This paper presents an approach to identify a weapon from a single image using a weapon ontology. Ontological nodes selected by experts store convex hull (CH) sequences for their descendants, whereas the ontological leafs are labeled with object boundary sequences. The latter are generated from object boundary vertices, while the CH sequences are generated from objects' CHs. The object's boundary and CH are extracted by an active contour model. Ontology search is performed top-down using cyclic sequence alignment, which provides a scaling and rotational invariant matching. Experimental results are given to validate the theory, and the paper concludes with a list of contributions and discussion.