, said row assignment data for determining the relevant row(s) of the scanning matrix and said column assignment and column correction data for determining the relevant column(s) of the scanning matrix.

2. An apparatus according to claim 1, in which the row identification means comprises means for forming the row sums of the image data of a plurality of columns of the scanning matrix and, from these, area sums of sub-areas of the scanning matrix are continuously formed, compared in a comparator means arranged to compare these sums in order to determine the centroid of a first reference mark and when these area sums are of sufficient value, the relative vertical position for a timing mark is deduced taking into consideration white zones above and below the analyzed scanning zone.

3. An apparatus according to claim 1 or 2, in which the row identification means comprises a first register block comprising a plurality of registers each of which accommodates the image data of one scanning column and which can be continuously read-out a part-row at a time, a row adder arranged to form a sum indicating the number of elements in said part-row which contain mark information and whose output is connected to a second register block which continuously stores the row sums supplied by the row adder, whereby a plurality of consecutive row sums are concurrently available, a first area adder serving to add the row sums of that half of said plurality of row sums which were formed before the other half, to form a first area sum, and a second area adder serving to add the row sums of the other half to form a second area sum and comparator means arranged to compare the two row sums and, when the first sum comes to exceed to second area sum to indicate that a first reference mark has been recognised, and to record the scanning row at which this occurs.

4. An apparatus according to claim 3, in which the comparator means is arranged in operation to indicate recognition of a first reference mark only when recognition of the fact that the first area sum has come to exceed the second is preceded by the first area sum being less than a first predetermined value, is followed by the second area sum being less than said first predetermined value and is accompanied by the sum of said area sums being greater than a second predetermined value.

5. An apparatus according to claim 4, in which the row identification means further includes a coder connected to the parallel outputs of a scanning row counter and which is supplied with control signals derived from the results of the comparisons whereby the relative position information for an established timing mark and in dependence thereupon the track format for the scanning pattern are derived in the coder on the occurrence of the control signals and are transferred to a recognition unit for analysis.

6. An apparatus according to any one of the preceding claims, in which the alignment recognition means includes means operable, in respect of predetermined scanning zones in each of which a code mark is expected to lie, to recognise the occurrence in a column of a predetermined plurality of consecutive scanning elements containing mark information and to recognise such occurrence in respect of a plurality of adjacent scanning columns as indicating the presence of a second reference mark and recording means operable in the event that the presence of second reference marks is not recognised simultaneously in the two zones to record information indicating in which of the zones a second reference mark was firstly encountered and the number of scanning columns between the recognitions of the two reference marks.

7. An apparatus according to claim 6, comprising a multi-position shift register into which the image data is input in serial form, and having parallel outputs connected to a conjunctive logic element which is enabled during scanning of said predetermined zones and is operable to recognise the occurrence in a column of a predetermined plurality of scanning elements containing mark information, a first control flip-flop which is reset after passage through of one of the given scanning zones, and is arranged between the conjunctive logic element and the serial inputs of two shift registers each arranged for serial input to occur at the end of the scanning of a respective zone prior to the resetting of the control flip-flop, the parallel outputs of each shift register being connected to a respective conjunctive logic element and the outputs of the latter being respectively connected to second and third control fl i p-fiops whereby the output signals of the second and third flip-flops characterise the occurrence of a code mark in one of the two scanning zones.

8. An apparatus according to claim 6 or 7, in which the recording means includes means operable to record information indicating in which of the zones a second reference mark was firstly encountered, so that the direction of inclination of the marking matrix of the document relative to the scanning matrix is recorded, and a binary counter which is caused continually to increase its count as long as only one of the two second reference marks has been recognised, and to cease counting for the remainder of the scan of the document as soon as both second reference marks have been recognised.

9. An apparatus according to claims 7 and 8, in which the outputs of the second and third control flip-flops are fed via an EXCLUSIVE-OR logic element to an enabling input of the binary counter, which is arranged to receive a count pulse at the beginning of each column cycle and whose parallel outputs are connected to the selection inputs of a first multiplexer in which, in dependence upon the signal combination supplied in this way, one of its data inputs is selected and switched through to a further binary counter, the data inputs of the first multiplexer are connected to a control signal output of a coder which carries row-dependent control signals for correcting the oblique positioning, and a further control flip-flop which is itself connected to the control flip-flops which are set on the occurrence of a second reference mark, in such manner that its two switching states each characterise the direction of inclination of the oblique positioning ofthe document.

10. An apparatus according to claim 9, in which the retrieval means comprises two further multiplexers whose data inputs are connected to outputs, which are each assigned to a scanning column, of the store and whose selection inputs - for the selection of one of the column outputs of the store - are connected to the further binary counter and to one of the two outputs of the further control flip-flop.

11. A method for automatically recognising marks on a machine-analysable document, comprising determining the vertical and horizontal positions of mark boxes printed in the form of a matrix on the document relative to the scanning pattern of the device, wherein the positions of the columns of the mark matrix are determined from the position of timing marks which are each pre-printed in alignment with the row direction, characterised in that for each mark column the row-wise assignment of quantised image data stored in an image store to the rows of the marking matrix is derived from the row position of the timing marks relative to the scanning pattern of a scanning unit that the position of code marks pre-printed on the document in the column direction of the marking matrix is evaluated and any column shift is stored, as binary value, with information concerning the inclination direction, for the compensation of the oblique positioning of a document, and that, in dependence upon the size and direction of this oblique positioning, the image data are read out with the correct column in respect of the marking matrix of the document which is obliquely positioned relative to the scanning position.

12. An apparatus for reading a document upon which information has been encoded by the making of markings at one or more of a plurality of predetermined positions of a marking matrix formed by notional rows and columns substantially as herein described with reference to the accompanying drawings.