Modified Adaptive Implicit Shape Model for Object Detection

Abstract

Automated threat object detection in X-ray images is needed urgently in baggage inspection at airports, railway stations and other public places. However, the works on object detection are still very limited to meet the needs of practical application. In this paper, we propose a modified adaptive implicit shape model (MAISM) to detect threat objects in X-ray images, in which the triangle patches are used to compute occurrence of the centroid of object instead of keypoints. This model is adaptive for object detection in images of variable scales through triangle patch matching. Experiments on three different threat objects images (razor blades, shuriken, handguns) of various scales demonstrate the effectiveness of the proposed method.