Polluted and Perspective Deformation DataMatrix Code Accurate Locating Based on Multi-features Fusion

In this paper we present a method for extracting the best candidate edge combination based on multi-features fusion, aiming at the challenges of accurately locating the DataMatrix (DM) code (hereinafter referred to as DM code) with pollution and perspective deformation. Firstly, DM code edges are transformed from image into Hough domain in which linear feature is more prominent. We are able to obtain the valid combinations of candidate marginal points after prior rules-based filtering. Then, we design and extract four boundary features of the finder pattern in image domain. Meanwhile, we establish the model of distorted DM code edge distribution in Hough domain and extract the corresponding features. Finally, we merge the multi-features according to the DS theory and make the final locating based on the fusion result. Compared with traditional methods, the experiments demonstrate the greater robustness and flexibility of our proposed approach to accurately detecting the contaminated Data Matrix coexisted with perspective deformation.