NO974285L - Machine-readable label - Google Patents

Description

Professional field

The present invention relates to techniques for identification, more particularly identification of labels, and further labels which are adapted for reading by machines (ie computers).

The position of the technique.

There is a need for identifying labels to keep track of merchandise during shipping or manufacturing. An optical or other type of "no-touch" label is preferable and these should be able to be used for the following areas of use;

(a) image reading is carried out under difficult conditions such as outdoors or when parts of the label are damaged, lost, soiled, folded or are in a shadow or the label is rendered crooked or blurred, (b) the exact position and orientation of the label is not defined (c) there may be many or a large number of tags in a registered image (d) a high degree of accuracy during the reading of the tag is required, and (e) the tags are out of range for barcode readers or the like.

Furthermore, it is especially important in cases where it is necessary to use a cheap and relatively low-resolution camera that the label is covered with data annotations that are as large as possible, given that a relatively small label such as 50 mm across is highly preferable.

A particular area of application is the tracking of felled logs (timber) from felling to export. In the felling area where the logs are collected before being loaded onto a truck, it is common in some forests to attach a previously printed label with a barcode to the end of the log where it has been cut in order to be able to identify the tree, which is described individually with regard to to owner, cubic, quality, type of wood and the like. The barcode and the associated data are later loaded into an administrator database. A barcode is not easy to read later when, for example, the log is one of many loaded on a freight truck or in a cradle that is loaded on board a ship. There is therefore a need to be able to read such labels in an easier way.

Purpose of the invention.

The purpose of the present invention is to describe an improved identification label of a readable type for machines, and/or set of improved instructions for reading labels of this type, or at least to be able to offer users a usable choice.

Disclosure of the invention

The various aspects of the present invention appear from the claims. For example, one aspect of the invention specifies a machine-readable label as stated in claim 1, another aspect of the invention specifies a set of labels as stated in claim 4, or another aspect of the invention that describes a method for identifying one or more objects out in the field as stated in requirement 9.

Another feature of the invention relates to a machine-readable tag or symbol for identification purposes, consisting of a matrix of machine-readable identifiers, where each identifier consists of at least one element of readable information, and which matrix of information includes an error-detecting code and where in at least one identifier that serves as a locator to indicate to the machine or computer where the remaining part of the matrix is located.

It is preferred that the error correcting code is a cyclic code capable of handling individual bit errors.

Furthermore, it is preferred that the error code is a BH code or a derivative thereof, such as an abbreviated BH code.

It is also preferred that the label is optically readable and consequently each of the machine-readable characteristics is provided with an optically recognizable and predetermined specific range of brightness.

It is also preferred that the label is capable of scattering or reflecting light and it follows from this that brightness level is synonymous with reflection level.

The label is therefore, for reasons of readability, capable of becoming luminescent by means of a transducer-compatible source of electromagnetic radiation.

It is preferred that the resolution or density of pixels in the device that acquires images for the reading machine is such that an area of at least three times three of the sample elements used by the reading machine is provided to at least partially cover each distinguishing mark.

It is preferred that the label is surrounded on at least one side by an outer edge with a relatively bright surface which, in the preferred embodiment, is at least three sample units wide.

It is preferred that this surface corresponds to the white level of the gray scale.

It is preferred that at least two sets of indicia serve as position indicators.

It is preferred that, upon recognition, the position transmitters are able to transfer information to the machine which defines the position (ie position and/or direction) for the rest of the matrix.

It is preferred that each such feature indicating position consists of at least one matrix element surrounded by a field of matrix elements which are in contrast.

Alternatively, the array or arrays of any of the positional characteristics may be used to indicate a step or steps on a scale of brightness.

As an option, no positioners need to be used, in which case the reading machine will try to decode the matrix of characteristics correctly by a process where repeated attempts are made at different positions and directions.

It is preferred that each cell in the array of computer-readable indicia be optically recognizable and furthermore it is preferred that each indicia has a predetermined level of reflectance.

It is preferred that each cell of the array of computer readable features is provided with one of two reflective contrasts.

Alternatively, each cell of the computer readable matrix is provided with one of more than two reflections of different contrasts.

Each cell of characteristic can therefore as a possibility be composed of a material that has an intermediate level of reflection.

It is preferred that the intermediate level or levels are located between the lightest and darkest levels.

Alternatively, the optical properties of a group of cells that give different levels on a scale can also consist of a type of reflection that is perceived as a contrast level by a number of sensors which in turn consist of more than one set, where each one of these sets has a reaction pattern that depends on the characteristic of the reflected energy in a different way.

It is possible to use the entire spectrum of colors.

As a possibility, each cell can consist of a row/row consisting of several dots that make up a contrast surface placed over a substrate of another contrast surface with varying sizes so that intermediate levels of reflection can thus be simulated.

It is preferred that each cell is of such a size that when the expected "working noise" is superimposed on it, the preferred error correcting process will perceive this interference as random noise rather than a burst of noise affecting adjacent pixels.

It is preferred that the matrix of information-bearing characteristics is read in a consistent manner so that the use of each characteristic consists of a predetermined part of a series of information, consisting of means for detecting and correcting all errors.

It is preferred that the error correcting code is an abbreviated BCH code.

As a possible embodiment, the error correcting code is an entire BCH code.

As a possible embodiment, any other error correcting code capable of replacing the intended information, in the case where the brightness of the cell is degraded, can be used.

It is preferred a writing unit which can produce writable patterns according to a themes abbreviated BCH coded information convention according to the present invention in order to thereby produce a series of unique labels which are compatible with a reading process for a computer.

Such a preferred writing unit for producing such writable patterns consists of a computer capable of receiving a series of characters and converting them into a matrix of cells together with an error correcting code as previously described in this section.

As a possible embodiment, this device for producing printable patterns may involve a translation of the design and the error correcting codes into a language or form suitable for use in a printer device for actually producing the image.

One such preferred language is "Postscript".

A preferred label consists of a damage-protective substrate and a display surface capable of retaining the characteristics of the label.

It is preferred that the label also contains a barcode and characteristics that can be read by humans.

Another overriding aspect of the invention relates to a reading unit which can capture an image from time to time

an area of a field of view, which may contain one or more tags, and retain the image internally in the form of a matrix of sample points in which each of these can be marked with a corresponding point within the image.

A device is preferred for flooding a field of the field of view with electromagnetic radiation at least as long as the imaging of the image is in progress.

As a possible embodiment, a scanning luminescent device can be used to illuminate a field in sequences.

It is preferred that the machine to be read consists of a fixed camera which is electrically interconnected with at least one addressable memory surface accessible to a digital computer working with a stored program and including an output interface.

It is preferred that outgoing information from the reading machine consists of information from each individual label.

As a possible purely physical embodiment of the reading machine, this can be divided into an image archive and an image recognition part, separated by a communication connection.

As a possible embodiment, the reading machine does not need to be optical, this means that it can use microwave radiation or sound (preferably ultrasonic sound) to thus illuminate a field and obtain radiation from the individual areas of fields using preferably phased and /or time-controlled lighting together with a suitable detector.

It is therefore preferred that the symbols and their characteristics should preferably contain varied but controlled reflections of the non-optical lighting.

It is preferred that the machine reader locates the position and orientation of any tag in an image containing one or more tags by first recognizing the characteristic appearance of a set of one or more accompanying locators.

It is preferred that in this case it reads the identified data cell matrix.

It is preferred that the reading machine determines the actual reflection of a label by a process where it is decided which part is the most reflective and the least reflective part and thus consequently scales the apparent reflection accordingly.

It is preferred that the reading machine compensates for unevenness in the illumination of a label by examining the apparent brightness of the outer edge and consequently also compensates for the apparent brightness of the adjacent matrix.

It is preferred that the reading machine should be able to determine the relative steps on a gray scale or color scale, and if any, use the band of reflection contained in the position sensors, so that the use of which as a whole characteristic can be assigned a corresponding level of reflection -tion scale that is in use.

As a possible embodiment, the reading machine may be able to locate arrays of data without the aid of locators, in which case a number of readings are made until it detects a reading that is valid.

Drawings

The following description is a preferred embodiment of the present invention, described only in the form of examples with reference to the attached sketches. Fig.l shows a machine-readable symbol according to the invention. Fig.2 shows a row of symbols readable by humans, a bar code row, and a symbol on a printed label according to the invention readable by a machine ready for use.

Fig.3 shows a block diagram of the method according to

the invention.

Fig.4 shows a gray scale representation and the method of the software for analyzing a symbol according to the invention.

Preferred embodiment.

This invention describes a tag for recognition that is readable by a computer. This mostly optical label is specially adapted to applications where (at the same time the label should be as small as possible): (a) the photographing of the image was carried out under difficult conditions, such as outdoors or where parts of the image are distorted, lost, soiled, disappeared in the shadow, folded or folded, or where the tag is rendered crooked or indistinct, (b) the exact position and orientation of the tag is not defined, (c) there are a large number or many different tags in a photographed image, and (d ) a high degree of accuracy during the reading of the label is required.

For example, an area of use for these tags could be the recognition of bundles of wood or logs. Each individual pre-printed tag (which simply serves as a unique identifier) is attached to the cut end of a log and an accompanying barcode is scanned or an alphanumeric character is recorded. Other information such as quality or owner is also noted and stored in a database together with the corresponding number of the tag, so that when the log passes through a number of stations it can be identified and (for example) the owner can be determined. Typically, such stations along a series of stages are driven away, then shipped, and then transported to a receiving sawmill.

At each of these stations, it may be required that the tags attached to a bundle of logs be able to be read in an efficient, rapid and secure manner and indicate their location at a given time and place to a management system comprising the master database. We have therefore developed a camera system and method for image recognition that is capable of taking pictures of the end of a bundle of logs loaded on a truck or cradle, recognizes all the tags, decodes the information, and provides this on to the main database. On a quay, the logs are usually placed in a cradle using a lifting device. A sling is tightened around the logs and lifted aboard the boat. The sticks can be imaged while they are placed in the bundle.

A prominent requirement is an error rate (especially a bad tag message, in cases where a wrong number has been supplied without this being known), of less than 1 in 100,000 tags. The biggest harm is providing the wrong information to the database, as opposed to providing a blank. Another requirement that conflicts with the previous one is low operating costs and consequently a low-resolution digital camera, possibly with a damaged CCD chip, to capture images. Another requirement that conflicts with the first is the sensitivity of labels that are damaged, distorted or soiled. And another requirement is sensitivity to varying levels of light.

There is therefore a need to design a label where the image area is well utilized given that the characteristics that are readable by humans, and a barcode, are also required. It is a requirement to design it so that it is easy to read for a computer or machine when an error occurs, and further that there are methods to correct errors should one exist.

We have produced a number of 29 bits, with an additional 17 bits available for error correction, by using (for the label given in the example) a 6 x 7 matrix of squares, which can be individually colored white, gray or black. Each cell is the smallest piece of information carried by each individual label. This label uses a ternary number system. Some of the 56 cells are reserved, for labels in examples, for the positioners (see later) and some are reserved for placing a regular barcode. Fig. 2 shows an example of a label that would be used where the "ternary BCH" code 10 0 in the present invention together with corresponding descriptive numbers (202, 207), name of owner and/or logo 206, and corresponding barcode 203 is indicated. By BCH is meant a preferred procedure for error correction. 205 indicates optional indications. 204 is a possible perforation to tear off part of the label to use in another treatment. This label has an outer boundary 201 which really only indicates the edge of the paper or similar substrate for the characteristics.

Fig.l shows the actual area with the data cell itself. The actual tag 100 has no real outer boundaries or limitations, however it is preferred that a space of at least three sample points wide which is preferably in the most reflective state (ie white) surrounds the entire barcode to return detailed luminous variations should this be required . Each individual cell (101, 105, 108, 109) is shown here as a dark grey, a variegated or a white square. (We cannot display black squares in this description due to limitations in patent drawings - no filled black squares are allowed). Nevertheless, in printed labels the darkest areas are actually black (although as will be shown more fully later, a ternary system having three gray levels need not be stretched towards any limit of reflection).

In Fig. 1, the positioners are shown as cells 101, 102, 103 and 104 together with the adjacent contrast-giving area such as area 107. This area is reserved. The data matrix 106 can extend over the position sensors. The area that is not used between the positioners 103 and 104 is reserved in our example by a label for a barcode 203.

In our tests, camera noise has been found to be in the range of + or - 5 in a 256-step level ladder, and combining this spread (which could possibly be noise or internal compensation) with difficult lighting effects in the outdoor environment allows us to conclude that more than one theme recognition code is prone to produce errors. It is of course that in a more controlled situation that a higher multiple of the recognition code may be advisable. However, we will strive to be on the safe side. We believe base-4 or base-5 or more cells are too easily prone to error in our chosen application area, at least, although the amount of information that can be stored increases considerably as the number of levels increases. Future editions of this tag may even include 256 steps in each of the red, green and blue channels thus providing

2<24>"<1>different combinations per cell, if the technology (not to mention fading ink in daylight) of printing and reading allows this resolution cost-effectively. A machine edition of human-visible colors with at least two channels, (and perhaps also extended to infrared or ultraviolet) can be used in any cell. For the present embodiment, a three-part scale made of black "ink" is preferred.

The label is dimensioned with regard to the device that will take the images and intermediate factors (distance, focusing elements and the like) so that each cell at least partially covers three pixels and a spatial resolution corresponding to four pixels per cell along both a horizontal and a vertical line, (or test points, applicable with a scanner and an A-D converter used instead) which are integrated into the machine. The requirement for dimensioning allows the rejection of overlapping pixels in a transition in the label or their possibilities for a more advanced analysis, or in a reconstruction of a label where the label is compressed. (It is usually not possible to repeat a photograph in the target environment). It also allows for a certain amount of wiggle room when focusing the label on the image plane for the sensor, or during movement shakes or the like. It also allows some wiggle room for using cameras with defects in their CCD rows. Such defects are well known and consist of individual cell defects, row defects, and/or column defects. (Defect-free cameras are available, but dust spots or defects in crystals can occur from time to time during manufacture and it is thus the case that defect-free cameras must be sold at top prices.)

The outer space (see later) is specifically used to determine the degrees of illumination of any label. It should preferably be at least one square matrix (or three pixels in the reading machine's camera) wide, although this can advantageously be wider.

Positioners

We prefer to use at least two, and preferably four sets of identifiers such as 101 with 107 as positioners. Each set of position-giving characteristics consists of at least one cell or matrix element 101 surrounded by a field of contrast-giving matrix element 107; and generally the surrounding fields are white. We make one pair of adjacent positioners 103, 104 assume black level, and the other two are assumed to be gray level. The software is able to determine the intended reflectance for any cell given the actual brightness of the locators and their adjacent fields, thus compensating for variations in light or exposure that otherwise deviate from accuracy.

It is preferred that the positioners, when recognized, are able to provide the reading machine with information that defines the position (ie position and/or direction) of the remaining matrix. The software uses less than one pixel accuracy to determine the position of the positioners.

As a possible embodiment, no characteristic is used in the cases where the reading machine tries to interpret a sensible one

postulated matrix by a method of repeated trials at different positions and directions.

BCH code.

The preferred BCH (Bose-Chaudhuri-Hocquenghem) family of codes, which were developed in 1959 and 1960 can be considered a general form of Hamming codes for the creation of multiple errors. They are cyclical, constructive codes suitable for communication channels where errors affect subsequent symbols independently. The well-known Reed-Solomon codes seem to be a special case of BCH codes.

We have developed an abbreviated (47, 29) BCH code to handle our preferred rectangular array of data cells (in 100) consisting of 30 ternary (3-level) cell blocks representing 47 bits of binary data. This BCH code has a guaranteed minimum distance of 7, so that at least 7 bits must be wrong before a code can be categorized as wrong. The abbreviated code does not cover all 26 million different variations in numbers, but is nevertheless sufficient to achieve the desired level of accuracy. We use 5 (binary) bits for the letter and 24 bits for the number which are adapted to a 47 bit BCH address area - in a spatially spaced format.

Of course, in some cases error-correcting codes are not necessary. In our application, it enhances the reliability of reading methods for damaged tags, but still allows the complete size of the tag to be smaller than if one were to use other methods such as data duplication.

We also deploy our matrix so that straight lines are minimized, we use the so-called Knight's Walk method to deploy the cells.

A complete system.

Fig. 3 shows a data processing system 300 for identifying logs. A device 301 for producing preferably unique patterns of "ternary BCH" symbols that include data, an error code, and position sensor, with a label of another material (see Fig.2) according to the present invention sends data to a printer 3 02 A preferred writing language is "Postscript". The printer can be located in the felling field, or simply prepare large piles of labels for use out in the field. 3 03 illustrates a bundle of marked logs in front of a camera 3 04 which sends a digital signal to the Wide Area Network user interface 305, which in turn brings the signals to a receiver 306, and then to an image recognition computer 307. This reproduces the data 308 as they originally existed in the ternary BCH symbols.

Writing

It is preferred that a device 3 01 which produces data cell matrices on the basis of the preferred ternary BCH code should be placed where the labels are written, so as to be able to produce a series of unique labels. A computer controlled printer 302 is preferred and one type is a conventional laser printer which uses toner to preferably a paper which is resistant to damage, while another preferred type is the "Printronics L5024" which uses a xenon flash lamp to fuse the toner to a substrate that may contain vinyl, a material that has a melting point that is lower than the temperature that occurs in ordinary laser printers. This machine has the additional advantage that the black is matte. Black that glitters will be perceived as white by a camera that has a side-mounted flash. As a possible embodiment, each cell that is printed can be assembled as a series of dots on a surface and placed over a surface with a substrate on another contrast surface with variable proportions so as to be able to simulate intermediate levels of reflection; a gray level. We use dots that are approximately 1 mm<2>, which are shown in Fig.1 at 101, for example. As the labels will be illuminated mainly perpendicularly along the optical axis, the black should preferably be matte and non-reflective.

The camera.

We need a portable station for data capture so that one can move on a transport area such as a quay; and one imagines a person carrying an equipment pack on their back and holding the camera. We prefer to use a battery-powered handheld camera with a CCD chip and an X-Y array with, for example, 1000 pixels in height and 1500 across. A "KODAK" DCS 42 0 camera (IR version) with 153 6 x 1024 pixels is in use. A higher resolution camera will of course have some additional advantages, but cameras of the preferred type are available off the shelf from several suppliers, and insignificant (using active error correction) or minor defects will possibly be found. This camera is assisted by a conventional flash, although an infrared light flash is preferred so that (among other reasons) other people in the area are not disturbed or blinded when the flash is used, especially at night. Some people drive 50-ton trucks with logs near the cradle being photographed.

We prefer to use a Wratten 25 filter on the camera preferably coupled with a filter that blocks infrared light with wavelengths greater than about 800 nm, and consequently allows the 550-900 nm range to the silicon-based CCD device. This red filter allows the user to aim. Red or infrared light can enhance the contrast between the label and the wooden background that normally forms the background. Ideally, a narrowband source and a narrowband filter over the camera lens would minimize the impact of outdoor light on the image, and we are experimenting with suitable filters. Because the camera is expected to take at least 240 pictures in the 4-hour period between each battery change, it is not practical to use a very bright flash (with large batteries) to compensate for the light outdoors.

The camera transfers the digital image data for processing as soon as an image has been captured, although it is delivered with a hard disk with a storage device that can store up to approximately 60 - 70 images. A preferred camera lens is an 18mm fisheye lens as this allows the user to approach the cradle or timber bundle and illuminate it well with a flash lamp, and minimizes focusing errors which are particularly caused by an irregular or uneven surface on the lens. The CCD sensitive area consists of a small central part (about 10 x 20 mm) of the 24 x 36 mm image area in a custom "Nikon" camera and a normal or telephoto lens would need too large a working distance for this use.

A reading device that scans an area with a pencil beam from a laser can in some use cases be used, but the total reading time for a handheld device would tend to be too long. It may therefore be more appropriate for a more controlled environment, such as a fixed camera platform that reads tags on smoothly moving objects such as railway carriages.

Non-optical images can be used in some areas of use and it is, for example, possible to read suitably designed labels on the inside of structures by using microwave radiation or ultrasonic holographic techniques. This can be useful for finding marked sheets of paper inside a stack, for example in an office.

Transferring the data to the image processing computer.

The output image information from the camera will not (in the preferred embodiment) go directly to a memory store for image analysis, there is an intermediate transfer step. The camera's output information can be compared with other additional information. For example, we can digitally decode variables such as the environmental impact of the portable batteries and transmit these to the base station. The boat and hold that the log cradle is unloaded into can be written on the cradle so that this data can be stored. As the computer that analyzes is relatively large and complicated, it is located far away and the data is transferred to a large wireless "Novell" network system 305 which operates at around 2.3 GHz with a 200 Kb per second transfer speed. The image data is (currently) on an uncompressed TIFF formatted file and each image is approximately 1.5 Mbyte in size, so the transfer time is approximately 8 seconds. The camera is connected to a dedicated laptop placed in a small case. The case with batteries is placed in a portable bundle. It works with 12 volts and is equipped with a self-starting booting program that downloads the operating systems onto the network, and then transfers image data together with additional information such as the number of logs and the camera and the condition of the camera battery to a base station 306. Number of logs can be independent of, possibly the user uses a "sheep counter" or perhaps a stick-shaped device with a counting mechanism which also sprays a paint spot on the stick when it is pressed against the device.

In time we will perhaps develop the computer that processes the images to such an extent that it is equipped with a camera and this would involve the beneficial faster notification of an unacceptable photograph.

We may also use software that compresses the image before it is transmitted. "Novell" network software takes error situations into account.

Image processing computer.

We currently use a "Digital Equipment Corporation" Alpha "AXP" image analysis computer to extract information corresponding to the tag currently in the field of view. This computer 307 completes the process within about 5.5 seconds, while the design of various reports and the drawing of lines on an image occur simultaneously. Lists and tracks for troubleshooting are available. A production release would normally allow the user to do the intervention and is therefore faster. We prefer to compile the same program with different switches so that we can develop special needs for one specific purpose and thus create a production edition from the same software.

Information into the computer preferably takes place through a "Novell" network and the outgoing information 3 08 can also be sent through a network or modem to administer the database located in a remote field. An image analyzing computer can handle images from several cameras that are in use at the same transport yard, which can for example be a quay where several cradles of timber are loaded at the same time.

Software.

Assuming the image has been loaded as a copy of the camera's pixel array (dimensions from about 1024 x 1024 to about 4096 x 4096 bytes, preferably 1024 high by 1536 wide), the software that extracts information from that image is best described in a series of steps.

We assume that any valid tag will be of a certain approximate size, whereby the image of each patch is approximately 4 pixels by 4 pixels, so as to encompass an entire cradle of logs in a field of view. The person handling the camera is told to stand at a certain distance from the object. It is obvious that other sizes can be accommodated either by the program using scanning, where one possibly starts with the last suitable "zoom factor" or by programming another constant zoom factor. Our preferred camera (which has a limited number (153 6) of image pixels across) provides a square of approximately 3 mm<2> (referring to the label) pixel size.

A large number of camera pixels will produce (a) smaller tags, or (b) a more accurate reading with more redundant data, or (c) more logs in a cradle, but on the other hand, 1.5 Mb of data is faster to transfer and analyze than 16 or even 4 Mb.

The description below assumes that four position sensors are used for each label. It is thus obvious that minor changes to the algorithm will facilitate other numbers of positioners.

It is also obvious that it will not be searched for outer limits as a method in itself to be able to locate tags. Some prior art computer-readable labels rely on outer boundaries to determine location, direction and, in some cases, information about distance between cells. The only case where the edge of the label is used in our system is when variations in the illumination of the surface of the label are investigated. The positioners are used to define the position and direction of the accompanying information. 1. The image (which may be reconstructed at this point should a defect be discovered) is scanned over perhaps every single line or row to locate dark spots of approximately the right size and characteristics for the edge. The X and Y coordinates of each detected spot are stored.

In more detail, a suitable spot is where the brightness decreases from a lighter background level to a darker level in the spot, maintains this dark level for at least one or two pixels, and then rises back to about the same level. A graph showing displacement versus brightness would have a trapezoidal shape. The computer can "hunt" by comparing adjacent pixels to find the point with the lowest brightness to get a first approximate x and y point.

The program is quite flexible at this stage on the actual value of a bright area or a dark spot, so that varying lighting conditions within a single image can be handled. For example, a short log with an attached tag might be rejected as a normal-sized log, so its tag would appear much darker than the others. 2. The actual center of each spot is located with less than 1 pixel accuracy by procedure that indicates the center of gravity by averaging the product arm x density in both directions. At this stage, the spot is examined for other acceptance criteria (such as symmetry as asymmetric spots do not resemble preferred locators, a contrasting white (usually white) locator edge, and the like). An accurate indication of the center helps to recover data from centered pixels in each cell of the information matrix. 3. Spots that are in close proximity to each other are now recognized and attempts are made to form "pairs" of spots that have an approximately suitable distance to each other that coincides with the expected distance between

positioners.

4. It becomes easy for "three-leaf clovers" of spots where the outermost spots are perpendicular to (approximately) a spot in the middle. 5. Rectangles are formed by trying to find a fourth spot at a suitable distance to both of the other spots. Note that we are tolerant of the actual accuracy of the rectangles, it would be more correct to call the formed areas squares. Errors in the lens, labels that are not perpendicular to the optical axis, and some damaged labels (for example) will contribute to valid, but not rectangular data sets. 6. Unique rectangles are thus formed and a list is saved, while apparently overlapping rectangles and other such errors are put into a "too difficult" group. 7. If you take each rectangle one by one, a data area is created inside this and the data cell is catalogued. In particular, depending on the density of the positioners and the immediately adjacent white area, a scale of what is black (ie falls within a certain range of densities), what is gray (a distinct and larger range), and what is white (another distinct and even greater span) is constructed. 8. The data cells are then cataloged again. Now a partially shaded label will appear clearly. These are also placed in a "too difficult" group. 9. The data cells are read as ternary information in a particular order, after the direction is determined by examining the relative levels of darkness of the positioners, which indicate direction. 10. The data cells for the "simple" tag are sent to a decoder which extracts the current ternary information, applies error correction, and reports the information in the usual decimal form. At this stage, error correction such as a mud splash or similar can be detected and corrected by a BCH error check. In one example : Data sent to the decoder : 101001 002200 0200220 2211011 0220011 1011021

The debugger reveals that : 2211011 should be 2111011 and changes this ternary bit. Possibly the mud stain made the cell appear darker. 11. The program returns to the "too difficult" group and attempts to categorize the data cells. Some of the content in this group need not be tags at all - just random identical sets of pixels. Specially shaded tags are handled by "walking" around the edge which is always a white stripe, collecting the density of dots. Once a shadow is detected, you can compensate for the data cells located there. 12. Splashes of mud over the data cells are reported as best as possible and all issues of "too difficult" labels are transferred to the decoder.

In the decoder, the error correction algorithm (BCH or equivalent) is used, even if too many bits are lost and the information may become unreadable. The decoder takes into account the direction indicated by the relative intensity of position-giving spots.

This procedure is relatively fast because normally none of the time-consuming processes of image enhancement or image reconstruction involving the use of adjacent operations are used. Our digital camera's incoming information is essentially noise-free. If in an application the incoming data should turn out to contain noise, methods for calculating averages, methods for ranking and the like will be used in an early phase, or at the same time the positions of the labels have been approximately determined, so that the image or interesting parts of it could be cleaned up.

Fig. 4 illustrates at 400, outgoing information from a further developed option of the program. For the label in this example, the sketch is inverted and was optically distorted with lens aberration. The size of the individual pixel that was used to take the picture is clearly shown as small squares, this sketch only shows a smaller part of the total picture that was taken.

In this example, the software has taken out four position sensors; 402, 403, 404 and 407. It has overlooked other dark masses that have not satisfied the criteria for a positioner, which criteria include a given range of sizes, and a contrast zone 4 05 that encloses each positioner. Dark masses that were overlooked include a data matrix 401, the barcode 410, and another dark material 409 in the image. The software has confirmed that the positioners belong to each other, after detecting the center at their centroids, by constructing a rather approximate rectangle (lines and diagonals are shown as 408). The software has then recognized a matrix of 6 x 7 points, and marked these with crosses 4 06 above the expected centers for data-bearing cells. On the computer screen, these crosses are red, green or blue corresponding to the ternary values (0, 1 and 2) recognized by reading the value of the underlying pixels. This is not shown in the procedure for assigning a point on a gray scale to each cell (which could possibly be shown as a frequency diagram) or compensating for uneven light.

Attempt

The entire system has not been tested out in the field when this application was submitted. A "worst-case" trial was arranged to check the failure rate, which should not be more than 1 in 100,000. A photographer went to a dock and took photos of about 105 barcode tags that had been subjected to some kind of damage in the field, from a sample of approximately 70,000 tagged logs. We made a transparent grid with a measured matrix to find out how many cells (they had the label printed on them) were designated as poorly decoded. We assumed the use of the preferred BCH code which in 47 binary bits fit into 30 ternary bits placed in a row in the cell of the label.

If there are misreadings by the computer, we can make an estimate of incoming cells with errors of this size of 0.09%. We estimate the probability of misreporting of data cells after error correction to be in the range of 1 to 200,000. As 3-bit error correction tends to raise the overall risk of errors in error correction. We simply prefer to reject a tag should more than 2, or preferably 1 bits need to be misaligned. This study ultimately resulted in a reporting error rate of approximately 1 in 200,000, and 99.76% of tags were read correctly.

Variations

Most variations have been discussed in the text as they have arisen, but we can still highlight the possibility of using non-rectangular cell markings or features: such as more prominent positioners, which can be circular or elliptical or any other shape ( the company logo could possibly be one of these). Reflective position sensors can be used to increase the contrast. Labels printed on a retroreflective surface (ie, a surface that essentially sends the light back to its source) can be used, in which case the intensity of the light flash will be reduced, which will at least help the person carrying the battery pack.

We will again state that a ternary base for cell information is preferred for this particular application, and a binary or higher base can be used for other applications.

Benefits.

The advantage of the preferred embodiment of the invention is: (1) While maintaining a small area, the symbol or label can be recognized and read accurately in most situations. (2) This applies even after degradation in the field and the image is read outdoors.

Finally, it should be noted that various changes and modifications can be made without departing from the object and purpose of the invention.

Claims (9)

1. Machine-readable label capable of being used in connection with an article, which label comprises a number of machine-readable characteristics, in that each characteristic can be distinguished from the background or the adjacent characteristics by using means for machine reading, characterized in that said machine-readable characteristics consist of: an array of machine-readable cells forming a data field, each cell containing at least one element of readable information, and in which array of cells includes an error correction code and at least one cell serving as a locator to provide the reading machine the location of the data field. 2. Device according to claim 1, characterized in that the position cell or each individual position cell is surrounded by said background field. 3. Device according to claim 2, characterized in that the label has one or more position-giving cells that contain directional information about the direction of the matrices relative to the location of each position-giving cell. 4. Device according to claim 2, characterized in that the label contains at least three position-giving cells separated from the field of the matrix by said background field. 5. A set of machine-readable labels, each label able to be used on an object, each label as according to claim 4, in that the configuration of at least one of the position-giving cells relative to the data field of each label is identical within the set . 6. A set of machine-readable labels according to claim 5, characterized in that the information on each label varies from the information on all other labels in the set. 7. A set of machine-readable labels according to claim 5, characterized in that each of the machine-readable labels is equipped with one or two contrast reflections relative to the reflection of the background. 8 . A set of machine-readable labels according to claim 5, characterized in that the error correction code is selected from a group which contains BCH codes, and a code derived from the BCH code. 9. A procedure for the identification of one or more objects out in the field, which consists of the following steps, characterized by marking the one or more items with a machine-readable label selected from a set of labels according to claim 6 and data fields containing relevant information for the respective item, registration of at least one image of at least one label on the field, transforming at least one image into machine compatible form, using machine vision algorithms to detect the presence and position of the or each of the position-providing cells, calculation of the position and direction of the data field, uncovering and decoding information-bearing characteristics in the data field to identify each object.