In this article we introduce a new method for automatic identification of suspicious objects in body scanner’s images in the leg’s regions. The method is based on a combination of texture and a classifier, which may be the KNN, K-means or the MLP neural network. The methods are applied on 20 body scanner’s images, in the leg’s regions of volunteers. The accuracy of the method is verified by the hit rate presented in graphical form and the presence of false positives. The results indicate that the method that combines textures with MLP provides a hit of 0.91 and 0.08 false positives, the KNN classifier hits 0.89 and provides 0.01 of falsepositive while K-means obtained a hit of 0,81 and 0.03 of false positives. Given the results, it is concluded that the method that combines texture and MLP achieved the identification of suspicious objects more effective than the other classifiers used in this work.
