Resumo

CASTRO PINOL, David; SANABRIA MACIAS, Frank; MARANON REYES, Enrique  e  RODRIGUEZ ARIAS, Felipe. Weapons recognition in X-ray images using Bag of Visual Words. Rev cuba cienc informat [online]. 2016, vol.10, n.1, pp.152-161. ISSN 2227-1899.

ABSTRACT An automatic system’s design that recognizes dangerous objects in baggage X-ray images has been a complex problem in recent years. X-ray inspection has difficulties because of the low efficiency in automatic recognition of dangerous objects and inspection process delay. It doesn’t exist a software application that automatically detects weapons in those images and reduce the workload of screeners. In this project was developed and implemented an algorithm for recognizing handguns in X-ray images using the Bag of Visual Words method. In order to achieve this, it was implemented a preprocess, was built a vocabulary of visual words with the better performance for this kind of images, it was represented a set of images by histograms of visual words and it was trained a Support Vector Machine classifier. This algorithm was developed in Matlab platform using VLFeat library. It was performed several experiments handling tunable parameters, getting the most relevant result a true positive rate of 97.12% and a false positive rate of 7.4%. These results show that the implemented algorithm could be a support for inspection screeners and hence increase inspection speed and increase the efficiency of weapons recognition in X-ray images of inspection system.

Palavras-chave : Bag of Visual Words; Suport Vector Machines; X-ray images.

        · resumo em Espanhol     · texto em Espanhol     · Espanhol ( pdf )