Close
Metadata

%0 Conference Proceedings
%4 sid.inpe.br/sibgrapi/2017/08.17.13.40
%2 sid.inpe.br/sibgrapi/2017/08.17.13.40.25
%T Detecting Modifications in Printed Circuit Boardsfrom Fuel Pump Controllers
%D 2017
%A Oliveira, Thomas Jose Mazon de,
%A Marco Aurelio Wehrmeister,
%A Nassu, Bogdan Tomoyuki,
%@affiliation Federal University of Technology - Parana
%@affiliation Federal University of Technology - Parana
%@affiliation Federal University of Technology - Parana
%E Torchelsen, Rafael Piccin,
%E Nascimento, Erickson Rangel do,
%E Panozzo, Daniele,
%E Liu, Zicheng,
%E Farias, Mylène,
%E Viera, Thales,
%E Sacht, Leonardo,
%E Ferreira, Nivan,
%E Comba, João Luiz Dihl,
%E Hirata, Nina,
%E Schiavon Porto, Marcelo,
%E Vital, Creto,
%E Pagot, Christian Azambuja,
%E Petronetto, Fabiano,
%E Clua, Esteban,
%E Cardeal, Flávio,
%B Conference on Graphics, Patterns and Images, 30 (SIBGRAPI)
%C Niterói, RJ
%8 Oct. 17-20, 2017
%S Proceedings
%I IEEE Computer Society
%J Los Alamitos
%K Image-based detection of modifications in printed circuit boards, Fraud detection in fuel pumps, Computer vision, image registration, machine learning, Points of interest.
%X Frauds involving illegal modifications to the printedcircuit boards from fuel pump controllers are a serious problem,which not only harms customers, but also connects to othercrimes, such as money laundering and tax evasion. The currentstate-of-practice for inspecting these boards is a visual analysisperformed by a human. In this paper, we introduce an image-based approach that can provide support to the human inspectorby automatically detecting suspicious regions in the boards.The proposed approach aligns a photograph of the inspectedboard to a reference view, partitions the image in sub-regions,extracts features using a variation of the popular Scale-InvariantFeature Transform, classifies the features against previouslytrained Support Vector Machines, and integrates the results forpresentation. In experiments performed on a dataset containing649 images from a board, with and without modifications, ourapproach achieved a precision of 0.7739, a recall of 0.9434, andanF-measure 0.8503. These results indicate that our approachcan effectively identify suspicious regions, providing invaluablehelp to the human inspector.
%@language en
%3 SibGrapi.pdf


Close