Resumen
The process of detecting hidden weapons is an important process right now due to the increase in terrorist operations, so the process of building an automatic weapons detection system is an important process to reduce errors resulting from manual detection. In the proposed work, the preprocessing was given high importance because the x-ray images contain noise and low resolution, therefore image smoothing has been used to reduce the noise where histogram equalization has been used for image enhancement and increase of contrast. The local algorithms: SIFT, SURF and ORB have been used to detect and describe the features from the region of interest, then KNN algorithm has been used to match and index the similarity between the query image and the extracted features from the data set. KNN and Random Sample used a consensus on the three methods to see which local algorithm performs best. RANSAC has been used to reject false matches that may be taken as correct matches. The performance of the SIFT algorithm with the KNN outweighed both of the algorithms in spite of the fact that it was slow. SURF and the ORB algorithms as a position in the result where SURF was the fastest one with high performance and showing its dominance in illumination changes and rotation