A Fast Image Matching Algorithm and the Application on Steel-Label Recognition

This paper presents a kind of image matching algorithm based on two steps matching process. This algorithm could measure the similarity between two images fast and accurately. In the first stage, the color ring-projection transforms the 2-D color image contained in a circular window into 1-D color signals as a function of radius. In the second stage process, a rotation invariance moments based on a set of complex polynomials are used to determine the final similarity. The proposed algorithm has shown promising result based on the experiments on steel-label recognition. It has a good robust performance and excellent computational efficiency.