Provided for non-commercial research and educational use only. Not for reproduction, distribution or commercial use.

Transcription

1 Provided for non-commercia research and educationa use ony. Not for reproduction, distribution or commercia use. This chapter was originay pubished in the book Advances in Computers, Vo. 80, pubished by Esevier, and the attached copy is provided by Esevier for the author's benefit and for the benefit of the author's institution, for non-commercia research and educationa use incuding without imitation use in instruction at your institution, sending it to specific coeagues who know you, and providing a copy to your institution s administrator. A other uses, reproduction and distribution, incuding without imitation commercia reprints, seing or icensing copies or access, or posting on open internet sites, your persona or institution s website or repository, are prohibited. For exceptions, permission may be sought for such use through Esevier's permissions site at: From: Amit Grover, Pau Braecke, Kevin Lindgren, Ha Berghe, and Dennis Cobb, Parameters Effecting 2D Barcode Scanning Reiabiity. In Marvin Zekowitz, editor: Advances in Computers, Vo. 80, Burington: Academic Press, 2010, pp ISBN: Copyright 2010 Esevier Inc. Academic Press

2 Parameters Effecting 2D Barcode Scanning Reiabiity AMIT GROVER Identity Theft and Financia Fraud Research and Operations Center, University of Nevada, Las Vegas, Nevada, USA PAUL BRAECKEL Identity Theft and Financia Fraud Research and Operations Center, University of Nevada, Las Vegas, Nevada, USA KEVIN LINDGREN Identity Theft and Financia Fraud Research and Operations Center, University of Nevada, Las Vegas, Nevada, USA HAL BERGHEL Identity Theft and Financia Fraud Research and Operations Center, University of Nevada, Las Vegas, Nevada, USA DENNIS COBB Identity Theft and Financia Fraud Research and Operations Center, University of Nevada, Las Vegas, Nevada, USA ADVANCES IN COMPUTERS, VOL Copyright 2010 Esevier Inc. ISSN: /DOI: /S (10) A rights reserved.

3 210 A. GROVER ET AL. Abstract This paper describes a case study to identify the various parameters that affect the ning reiabiity of 2D barcodes for high content density appications. A review of the numerous options for 2D symboogies showed that, in theory, these symboogies are capabe of encoding reativey arge amounts of data, but in practice, barcode -abiity decreases as the encoded content increases. With specific attention to higher content (or capacity) appications, a case study was performed to identify the 2D symboogies with the highest ning reiabiity, to be termed -abiity, as we as the various parameters that impact this -abiity. The paper is divided into four ogica sections. The introduction section covers the goa and the specific requirements of the project. Section 2 discusses in detai the various parameters that impact the reiabiity of high-capacity 2D barcodes. Section 3 focuses on the testing methodoogy empoyed and Section 4 summarizes the concusions of the detaied testing. Based on our requirements, our findings indicated that the best three reiabiity measurements of 96% 99% were a obtained using the PDF417 symboogy with different media types and ners. 1. Introduction Factors that Affect Barcode Scan Abiity Barcode Symboogies Content Density Error Correction Scan Grade and Refectance Profie Symbo Grade Encoder Printer Media Decoder/Scanner Characteristics Testing Methodoogy Interna Testing Externa Testing Concusion References

4 PARAMETERS EFFECTING 2D BARCODE SCANNING RELIABILITY Introduction A barcode is a symbo that contains encoded paintext data that can be read by standard optica ners/decoders, thereby automating data representation and retrieva whie eiminating potentia human error. Common barcode appications incude tracking of consumer goods at grocery stores, checkout terminas, document management toos, inventory management, ticketing, mobie tagging, mobie airine boarding passes, etc. The coreation between the barcodes and the corresponding messages is termed as symboogy. Symboogies are cassified as inear and twodimensiona (2D). The goa of this case study was not ony to identify the various parameters that affect the ning reiabiity of 2D barcodes but aso to identify the most reiabe barcode and its appropriate configuration for high content density appications. The need for this study is rooted in a project for deveoping a secure credentiaing system, where there is a need to have data encoded in high-capacity 2D barcodes on viz. Poyviny choride (PVC) and poyester media. Initia experimentation showed that the ning reiabiity of these high-capacity 2D barcodes was not as high as a typica grocery store checkout system. Since the average user s typica experience with barcodes is the checkout ine at the grocery store, the expectation from using a barcode is quick and reiabe ning. The primary difference between this expected resut and a second-generation 2D barcode was in the data content density. The typica grocery store items use a very sma amount of data in the barcode, mainy a primary key, which requires referencing a backend database to read the compete information about the ned item. Our particuar appication required eiminating this dependency on a backend database, thereby requiring a arge amount of data to be encoded directy in the barcode itsef. Even though the barcodes that we used were within the theoretica data capacity imits for different barcode symboogies, the ess than perfect reiabiity prompted us to expore other potentia factors. These secondary factors incuded the barcode symboogy itsef, the encoder which conforms to the reevant barcode symboogy, the printer which creates the physica barcode, the media on which the barcode is printed, and the decoder that interprets the encoded content. This case study describes a series of internay designed tests, and interpretation of the test resuts, to precisey identify the impact of data content density, error correction eve, type of encoding, ner characteristics, type of media, and printer on the -reiabiity of 2D barcodes.

5 212 A. GROVER ET AL. 2. Factors that Affect Barcode Scan Abiity As per the Layman s guide to ANSI, CEN, and ISO bar code print quaity documents pubished by the Association for Automatic Identification and Mobiity (AIM) [1], through the years, bar codes had been printed that met the existing standards, but woud not. And often bar codes printed out of specified standards did. Years of extensive testing by different groups from ANSI, Committee for European Normaization (CEN), and ISO, have identified the foowing critica parameters for bar code reiabiity: Aperture and waveength of the ner, refectance and surface opacity of the media, printer characteristics, grade and refectance profie, and symbo grade and the operative ning environment. In addition to these factors, our case study considered other parameters such as the type of symboogy used, the content density of the encoded data, the error correction eve used, and the method of encoding. Each of these factors, as they reate to this particuar case study, is examined in detai in the subsequent sections. The symboogies used may be inear or 2D. Error correction techniques are used to increase the -abiity to hande partia data corruption. The barcode may be printed on different kinds of media such as paper, poyester, and PVC. Each of these media offer different benefits and are typicay seected based on the intended usage profie for the barcode. Once the barcode is physicay printed on the media, various decoders or ners may be used to decode it based on the type of barcode used. 2.1 Barcode Symboogies Whie inear symboogies, such as Universa Product Code (UPC), Code 39, Code 128, EAN, MSI, Inteigent Mai Barcode, Pharmacode, to name a few, were imited in their data-carrying capacity, the 2D symboogies enabe arge amounts of data to be encoded and decoded in machine-readabe formats [2, 25]. Common exampes of 2D symboogies incude DataMatrix, PDF417, Aztec Code, Codabock, MaxiCode, QR Code (Quick Response Code), Datastrip 2D, etc. [3]. Some common exampes of various barcode symboogies are shown in Tabe I (the Kaywa website [20] was used to generate the QR Code). Based on the voume of usage, the symboogies may be cassified as: 1D or stacked symboogies, such as UPC and Code 128. These barcodes are not the focus of this study and wi not be discussed in detai. Symboogies having widespread industry usage such as PDF417, QR Code, DataMatrix, and Aztec Code. Proprietary and Emerging symboogies such as Datastrip 2D [24] and Microsoft s High Capacity Coor Barcode (HCCB).

6 PARAMETERS EFFECTING 2D BARCODE SCANNING RELIABILITY 213 Tabe I BARCODE SAMPLES UPC-A Code 128 PDF417 QR code DataMatrix Aztec code High capacity coor barcode (HCCB) DataStrip 2-D Symboogies Having Widespread Presence One of the most widey known barcode symboogies is the UPC symboogy, which is a inear barcode used for tracking retai merchandise and other point-of-sae management functions. UPC impementation has different variations, incuding UPC-A, UPC-B, UPC-D, UPC-E, and UPC-5, athough UPC-A is by far the most widey used symboogy [28]. It is abe to encode 12 bytes of numeric data wherein 11 digits constitute the data and 1 extra digit is used as a check digit for error correction [4]. Each digit is represented graphicay as a combination of two bars and two spaces. Whie UPC-B does not use a check digit, UPC-D differs from UPC-A in that it uses a variabe-ength code instead of the standard 12 digits used by UPC-A. UPC-E is optimized for appications that require a smaer barcode and produces a compressed code with ony 6 digits as opposed to 12 digits [5]. UPC-5 is a 5-digit extension to standard UPC codes used for encoding retai pricing for books. The printed symbo contains both a machine-readabe part as we as a human-readabe part. The structura breakdown of a typica UPC barcode is shown in Fig. 1. The foowing sections provide a brief overview of the 2D symboogies evauated for this particuar project as we as a comparative summary, which may be referenced for quick interpretation.

7 214 A. GROVER ET AL Product prefix Manufacturer identification number Product identification number Error correction check digit Structure of a UPC-A barcode symbo highighting different parts FIG. 1. UPC-A barcode structure The PDF417 Symboogy. Portabe Data Fie 417, though cassified as a 2D barcode, is in fact a mutirow, variabe-ength stacked symboogy deveoped in 1992 by Symbo Technoogies. This unique characteristic aows it to be decoded by many 1D hybrid ners, apart from reguar 2D ners. The symbo is composed of 3 90 stacked rows. A PDF417 symbo character, or codeword, is the individua buiding bock for the barcode and consists of 17 modues arranged into four bars and four spaces, thereby giving it the name of 417 [17]. The integra sections of the barcode as depicted in Fig. 2 incude ceary defined Start and Stop patterns, the data coumns, and the quiet zone. PDF417 offers encoding a maximum data character capacity of 1850 text ASCII characters, 2710 numeras, or 1108 bytes [2]. This amount of data encoded is a resut of the manner in which the agorithm encodes the type of data, for exampe, numeras require a smaer codeword size to encode than an aphabet etter (Tabe II). PDF417 uses Reed Soomon error correction [6]. Error correction eves are user seectabe and can be set from 0 (zero) for no error correction to 8 (eight), which is the highest eve. This eve indicates the amount of redundancy that is added to the encoded barcode. The benefit is the increased -abiity; however, the downside is that the effective content size is reduced because the error correction takes up content space. Tabe III iustrates the specifications [2] pertaining to codewords and error correction.

8 PARAMETERS EFFECTING 2D BARCODE SCANNING RELIABILITY 215 Quiet zone Start pattern Stop pattern Start and stop pattern, data coumns, and quiet zone Enarged portion of barcode indicates one codeword (depicted by the red rectange) FIG. 2. PDF417 barcode characteristics. Tabe II PDF417 CODEWORDS PER CHARACTER TYPE Character type Character Codeword Aphabetica Numeric ASCII Tabe III COMMONLY USED ERROR CORRECTION LEVELS FOR PDF417 No. of data codewords Error correction eve No. of error correction codewords The QR Code Symboogy. QR Code, or Quick Response code, is a 2D matrix code created in 1994 by the Japanese Denso Corporation. As opposed to a stacked barcode, a matrix code is one in which encoding is based on the positioning of the eements or back dots in the matrix. These are among the most popuar barcodes being used for mobie tagging (providing information using 2D barcodes on ce phones) appications today. The QR Code data capacity for different encodings is shown in Tabe IV [7].

9 216 A. GROVER ET AL. Tabe IV DATA ENCODING CAPACITY FOR QR CODE Data encoding Max. number of characters Numeric 7089 Aphanumeric 4296 Binary 2953 Japanese Kanji characters 1817 Tabe V QR CODE ERROR CORRECTION LEVELS Leve of error correction Data retrievabe (%) (% CW that can be restored) Leve L 7 Leve M 15 Leve Q 25 Leve H 30 Tabe VI DATA ENCODING CAPACITY FOR DATAMATRIX Data encoding Max. number of characters Numeric 3116 Aphanumeric 2335 The QR Code uses Reed Soomon error correction and supports four different eves of error correction as detaied in Tabe V [29] The DataMatrix Symboogy. DataMatrix, shown in Tabe I, is a highy scaabe 2D matrix symboogy popuary used for marking sma inventory artices and eectronic components [18,19]. As shown in Tabe I, it is characterized by an L shaped centering pattern which has soid back ines aong the eft and bottom sides of each of its square or rectanguar data regions [31]. Each data region is composed of a coection of modues arranged in an even number of rows and coumns. Based on data content density, a data matrix code can have a symbo from 1 mi to 14 in. per side [26]. The newer version of DataMatrix supports Reed Soomon error correction and reconstructs data by using poynomia oversamping [21]. The DataMatrix Code data carrying capacity for different encodings is shown in Tabe VI [8].

10 PARAMETERS EFFECTING 2D BARCODE SCANNING RELIABILITY 217 Tabe VII DATA ENCODING CAPACITY FOR AZTEC CODE Data encoding Max. number of characters Numeric 3832 Aphabetic 3067 Bytes 1914 Tabe VIII COMPARISON OF THE DATA CAPACITIES OF VARIOUS SYMBOLOGIES Data carrying capacity (max number of characters) 2-Dimensiona barcode symboogy Numeric Aphanumeric/ aphabetica ASCII/binary/ bytes Japanese Kanji characters PDF QR Code DataMatrix Aztec Code The Aztec Code Symboogy. The Aztec Code, shown in Tabe I, is another popuar 2D matrix symboogy that has been in use since Based on a square grid, its characteristic bus-eye pattern of concentric square rings faciitates quick centering for the encoder irrespective of the barcode orientation. The grid grows in size around the bus-eye center with additiona square modues as more data are added. Based on the data content density, the symbo size varies from 1515 to modues per square. The Aztec Code supports Reed Soomon error correction and the permissibe vaues are 5 95% [9].The Aztec Code data capacity for different encodings is shown in Tabe VII. A quick comparison of the data capacities of these symboogies is shown in Tabe VIII Proprietary and Emerging Symboogies A number of emerging 2D symboogies are optimized for high content density appications. The foowing symboogies were considered but not incuded for actua testing as they incude proprietary eements that did not fit our vision of the secure credentiaing project.

11 218 A. GROVER ET AL The Datastrip 2D Symboogy. Datastrip 2D is a highdensity 2-dimensiona matrix symboogy optimized to hod high-capacity data incuding encrypted coor photographs and other biometric data used for secure credentiaing [3,27]. It was deveoped by Softstrip Systems and was originay known as Softstrip. It can support a data density of up to 1000 bytes/sq. in. Its biggest advantage is that it can store arge amounts of data in a fraction of the space used by other popuar 2D barcode symboogies [30]. However, since certain impementation aspects were not in conformance with the project requirements, this symboogy was not considered for our testing The HCCB Symboogy. Announced at CES in 2009, Microsoft s High Capacity Coor Barcode (HCCB) is currenty one of the newer symboogies that was deveoped with the intention of enhancing the encoding of higher capacity content. Simiar to the matrix barcodes, HCCB uses a grid of coored trianges referred to as symbos to encode data and has an announced capacity of encoding 3500 characters per square inch [10]. The factors that impact on this encoding amount are the grid size, the symbo density, and the number of coors used for the symbos, which is either eight or four. The actua encoding and decoding however is proprietary and must be icensed. Printing may be performed with off-the-shef inkjet or coor printers. Current usages of this symboogy is imited because of its proprietary status and incude the Microsoft Tag, which is a mobie tagging service that accesses product information, and the ISAN-IA, which is a version method for audio and video products. Somewhat unique to this particuar symboogy, the HCCB incorporates tamper-proof quaity through the use of digita signing based on Eiptic Curve Cryptography. A in a, the HCCB addresses the concern of encoding higher capacities and is aso rumored to be more forgiving when it comes to poor barcode image quaity. 2.2 Content Density Testing showed that data content density payed an important roe in the ning reiabiity of 2D barcodes. Barcodes can be cassified based on their data content density or capacity. For the purpose of this case study, ower content density is defined as content density, incuding optimum error correction, that does not exceed 5 of the theoretica data capacity imit of the respective barcode type. On the other hand, higher content density is defined as content density incuding optimum error correction that exceeds 5 of the theoretica data capacity imit of the respective barcode type.

12 PARAMETERS EFFECTING 2D BARCODE SCANNING RELIABILITY Error Correction Error correction is an important feature supported by most 2D barcode symboogies. This invoves encoding additiona data in the barcode that heps in reconstructing the data in case of partia damage or defect in the generated barcode. This abiity to compensate for partia damage improves the credibiity of barcodes as a reiabe machine-readabe format. There are error corrections settings associated with most symboogies with the intention of increasing the readabiity. However, there is a fine baance when seecting the desired setting because of an increase in the overa encoded barcode content (and a corresponding decrease in the actua information that can be encoded within the prescribed capacity imits) that resuts from using the error correction. One must baance this vaue with the string to be encoded in order to stay in ine with the capacities associated with the specific symboogy. Error correction eves for PDF417 and QR Code are tabuated in Tabe III and Tabe V, respectivey. 2.4 Scan Grade and Refectance Profie A refectance profie (SRP) is a coection of % refectance vaues measured across a barcode by a ine. As per the American Nationa Standards Institute bar code print quaity specification, ANSI X3.182 [11], the SRP considers the foowing eight parameters to obtain a grade: Edge determination Minimum refectance Minimum edge contrast Symbo contrast Moduation Defects Decode-abiity (printing accuracy as compared to the agorithm) Decode (pass/fai) The minimum grade achieved by any of the parameters above represents the overa SRP grade. 2.5 Symbo Grade The average of 10 SRP grades gives what is known as the Symbo Grade. As per the ISO/IEC and ANSI X3.182, and EN 1635 standards, these 10 s shoud be conducted at different heights [16]. The numeric vaues and the corresponding symbo grades are defined as given in Tabe IX.

13 220 A. GROVER ET AL. Tabe IX SYMBOL GRADE AND NUMERIC VALUE CORRESPONDENCE Numeric vaue (x) Symbo grade 4.0=x=3.5 A 3.5> x = 2.5 B 2.5> x = > x = >x C D F 2.6 Encoder For the purpose of this chapter, barcodes were printed using two separate encoders, viz, Bartender Enterprise edition 9.01 from Seagu Scientific and CardFive Vision 8.1 Professiona from Number Five Software. Whie Bartender is an industry-standard software optimized for encoding barcodes, CardFive Vision is primariy an ID card design software that aso supports encoding of barcodes. Whie initia testing suggested 10 positive resuts for barcodes generated using Bartender, the encoding quaity achieved through CardFive was not found to be that encouraging. Hence, for reiabiity and consistency, a further tests incuding the preiminary as we as extensive tests detaied beow were done using barcodes generated by Bartender. It may be noted here that apart from Bartender, any standard commercia barcode encoder may be suitabe as ong as the barcodes generated are of a high quaity. To eiminate any potentia of the encoder adversey impacting the readabiity, independent externa testing using barcode verifiers indicated that our barcodes received a quaity grading of A. 2.7 Printer The ning reiabiity of barcodes is impacted significanty by the quaity of printing which is determined to a great extent by the printer resoution and the type of print head being used. As per the AIM and GS1 pubications [1,12], the foowing print considerations are important: The generated barcodes shoud aways be an even mutipe of printer pixes. There must be sufficient quiet zones around the bar code as required by the respective barcode specifications

14 PARAMETERS EFFECTING 2D BARCODE SCANNING RELIABILITY 221 Tabe X COMPARISON OF PRINTER SPECIFICATIONS Name and mode number Media type Resoution (DPI) Dot shape Printing technoogy Zebra S4M Labe printer 203 Square Direct therma DataCard SP75 Card printer 300 Square Dye-subimation/resin therma transfer Xerox Phaser 6360 DN Document printer 600 Round Laser The distortion introduced by genera-purpose printers (with round dot shape) shoud be catered for. The round dot-shape resuts in printing wider bars and narrower spaces because the printed dot size is bigger than the pixe size. For the purpose of this case study, the foowing printers were used for printing the barcodes: (i) Zebra S4M abe printer (ii) DataCard SP75 card printer (iii) Xerox Phaser 6360 DN aser printer A brief comparison of the reevant specifications of these printers is tabuated in Tabe X [13 15]. When considering the printer as a factor, it must be noted that a direct comparison of resoution aone without taking the printing technoogy into consideration woud be miseading, since a square dot-shape produces a straight edge thus having the abiity to print higher density barcodes more accuratey as opposed to a printer with a round dot shape of the same DPI vaue. Since the direct therma and therma transfer technoogies are optimized for printing high-capacity bar codes and since our target media for the case study was abes and PVC cards, ony the first two printers mentioned above were used for most of the tests. The aser document printer was used for testing the effect of gossiness on reiabiity as it aowed a direct comparison between gossy and nongossy media. 2.8 Media For the purpose of this case study, we focused on our target media, viz. abes and PVC cards that are both gossy in nature. The absorbance or the optica density of the media affects the quantum of refraction and is thus another important factor that impacts on the -abiity of the barcodes. There is a marked difference in

15 222 A. GROVER ET AL. the optica density of gossy as we as nongossy media and though gossy media provide a much higher optica density, they aso provide a higher eve of refectivity from the printed surface. Whie refectivity is the fraction of incident ight refected by the surface in question; the optica density or absorbance, A, is defined as A ¼ og 10 ði=i 0 Þ where I is the intensity of the transmitted ight at a given waveength and I 0 is the intensity of the incident ight. To cater for the difference in absorbance and refectivity of different media, a specific test, the Effect of Gossiness on reiabiity test, was carried out with the foowing two types of media: Norma nongossy paper Gossy paper 2.9 Decoder/Scanner Characteristics The reiabiity of barcodes is aso affected by the type of ner/decoder, or the actua hardware that wi decode the encoded barcode. The reiabiity is greaty affected by the combination of the ner ight-waveength and aperture size (size of the ning spot reative to the bar-width) used. The ange at which ner ight is incident on the surface affects the refectance. Light incident at 45 to the surface wi minimize refection from gossy surfaces, thereby producing optimum reiabiity. As per the Layman s guide [1], variations in these two parameters might drasticay ater the overa grade quaity (Tabe XI). As per GS1, a goba eader invoved in the design and impementation of standards incuding those of bar codes, depending on ner technoogy, there are six functiona bands. For optimum -reiabiity, ners shoud be chosen on the basis of appications that pertain to these bands [12]. Tabe XI SCAN GRADE W.R.T. APERTURE DIAMETER AND WAVELENGTH Aperture diameter (mi) Aperture diameter (mm) Waveength (nm) Scan grade Quaity /D Poor /B Good /F Fai

16 PARAMETERS EFFECTING 2D BARCODE SCANNING RELIABILITY 223 Tabe XII COMPARISON OF SCANNER SPECIFICATIONS Characteristic Symbo DS 6707SR Symbo DS 3478 Light source 650-nm aser diode 650-nm aser diode Fied of view 40 horizonta 30 horizonta 30 vertica 22.5 vertica Yaw toerance 60 from nomina 50 from nomina Pitch toerance 65 from nomina 60 from nomina Ro toerance 360 from nomina 180 from nomina Since Motoroa (a.k.a. Symbo) ners possess a majority of the ner hardware market share; the foowing two different Symbo ners were used for testing: Symbo DS 6707SR wired ner with SE6707 engine Symbo DS 3478 wireess ner with SE4400 engine The specific difference between these two ners is the embedded engine, which is the heart and sou of the ner and responsibe for the actua decoding process [22, 23]. A brief comparison of the technica specifications that are directy reevant to our tests is given in Tabe XII. 3. Testing Methodoogy Our comprehensive testing invoved interna as we as externa testing. The interna testing methodoogy, which is described in Section 3.1, used a series of carefuy designed tests to evauate the impact of various parameters that infuence barcode ning reiabiity. The interna testing comprised severa stages incuding preiminary samping foowed by more extensive testing with the resuts of each test eading to the need for the successive tests. The externa testing, which is described in Section 3.2, was used to assess the quaity grade of the barcodes by using an independent and verifiabe industry standard process. This externa testing augmented the interna testing where resources were not avaiabe to actuay perform the testing. 3.1 Interna Testing To provide a point of reference and for the sake of comprehensive testing, 1D as we as 2D barcodes with ower content density were aso tested. Since the focus of this paper is higher capacity barcodes, the resuts of the ower capacity testing are

17 224 A. GROVER ET AL. mentioned without going into further detais. Testing was performed by performing individua s, termed iteration, of the sampe barcodes and then cacuating success rate, which is caed the barcodes -abiity Barcodes with Lower Content Density Mutipe iterations using different types of ners as we as encoders for a 1D as we as ower content density 2D barcodes tested produced 10 reiabiity Barcodes with Higher Content Density Sampes of various barcode symboogies were subjected to preiminary testing invoving twenty -iterations each and a sampes that achieved a reiabiity of 4 or more were then subjected to extensive testing invoving fifty iterations each. The testing methodoogy is expained in detai in the succeeding sections Preiminary Testing. In order to narrow the scope of viabe barcodes, and based on the specific requirements of our project wherein the physica print size was a imiting factor, the four commony used 2D barcodes (Tabe XIII) were encoded for preiminary testing. This test invoved twenty iterations of each barcode sampe and the resut was a percentage of successfu Tabe XIII PRELIMINARY TESTING S. No. Barcode type 1 PDF mi 2x, 13 co EC : 6 2 QR cvode 14.8 mi error correction M (15%) 3 Aztec 14.8 mi with error correction 4 4 Aztec 14.8 mi with error correction 45% 5 Aztec 19.7 mi (no image) 6 DataMatrix 14.8 mi with ECC DataMatrix 9.8 mi Scanning reiabiity 95 % 90 % 45 % 80 % 90 % 50 % 0 %

18 PARAMETERS EFFECTING 2D BARCODE SCANNING RELIABILITY 225 reiabiity. This preiminary test served as a smoke test to provide a quick initia point of reference, based on a wide range of possibe barcodes and the settings associated with each particuar barcode. In particuar, the suggested settings for each symboogy were used to encode the barcode. Based on the resuts of this preiminary testing, the ow-performing barcode namey DataMatrix 9.8 mi (S. No. #7) was eiminated from further testing. On the other hand, for the mediocre-performing Aztec 14.8 mi with error correction 4 (S. No. #3), an additiona sampe with a 5 error correction was added for further testing. Simiary, an additiona QR Code sampe with an error correction of 3 was aso incuded for further testing Extensive Testing. This invoved fifty iterations for each sampe barcode and enumerating the resuts as a percentage reiabiity vaue. The number of s was suggested by the ner vendor as being sufficient to successfuy sampe the success rate. For a better understanding of resuts and to ensure that our testing resuts were not biased by the use of ony one type of ner, reiabiity was enumerated separatey for different ners, viz. the Symbo DS 6707SR (SE6707) wired and the Symbo DS 3478 (SE4400) wireess ners. For the purpose of this testing, a good is defined as a successfu in ess than 2 s, a deayed is a successfu that takes 2 7 s, and a No is faiure to even after 7 s. The deayed measurement was introduced to account for the fact that whie not actuay a No Scan, onger than a 2-s time is ess than optima in a depoyed situation; however, the tracking of such s becomes vauabe for measurement when onger times are acceptabe. Deayed s are acting as a buffer between the two possibiities Reiabiity testing. The first in a series of specific extensive tests, Reiabiity Testing invoved extensive ning of different types of 2D barcodes, viz. the PDF417, DataMatrix, Aztec, and QR Code with both the engines. The goa of reiabiity testing was to identify barcode symboogies aong with the corresponding engine that produced the highest percentage reiabiity. These test resuts, summarized in Tabe XIV with the reevant rows highighted with the arrow marking, indicate that PDF mi 13 co, error correction: 6 with SE6707 and QR Code 14.9 mi, error correction (3) with SE4400 emerged as the most reiabe s, which were then subjected to further testing Content density testing. Having shortisted PDF417 and QR Codes based on the reiabiity testing resuts, each of the two barcode symboogies was subjected to rigorous content density testing. The purpose of this test was to determine the impact of the content density on the reiabiity of the barcodes.

19 226 A. GROVER ET AL. Tabe XIV RELIABILITY TESTING Symboogy PDF mi 13 co EC:6 (abe) QR code 14.8 mi error correction M (15%) (Labe) QR code 14.8 mi, error correction M (15%) (Card) QR code 14.8 mi, error correction (3) (card) Aztec 14.8 mi with error correction 4 (abe) Aztec 14.8 mi with error correction 45% (abe) Aztec 14.8 mi with error correction 5 (abe) Aztec 19.7 mi (no image) (abe) DataMatrix 14.8 mi with ECC 200 (abe) Scan engine % Good % Deayed SE % 6% SE4400 SE6707 SE4400 SE6707 SE4400 SE6707 SE % % 94% 2 38% 18% 38% 16% 3 SE6707 SE4400 SE6707 SE4400 SE6707 SE4400 SE6707 SE % 26% % 8% 26% 48% % No % 14% 58% 8% 56% % 24% We designed different capacity barcodes ranging from 162 bytes to 929 bytes. The different byte sizes were seected based on the specific requirements of the project and thus subsequent test sampes do not have a uniform data capacity difference. In Tabe XV, the rows indicating test resuts for a data capacity of 844 bytes are highighted with the arrow marking, as that is the preferred data capacity for our project. This test indicated that for PDF417, the reiabiity starts dropping after content density increases more than approx. 6 when the wireess ner (SE4400) is used. However, the content density has no effect whatsoever on reiabiity with the wired ner (SE6707). For the QR Code, the testing indicated a substantia drop in reiabiity as the content density increased to more than 844 bytes with either the SE6707 or SE4400 Scan Engines (Tabe XVI). This indicates mixed resuts for the effect of content density on reiabiity of QR Code. There is a trend toward a decrease in the reiabiity as the content of the barcode increases. The success rate is higher with the

20 PARAMETERS EFFECTING 2D BARCODE SCANNING RELIABILITY 227 Tabe XV CONTENT DENSITY TESTING FOR PDF 417 SE 4400 Size in bytes, %capacity % Good % Deayed % No PDF mi EC = 6 162, 14% 224, 2 492, 45% 625, 56% 694, 63% 844, 76% 929, 84% Size in SE 6707 bytes, %capacity PDF mi EC = 6 162, 14% 224, 2 492, 45% 625, 56% 694, 63% 844, 76% 929, 84% % 10 96% 88% 78% % Good % 16% % Deayed 4% 6% % No wired ner when one takes into account the deayed s because in essence these are successfu s though just not as idea Error correction eve based testing for QR Code and PDF417. To fine-tune the resuts of the previous test and to achieve high reiabiity with high content density barcodes, the required 844 byte sampes (of PDF417 and QR Code) were tested with different eves of error correction encoded into them. The purpose of this test was to determine the impact of error correction eve on the reiabiity of the 2D barcodes. For QR Code, sampes with a content density of 844 bytes with error correction eves of 7%, 15%, 25%, and 3 were tested. The resuts are tabuated in Tabe XVII. The rows highighted with the arrow marking indicate the best resuts for the two different ners. It was observed that the size of the actua barcode, not just the amount of data, seemed to pay a roe in the ning reiabiity. It was observed that barcodes with a 3 error correction were so big that it seemed to

21 228 A. GROVER ET AL. Tabe XVI CONTENT DENSITY TESTING FOR QR CODE QR code 14.8 mi EC = 15% Size in bytes % Good % Deayed % No SE % % 6% % % % % % 24% 3 SE % % % 4% % 26% % % 48% 6% adversey affect the reiabiity. Based on data content requirements of the present project, our recommendation is to use 25% error correction to achieve optimum reiabiity. For PDF417, sampes with a content density of 844 bytes with error correction eves of 3, 4, 5, and 6 were tested with both the engines. The resuts are tabuated in Tabe XVIII. For PDF417, the resuts indicate that the higher the error correction, better the reiabiity Effect of Gossiness on reiabiity. After taking into consideration the primary factor of data content density, we shifted our focus to other secondary factors that impact on the reiabiity of barcodes. Since the primary technique used by most 2D ners is imaging, which invoves taking a photo of the barcode before trying to decode it, the difference in absorbance and refectivity of gossy and nongossy media impacts on their readabiity. The purpose behind this

22 Tabe XVII IMPACT OF ERROR CORRECTION LEVEL FOR QR CODE QR Code 844 Bytes SE 6707 Error correction % Good 7% 88% 15% 84% 25% 98% % Deayed % No 1 16% SE % % 88% 1 25% 9 6% 3 74% 2 6%

23 230 A. GROVER ET AL. Tabe XVIII IMPACT OF ERROR CORRECTION LEVEL FOR PDF 417 PDF bytes Error correction % Good % Deayed % No SE % 16% 4 94% 6% 5 98% 6 98% SE % 26% 48% % % 14% 8% 6 86% 1 4% test was to determine the potentia impact of the finish of the physica media, in particuar gossy versus nongossy, on the ning reiabiity. To resove this concern, the symboogies were ned separatey on a gossy as we as a nongossy media. The resuts are tabuated in Tabe XIX Effect of gossiness testing for QR Code 844 bytes data capacity. The highighted rows indicate the best reiabiity for gossy as we as nongossy paper using different engines. The resuts indicate a cear drop in the reiabiity of QR Code barcodes when printed on gossy surface (cards as we as abes) as compared to nongossy pain surface Effect of gossiness testing for PDF bytes data capacity. The test resuts indicate a significant degradation in the reiabiity on gossy surface compared to nongossy surface at ower error correction eves for the SE6707 as shown in rows highighted by a marking. For the SE4400, the corresponding rows highighted by the arrow marking indicate significant degradation in the reiabiity on gossy surface at a eves of error correction (Tabe XX) Comparison of target media (card and abe) with optima settings. Since our target media (abe or PVC card) are gossy, we decided to do yet another test to determine which of the two target media (card or abe) produced a

24 PARAMETERS EFFECTING 2D BARCODE SCANNING RELIABILITY 231 Tabe XIX EFFECT OF GLOSSINESS TESTING FOR QR CODE Non-gossy paper SE 6707 SE 4400 Error correction 7% 15% 25% 3 7% 15% 25% 3 % Good 88% 84% 98% % 9 74% % Deayed 1 16% % 2 % No 6% Gossy paper SE6707 SE4400 Error correction 7% 15% 25% 3 7% 15% 25% 3 Good % % 44% 96% 98% 8 44% Deayed No % 28% 3 34% 44% 1 4% 14% 2 4% 36% better reiabiity. It may be pertinent to mention here that there is a difference not ony in the gossiness of the media (card and abe) but aso in the printing quaity as cards were printed using the DataCard SP75 card printer whie the abes were printed using the ZebraS4M abe printer having different characteristics (as mentioned in Section 2.7). These resuts in Tabe XXI indicate that PDF417 is the better choice for abes; however, for badges, the wireess ner produced identica reiabiity resuts of 76% for both PDF417 and QR Code.

25 232 A. GROVER ET AL. Tabe XX EFFECT OF GLOSSINESS TESTING FOR PDF 417 Media Non gossy Scan engine SE6707 Error correction 3 4 % Good 94% 10 % Deayed 6% % No 5 98% SE % 6% 5 94% 4% 6 94% 4% Gossy SE % 16% 4 94% 6% 5 98% 6 98% SE % 26% 48% % % 14% 8% 6 86% 1 4% 3.2 Externa Testing In order to address the potentia for the encoder or the printers to infuence the performed testing, it is necessary to set a baseine for the quaity of created barcodes. This took the form of grading the test sampes that was performed independenty and externay by Motoroa Technica and Engineering Services using barcode verifiers and they confirmed the quaity of the internay encoded barcodes as Grade A. This eiminates the potentia of encoder and printer errors on the reiabiity of the barcodes.

26 PARAMETERS EFFECTING 2D BARCODE SCANNING RELIABILITY 233 Tabe XXI COMPARISON OF TARGET MEDIA Symboogy PDF EC 6 QR code % Scan engine SE 6707 SE 4400 SE 6707 SE 4400 Media printed on Labe Labe Good % Card 96% Card Labe Card Labe Card 99% 98% 76% % Deayed % 4% 36% 1 76% 14% No % 1% % 1 26% 1 4. Concusion For our target materia (card or abe, which are both gossy), and with a data content capacity of 844 bytes, The best case scenario is a 96 99% reiabiity for cards and abes, respectivey, using PDF417 Barcodes with the SE6707 engine. The ning reiabiity of PDF417 with the SE4400 engine varied from 76% for cards to 98% for abes. For QR Code with 844 bytes and 25% error correction, the SE4400 engine produced a success rate of 76% for cards and 64% for abes, whereas the SE6707 Scan engine produced a success rate of 5 for both abes and cards. These resuts are summarized in Tabe XXII. Comparing these resuts to the standards for each of the tested symboogies, theoreticay they are capabe of encoding high content-data; however, testing showed that the content for readabe symbos was greaty impacted by many factors. The resuts thus indicate that for our specific appication, the reiabiity of PDF417 barcodes was much better than that of QR Code barcodes with highcapacity data encoding (more than 800 bytes) printed on gossy media and read using either SE6707 engine or SE4400 engine.

27 234 A. GROVER ET AL. Tabe XXII SUMMARY OF RESULTS Rank Best Scan-reiabiity 99% Symboogy Scanner Media PDF 417 SE 6707 Labe 2 nd 98% PDF 417 SE 4400 Labe 3 rd 96% PDF 417 SE 6707 Card 4 th 76% QR Code SE 4400 Card 5 th 76% PDF 417 SE 4400 Card 6 th 64% QR Code SE 4400 Labe Worst 5 QR Code SE 6707 Card & Labe References [1] Layman s Guide to ANSI, CEN, and ISO Bar Code Print Quaity Documents, Association for Automatic Identification and Data Capture Technoogies (AIM), Pittsburg, PA, 2002 (November). [2] Sizing Appications for 2D Barcode Symbos, ibrary/white_papers/index.aspx (Intermec Technoogies Corporation, Avaiabe from: accessed ). [3] (accessed ). [4] (accessed ). [5] (accessed ). [6] (accessed ). [7] (accessed ). [8] (accessed ). [9] (accessed ). [10] (accessed ). [11] American Nationa Standards Institute, Bar Code Print Quaity Specification, ANSI X3.182, [12] GS1 Genera Specification, v 7.0, v 7.0, Jan 2006: Bar Code Production and Quaity Assessment, [13] htm (accessed ). [14] (accessed ). [15] (accessed ).

28 PARAMETERS EFFECTING 2D BARCODE SCANNING RELIABILITY 235 [16] GS1 Systems, GS1 Bar Code Verification for Linear Symbos, v 4.3, (May). [17] ISO/IEC 15438:2001, Information Technoogy Automatic Identification and Data Capture Techniques Bar Code Symboogy Specifications PDF417. [18] Laser Marking: Matrix Codes on PCBS, (Rick Stevenson, Printed Circuit Design & Manufacture, December). [19] (accessed ). [20] (accessed ). [21] (accessed ). [22] Symbo DS 6707SR Manua. Avaiabe from: do (accessed ). [23] Symbo DS 3478 Manua. Avaiabe from: (accessed ). [24] (accessed ). [25] (accessed ). [26] (accessed ). [27] (accessed ). [28] (accessed ). [29] (accessed ). [30] (accessed ). [31] (accessed ).