Abstract
This paper evaluates a new and useful tool which can be used to facilitate trade while simultaneously enhancing supply-chain security. It describes a simple algorithm to characterize regularities in the appearance of particular cargo types in radiography images and verify that new images with the same cargo type match the earlier characterization. We find that it is possible to identify containers inconsistent with their label to a substantial degree of confidence. This evaluation was made possible by analyzing a novel data set consisting of radiography images of real stream-of-commerce container cargo which is labeled with the contents from the shipment manifest. These data were gathered from containers passing through U.S. ports of entry and from customs data filed by importers. The pairing of radiograph and manifest data is shown to hold great potential for assisting human customs inspectors and detecting anomalous shipments.
This is a preview of subscription content, log in via an institution to check access.
Similar content being viewed by others
Notes
CBP form 7501, which is filled out by shippers.
In the European Union, Regulation 1875/2006 has required similar data to be available to customs authorities since December 31, 2010.
Median and average absolute deviation (AAD) of bin intensity were chosen as measures of the moments because they are more robust to outliers than mean and standard deviation. AAD performed very similarly to the better known median average deviation (MAD), but the latter has the complication of possibly equaling zero in our data.
Several alternative distance metrics were evaluated, including Euclidian, chi-square, and earth-mover’s distance (Rubner et al. 1998, 2000), both before and after weighting by variance. Comparisons of predictive value were generated by calculating the fraction of samples for which the closest model is of its own stated HS code as in “Automatic recognition of HS-codes from images”; the above measure performed best.
Sub-models are retained if the following three criteria are met. (1) Each sub-model represents at least 75 exemplars. (2) Each sub-model represents at least 10 % of population exemplars. (3) The distance from the resulting sub-model to each other sub-models is at least 1.15.
In total, five of 92 models having AUC less than 0.5 were dropped (see “Outlier detection and rejection”).
Cars are easily identifiable with certainty by eye in radiography images. Cars and other outliers which are likely incorrectly labeled were common enough to show up as a secondary peak on the distribution of distances between samples and the models to which they belong in Fig. 5.
These ROC curves are not cross-validated. Although, in theory, this means they may be biased, in practice it is unlikely to have much effect. Since our models are based on medians for the population of at least 100 samples, any individual sample has little effect on the model.
References
Barnard K, Duygulu P, Forsyth D, de Freitas N, Blei DM, Jordan MI (2003) Matching words and pictures. J Mach Learn Res 3:1107–1135
Barnard K, Fan Q, Swaminathan R, Hoogs A, Collins R, Rondot P, Kaufhold J (2007) Evaluation of localized semantics: data, methodology and experiments. Int J Comput Vis 77(1–3):199–217
Bersin A (2010) Commissioner shares trade vision at CBP conference. Customs and border protection. http://www.cbp.gov/xp/cgov/newsroom/highlights/trade_conf.xml. Accessed 6 July 2011
Bosch A, Zisserman A, Munoz X (2007) Image classification using random forests and ferns. Proceedings of the 11th International Conference on Computer Vision, Rio de Janeiro, Brazil
CBP (2006) Privacy impact assessment for the automated targeting system. http://www.dhs.gov/xlibrary/assets/privacy/privacy_pia_cbp_ats.pdf. Accessed 6 March 2012
CBP (2009) Importer security filing and additional carrier requirements information. http://www.cbp.gov/xp/cgov/trade/cargo_security/carriers/security_filing/. Accessed 30 October 2011
CBP (2011a) Container security initiative. Customs and border protection. http://www.cbp.gov/xp/cgov/trade/cargo_security/csi/. Accessed 30 October 2011
CBP (2011b) Customs-trade partnership against terrorism. Customs and border protection. http://www.cbp.gov/xp/cgov/trade/cargo_security/ctpat/. Accessed 30 October 2011
Customs and Border Protection (2011c) DHS and DOE launch secure freight initiative. http://www.dhs.gov/xnews/releases/pr_1165520867989.shtm. Accessed 30 October 2011
Decker SH, Chapman MT (2007) Drug smugglers on drug smuggling: lessons from the inside. Temple University Press, Philadelphia
Descalle MA, Manatt D, Slaughter D (2006). Analysis of manifests for containerized commodities imported through US Ports. In Nuclear Science Symposium Conference Record, 2006. IEEE (Vol. 1, pp. 275–280). IEEE.
Donner M, Kruk C (2009) Supply chain security guide. The International Bank for Reconstruction and Development (The World Bank)
Duygulu P, Barnard K, de Freitas N, Forsyth D (2002) Object recognition as machine translation: learning a lexicon for a fixed image vocabulary. Seventh Eur Conf Comput Vis IV:97–112
Ebeling CE (2009) Evolution of a box. Invent Technol 23(4):8–9, ISSN 8756–7296
European Commission (2006) Customs related security initiatives of the EU. European Taxation and Customs Union. http://ec.europa.eu/taxation_customs/customs/policy_issues/customs_security/security_initiatives/index_en.htm Accessed 30 October 2011
Fawcett T (2006) An introduction to ROC analysis. Pattern Recognit Lett 27:861–874
Feller G (2009) Lessons from the secure freight initiative. Supply Chain Asia Magazine July/August 2009
Fergus R, Perona P, Zisserman A (2005) A sparse object category model for efficient learning and exhaustive recognition. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, San Diego
Fleck D, Duric Z (2009) Affine invariant-based classification of inliers and outliers for image matching. LNCS 5627:268–277
Government Accountability Office (2006) Cargo container inspections: preliminary observations on the status of efforts to improve the automated targeting system. GAO-06-591T. March 30, 2006. http://www.gao.gov/products/GAO-06-591T. Accessed 6 July 2011
Government Accountability Office (2009a) Supply chain security: feasibility and cost-benefit analysis would assist DHS and congress in assessing and implementing the requirement to scan 100 % of U.S.-Bound Containers. GAO-10-12 October 30, 2009. http://www.gao.gov/products/GAO-10-12. Accessed 6 July 2011
Government Accountability Office (2009b) Combating nuclear smuggling: lessons learned from DHS testing of advanced radiation detection portal monitors. GAO-09-804T June 25, 2009. http://www.gao.gov/products/GAO-09-804T. Accessed 30 October 2011
Government Accountability Office (2009c) Combating nuclear smuggling: recent testing raises issues about the potential effectiveness of advanced radiation detection portal monitors. GAO-10-252T November 17, 2009. http://www.gao.gov/products/GAO-10-252T. Accessed 30 October 2011
Government Accountability Office (2010) Combating nuclear smuggling: inadequate communication and oversight hampered DHS efforts to develop an advanced radiography system to detect nuclear materials. GAO-10-1041T September 15, 2010. http://www.gao.gov/products/GAO-10-1041T. Accessed 30 October 2011
HSNW (2011) U.S. no longer mandating 100 % screening of cargo containers. Homeland Security News Wire. http://www.homelandsecuritynewswire.com/us-no-longer-mandating-100-percent-screening-cargo-containers. Accessed 30 October 2011
Jeon J, Lavrenko V, Manmatha R (2003) Automatic image annotation and retrieval using cross-media relevance models. SIGIR 2003, Proceedings of the 26th annual international ACM SIGIR conference on Research and development in information retrieval
Kaufhold J, Collins R, Hoogs A, Rondot P (2006) recognition and segmentation of scene content using region-based classification. Proceedings of the 18th International Conference on Pattern Recognition, IEEE
Kennedy H (2005) U.S. customs goes high-tech for cargo security. National Defense. Nov, 2005. http://www.nationaldefensemagazine.org/archive/2005/November/Pages/UF-US_Customs3077.aspx. Accessed 28 October 2011
Krupinski EA, Berbaum KS, Caldwell RT, Schartz KM, Kim J (2010) Long radiology workdays reduce detection and accommodation accuracy. J Am Coll Radiol 7(9):698–704
Lowe D (2003) Distinctive image features from scale-invariant keypoints. Int J Comput Vis 60(2):91–110
Malone R (2009) Bush signs safe port act. Forbes.com. http://www.forbes.com/2006/10/13/safe-ports-act-biz-logistics-cx_rm_1013ports.html. Accessed 6 July 2011
Miller EA, et al. (2011) Scatter in cargo radiography. Appl Radiat Isot (2011), doi:10.1016/j.apradiso.2010.12.006
Murphy K, Torralba A, Freeman W (2003) Using the forest to see the trees: a graphical model relating features, objects, and scenes. Commun ACM 53(3)
Office of National Drug Control Policy (2001) Measuring the deterrent effect of enforcement operations on drug smuggling, 1991–1999. http://www.abtassociates.com/reports/measure_deter_effct.pdf
Qi Y, Salehi Y, Wang Y (2011) Investigate existing non-intrusive (NII) technologies for Port Cargo inspections. http://swutc.tamu.edu/publications/technicalreports/161042-1.pdf. Accessed 9 March 2012
Rubner Y, Tomasi C, Guibas L (1998) A metric for distributions with applications to image databases. Proc ICCV 1998:59–66
Rubner Y, Tomasi C, Guibas L (2000) The earth mover’s distance as a metric for image retrieval. Int J Comput Vis 40(2):99–121
Russell BC, Torralba A, Murphy KP, Freeman WT (2005) LabelMe: a database and web-based tool for image annotation. Comp Sci and Artificial Intelligence Laboratory Technical Report, MIT-CSAIL-TR-2005-056, September 8
Smeulders AWM, Worring M, Santini S, Gupta A, Jain R (2000) Content-based image retrieval at the end of the early years. PAMI 22(12):1349–1380
Swedish National Board of Trade (2008) Supply chain security initiatives: a trade facilitation perspective. Kommerskollegium 2008:1
Tourassi GD, Vargas-Voracek R, Catarious DM, Floyd CE (2003) Computer-assisted detection of mammographic masses: a template matching scheme based on mutual information. Med Phys 30(8)
Varma M, Zisserman A (2005) A statistical approach to texture classification from single images. Int J Comput Vision Spec Issue Texture Anal Synth 62(1–2):61–81
World Development Indicators (WDI) (2011) World bank. http://data.worldbank.org/indicator/IS.SHP.GOOD.TU. Accessed 3 November 2011
Acknowledgments
The Department of Homeland Security (DHS) sponsored the production of this material under contract HSHQDC-07-F-00096. Notwithstanding this sponsorship the views expressed in this paper are those of the authors and do not necessarily represent the views of and should not be attributed to DHS.
Author information
Authors and Affiliations
Corresponding author
Rights and permissions
About this article
Cite this article
Tuszynski, J., Briggs, J.T. & Kaufhold, J. A method for automatic manifest verification of container cargo using radiography images. J Transp Secur 6, 339–356 (2013). https://doi.org/10.1007/s12198-013-0121-3
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s12198-013-0121-3