Font Size: 

This paper introduces pattern matching algorithm from industrial camera based on geometric features. In industry production line, object recognition using vision is a challenging problem. In the object recognition, feature is an important element that represents object’s state. Although its large amount of information contains location, rotation and scale difference, geometric feature is hard to get because it is sensitive to noise. To overcome the weakness, we propose two types of geometric features. Then geometric features are detected in order to construct descriptor. The geometry based Vector Mapping Descriptor (VMD) for pattern matching is proposed to effectively match salient feature points between different images under geometric transformation regardless of missing or additional feature points. VMD represents the correlation of features that includes Euclidean distance and angle. The group of one to one corresponding feature points on different images results in the completed object matching. The proposed algorithm is invariant to translation, rotation and scaling difference. To demonstrate the performance of the proposed algorithm, we conducted an experiment with both reference image data and real-time industrial camera. The result provides accurate and robust feature matching.

Pattern Matching, Industrial Camera, Geometric Feature, Vector Mapping Descriptor