DE19534009C2 - Method for reading barcodes moving very quickly past the reading device - Google Patents

Description

The invention relates to a method for reading very Barcodes quickly moved past the reader.

Reading barcodes to identify individual ones Objects are already common, so can Many supermarket checkouts already have one on the product brought barcode to be scanned for identification. For this scan, the barcode must be in the scanning area  of a scanner for a certain period of time will.

An application in which reading barcodes also the processing of the objects or an elaborate Identification of the items could be easier the processing of returns from unsold newspapers, Magazines, paperbacks, novels, comics and puzzles point of sale specific for individual titles. Middle one Large press distributors have to pay half a million a week on such remitting ends. The ak current press range of the federal German market about 2000 objects, so that through the differentiated Er apparently early and later missions the number of the titles to be considered to more than 15,000 can swell. Because the remission processing of this large flood of titles is very personnel-intensive and big Objek Smaller runs often do not include sales determined so that the market transparency is missing.

Competitors have optronic for this purpose Remission processing proposed that the automati fully under secure data collection in remission processing with the help of an image recognition of the front pages of all press products creates. This pictorial representation requires a Ge velvet scan of the magazine, as well as the comparison of the resulting optical pattern with pre-stored values. This pre-storage of certain Mu stern is complex, and especially in the daily newspapers with their fairly similar front pages, this requires some effort high computing effort.  

Nevertheless, this plant is already capable of approximately 8000 Recognize copies per hour. A subsequent one Depending on the detection, the copies are ejected to a maximum of 49 Sorting enables goals. Nevertheless one appears higher throughput is not possible because of the Re performance the evaluation speed of the quite com complicated and large data volumes is complex.

The aim is therefore to standardize on the Ti existing barcodes on the remission system plaren to scan. The problem arises that the barcodes both horizontal (like a fence) and also vertically (like a ladder) on the front pages are ordered. In addition, essentially the whole Title page can be scanned for such a barcode.

EP 0 565 214 A2 already describes a barcode Reader for remitting, where a variety of scan scanners and line scanners is used make the barcode reader so expensive that already in the Patent specification is recommended, preferably 10% of the items to be reoriented by hand as the highest expansion stage to use the reader. The necessary for this device Data processing facilities are either very complex, or they slow down the entire sorting process process significantly.

Furthermore, one is already known from US Pat. No. 4,873,426 a line scanner provided device known in a first circuit barcodes on the title page recognized, however, the barcode area is only on due to its pixel sequence in a one-line scan knows. This has the disadvantage that normal  Writing in a scan line the impression can be generated can, it is a barcode.

The invention has now set itself the task Barcodes that pass a reader very quickly be removed, with little equipment, above all however, read quickly.

It is now possible to change the title page into a number of sub-units to be considered, each  because scanned with a barcode reader, however this is still very computationally time-consuming and high computing power several times because the scanning a barcode or calculating the information the sampling data is not trivial. Since also during the scanning of magazines to be transported further should not be stopped in between, but yourself with a moving speed of currently Moving about 1.5 m per second is special scanning with barcode readers in both horizontal and also in the vertical direction on certain surfaces of the Magazine copies a problem. As a boundary condition After all, very fine lines should be recognized where at must be required that per thin line width at least three pixels are recognized for reliable detection will.

According to the invention, this task is accomplished by a Vorrich tion solved with the features of claim 1, wherein It is particularly advantageous that the large number of data which is captured, first of all on a linear structure tures is examined, and only this small subset of data representing the share of a barcode in the total area of the front page corresponds to that of the calculator subject to further processing and decoding becomes. This results in a significantly enlarged Speed of detection as well as the ability to enable decoding with reasonable computing effort lichen.

Because the entire detection area of, for example, 300 mm × 400 mm must be scanned, this will be a time steering camera with a number of pixels, for example Time 4096 elements and a frame rate of more used as 2 KHz. Taking the Bewe  speed of 1.5 m per second are over 600 scan lines with a distance of, for example 0.625 mm generated. With an 8-bit gray scale digitization So there is one magazine cover for each front page a storage area of approx. 2.5 megabytes. This leaves it understand that this large amount of data that much send times per hour, can be significantly reduced got to.

The subclaims give advantageous embodiments the invention again, it is advantageous to use the gray value levels, which can include pixel data from 0-255, by a differentiator on ascending and descending to reduce the leading edges, which are now from minus 250-255 Data can include and then by a parameter threshold for yes / no information (for example only rising edges). These include data now only zeros and ones, yes / no- Information.

In principle, it is now possible to scan the magazine in two directions in order to scan both the garden fence-like and the ladder-like barcodes. However, it has been shown that a transverse scan with an angle of approximately 45 ° to the object movement direction is sufficient to read both the transverse and longitudinal arrangement of the bar codes. In particular, it is proposed to use a flip-flop array with a 45 ° alignment, in which a plurality of flip-flops the data stream of the scan line and further rows of flip-flops the data streams of the previous and the previous one and others Received rows. These flip-flops are implemented, for example, in hardware in a 'programmable array' logic module, as is indicated in FIG. 4 with the flip-flops which receive data in the transverse direction, that is to say with the data in the 45 ° direction is. Of course, larger flip-flop arrays can also be used, such as is actually done afterwards for detection even with alignment tolerances or by a software evaluation.

However, the proposed hardware evaluation has the advantage of being particularly quick and cheap. The output signals A and B are each one if there is a linear structure either in the transverse or in the longitudinal direction. For slightly twisted arrangements of barcodes, for example if the front page is not completely aligned with the conveyor belt, a 4 × 24 flip-flop array that is approximately six times larger in one direction is proposed. In Fig. 5, the then possible interconnection directions for the flip-flops are indicated by the respective fanned beams in the block "matrix".

It is also proposed to use this first circuit special embodiment to connect a counter, by switching on further line retarders and the additive linking of their results decides whether there is a certain minimum number of transitions is whether a linear structure with a more number of black and white changes, or only one small number of black and white changes.

In a further proposed third circuit ( FIG. 7), the line events are enumerated and compared in a comparator with a preset value in order to determine whether the minimum number for a bar code has been reached. If so, a barcode detection result is achieved so that the corresponding data can now be marked in the data stream and its information value can be determined in a further decoding circuit.

In a preferred embodiment, it can also be determined by questioning what the ratio of the thick to the thin lines is, whether this fits a barco-like structure or not. In fact, however, the device functions even without this query, since it is provided that up to eight barcode detections can still be recognized on a front page in the decoding device in a timely manner. In Fig. 2, the case is further shown that in two successive magazines two barcodes in the 45 ° transverse He detection area. In order to clearly assign these to the correct magazines, a detection of the belt nubs between the magazines for transport is also provided by software.

The circuit for noise suppression in FIG. 5 is a further circuit for edge correlation, which is included in a preferred embodiment and has a delay circuit and a determination of the number of edges which is to be below a maximum value. From the combination of the two results, it is determined whether the edges just found are valid straight lines or whether, for example, they are rather lines of text that can also be recognized as linear structures in the angular range typical for barcodes. Due to the noise suppression, these are not recognized as straight lines in the causality check and rejected. With a regular barcode, it can be assumed that two straight lines recognized one after the other must have approximately the same angle. If this is not the case, the currently recognized straight line is interpreted as noise and discarded as a bar code line.

5, which is subsequently shown in FIG. 5, for output signals of the two matrices, is carried out separately for each channel (vertically and diagonally), all essential statistical parameters of the barcode recognition being included. It determines the frequency of the detected barcode lines per area in order to signal the presence of barcodes. The last three lines serve as the area. The averaging is implemented as a counter. The counter reading can change with every pixel of the line, whereby the following criteria must be met by the current counter reading:

Increases by the number within the current Detected barcode lines (each Area section consists of 1 pixel × 3 lines, so that the counter reading increases by a maximum of 3 values per pixel hen). It increases by 1 for every hold exceeded low, with the parameter Hold in the range of 1.255 is freely programmable. It may have the value OnLevel + 1 not exceed, it must not be less than zero who den and it is zeroed as soon as the end of a Barcodes is reached.

This behavior of the counter reading over a line is shown in FIG. 6. This averaging enables barcodes to be separated from previous linear structures, for example changing the background color, in order to print a white background behind the barcode.

Further features and advantages of the invention result from the following description of a preferred embodiment Example using the attached drawing. It shows:

Fig. 1 shows the schematic conditions in the detection area, wherein the data traffic is apparent from the typical geometric standpoint distances,

Fig. 2 illustrates the relationship between the scan lines raster under orientation of the bar codes,

Fig. 3 shows the worst case, two barcode detection areas that are each located in Benach disclosed compartments of the conveyor belt,

Fig. 4 shows the schematic structure of the first circuit for detecting line-like struc reindeer,

Fig. 5 shows the basic structure of an advantageous embodiment with additional noise suppression and averaging,

Fig. 6 is a circuit which detects the presence of a multiple dashed area from the off gave a line-like structure of Fig. 4,

Fig. 7 is a counter that counts and compares the number of line events, so as to make the decision as to whether or not there is a bar code, and

Fig. 8 shows the meter reading over a line at the middle, which results to determine the beginning and end of a barcode.

In Fig. 1, the resulting conditions are shown for the data volume in the detection range of typically 300 × 400 mm with a line camera with 45 ° used with 4096 elements and an image repetition rate of 2.3 KHz. Taking into account the speed of movement of all magazines of a maximum of 1.5 m / s, the area is broken down into 640 scan lines that are 0.625 mm apart. With 8-bit gray value digitization, this results in a storage area for the image data of approximately 2.5 megabytes (640 × 4096 bits), which occurs four times per second.

The resulting resolution essentially becomes determined by case 2, like a ladder, that is Barcode arranged parallel to the direction of movement. For the code recognition must be a mini in the stationary case male line width of 0.23 mm, diagonally 0.32 mm be solved. Taking into account the influence of the movement blur must be recognized to detect 0.23 mm thick Li stationary barcodes with moving barcodes still be increased.

From the maximum required length of a scan line of 425 mm and the selected line sensor of 4096 elements this results in a stationary resolution of approx. 0.075 mm at the 45 ° viewing angle. From the above The minimum module width listed therefore follows Code detection with triple (oversampling) resolution. The scan line spacing perpendicular to each other without movement This is 0.460 mm.  

Regarding the motion blur may be during a Exposure time along the module only 1 to 1.5 pixels of the line sensor. This requires due to the Movement speed for 1.5 m per second a ma maximum exposure time of approx. 70 µs. This short be glade time can be either mechanical or electronic shuttering of the line sensor or by Pulses of the lighting unit can be achieved.

A so-called Morphology processor used as the first circuit. This compares sections of the Hellig one after the other signal from several successive scan Lines and creates an information log if same che signal structures, i.e. a barcode can be recognized. This information protocol contains the location and extent the linear structure. Building on that possible barcode areas become from the scan lines composed (windowing of the line information), the then to decode the code from a decoding one Device can be used.

Fig. 2 illustrates the relationship between the scan line raster with the orientation of the bar codes. It can be seen from this that the orientation of the barcodes does not influence the information content of the detection, since the sum of the barcode width and height determines the scanning area. The maximum number of detection of a module results from the quotient of the module height and the scan line grid.

In the current specification, a module with 5 mm height in 7 to 8 scan lines, which ent speaking a redundancy or measure for the safe detection  tion of a module - hence the entire barcode leaves.

Fig. 3 shows the less favorable case of two barcode detection areas, each located in adjacent compartments of the conveyor belt. The belt knobs are recognized for a safe assignment due to the movement speed of the conveyor belt and the belt knob width with a time certainty of better than 2 ms. If there are no belt knobs, other suitable optoelectronic means must ensure that successive magazines can be viewed separately from one another.

In Fig. 4 it is clear from the schematic representation and in Fig. 5 based on the basic structure that only the edges are considered first by differentiating the determined data and these edge events then as logical 1 and 0 to a flip-flop matrix are forwarded, in the successive flip-flops of one line are supplied with the data stream one after the other, while the flip-flops of the next line each receive the corresponding data determined in the preceding scanning line by a line retarder.

In FIG. 4, as the pixel data from the left is illustrated, Coming, 8 bits, that is, having a value from 0 to 255 to a differentiator be supplied thereafter may accept values from minus 250 to 255, that is both negative and positive edge of events show, then by a threshold value analysis on 1-bit data, which only contain zeros and ones and symbolizing the change in brightness who converted to then be processed in the flip-flop matrix with downstream AND gates. The first AND gate, which produces the output signal A, is thus with the last flip-flop of the last line, with the before last flip-flop of the penultimate line, with the third last flip-flop of the third last line, etc., and with connected to the first flip-flop of the first line so that it describes a diagonal line across the flip-flop matrix, which actually corresponds to a straight line due to the 45 ° orientation of the scan line. The other straight line will extend transversely to it in the flip-flop matrix essentially at a 90 ° angle there. These corresponding flip-flops are marked by crosses. If there is some blurring of the alignment of the bar codes, larger arrays should be used, each with the detection of a certain area, as indicated in FIG. 5.

In FIG. 6 is further illustrated how this inputting can be closed in an AND circuit to a plurality of scheduled events from the off transition A and B signals by inputting in an OR circuit and Zeilenverzögern this result and in turn.

Finally, FIG. 7 shows how these line events can be enumerated positively, while a down counter in the pixel clock counts down to zero. The counter output is then compared in a comparator with a preset value to determine whether it can be a bar code structure. If this is the case, as is shown in FIG. 5, by means of noise suppression and averaging schematically, further criteria are used to define a barcode more precisely, in order to subsequently carry out output formatting, that is to say the image data with information about the position of the barcodes etc. is increased complete, so that a further deco diereinrichtung only needs to evaluate small portions of the large amount of data.

This evaluation can then be done in software, wäh rend due to the required speed preceding units preferably implemented in hardware will.

The noise suppression is particularly by Filter out text caused barcode events. If Noise is present at the matrix output scanned text or the like. usually several valid straight lines at the same time. This uses the noise reduction there to, the straight lines currently found as noise to in interpret. The maximum number of times he knew straight lines that are necessary to recognize a noise can be set using a parameter. In addition, an edge correlation offers another possibility ability to detect noise. With regular barcode it can be assumed that two successively recognized Ge raden must have approximately the same angle. Is this is not the case, the currently recognized straight line is marked as Noise interpreted and discarded as a barcode line. This correlation can be activated in software.

The averaging determines the frequency of the detected barcode lines per area in order to signal the presence of barcode. The last three lines serve as the surface, so that bar codes of different sizes can be recognized in the title pages of different magazines. The averaging is implemented as a counter that can change with each pixel of the line. The current meter reading must meet the following criteria:

• - Increases by the number of within the current Detected barcode line section. Everyone The area section consists of 1 pixel × 3 times len, so that the counter reading by a maximum of 3 values per pixel can increase.
• - If the parameter Hold is exceeded, which is in the range  1-255 to choose. 1 lowered.
• - It must not exceed the OnLevel + 1 value.
• - It doesn't get less than zero.
• - It is set to zero as soon as the end of a bar codes is reached.

A barcode is now considered to be present if the counter has exceeded the OnLevel value and is no longer available if the counter has fallen below the OffLevel value. Both values are freely programmable in the range 0 to 255. This fact is shown in Fig. 8 Darge.

The position information of the areas of the possible bar codes is appended to the gray image data stream. Per line Up to 8 such areas can be identified. Each area is over a 4 byte data block described, and a data block contains next to the Posi indication of the area within the line the information mation, whether it is horizontal or vertical Barcode deals with whether the barcode starts here or ends and an approximate angle. To be noted is that the additional data is a pipeline delay of 2 lines, that is, the position information barcode not to the current line but to  penultimate. Line heard. The angle is made up of 4 bits plus 1 bit, which indicates whether the angle is valid is. To do this, the number of a straight line within a Lines within the current barcode into a 4-bit Coded angle. The straight lines change from small to numbered at large angles. Has vertical barcode the straight line range 0 to 8, the straight line is exactly vertical  4. If no clear angle coding is possible bit 28 is set to zero.

In conclusion, it should be noted that for decoding of the barcode and the correct control the sorting system now for a comparatively long time up to ½ second is available in which the periodical from the scanner to Sorting device is transported.

Claims (10)

1. A method for reading bar codes that have moved very quickly past a reading device with a line scanner, with the following method steps:

• 1. recognition of linear structures with a circuit matrix in a first circuit,
• 2. - cross-line comparison of image data from neighboring pixels and insertion of additional marking data in the image information to identify the corresponding data,
• 3. - Storage of the image information of many scanned lines for the entire area on which barcodes can be located in a memory, and filtering the areas recognized as barcodes from the image data in a second circuit, converting the data into bit sequences and forwarding them to a device decoding the barcode information.

2. The method according to claim 1, characterized in that that the first circuit differentiated, on yes / no Information reduced image data in the matrix from be neighboring pixels.   3. The method according to claim 1, characterized in that the line scanner opposite the direction of movement the barcodes are slanted. 4. The method according to any one of the preceding claims, characterized in that

• 1. the circuit matrix is a flip-flop matrix, the flip-flops of which are connected one row after the other,
  while the next line in each case is processed by a line retarder correspondingly delayed image data of a preceding scanning line, which therefore originate from locally adjacent pixels on the scanning surface which have already been moved in the transport direction,
  those flip-flops, the arrangement of which in the matrix corresponds to the inclination of the scanning sensor, so that data of an object parallel line there are simultaneously processed, are coupled to at least one AND gate in order to determine the presence of lines transverse to the transport direction and / or Longitudinal events to be recorded.

5. The method according to any one of the preceding claims 2  to 4, characterized in that a matrix with we at least several times as many in one direction cell matrix elements is used, which also includes the ei NEN angle of the order of ± 10 ° to the  Transport direction tilted barcodes as lines sums up. 6. The method according to any one of the preceding claims, characterized in that by a counter after the first circuit is decided whether a be agreed minimum number of transitions is present. 7. The method according to any one of the preceding claims, characterized in that the first circuit and the format of the marking data for recording or Identification of up to eight barcodes per scan line are designed. 8. The method according to any one of the preceding claims, characterized by recognizing the between individual Distances bearing barcode objects. 9. The method according to any one of the preceding claims, characterized by segregation of events at which recognize several straight lines at the same time, by means of noise suppression. 10. The method according to any one of the preceding claims, characterized by averaging the output signals from the first circuit to determine the End and beginning of a barcode.