Rotation Invariant Local Binary Pattern Based On Glcm For Fluorescent Tube Defects Classification

In this paper, a machine vision system is developed for fluorescent tube defects classification, a new rotation invariant method is presented for texture analysis. The objective of research is to study a new texture analysis method and classify the surface defects of fluorescent tubes. Tri-phosphor fluorescent powder is sprayed onto the surface of glass tube; some defective products are made during the spraying process, how to find and classify them are important. In this research, the characteristics of fluorescent tube are studied; an algorithm of rotation invariant is presented to find the differences between the defects. In the algorithm, GLCM (Gray-Level Co-occurrence Matrices) is calculated to get the orientation of defect, LBP (Local Binary pattern) vector is got along the orientation, then the detects are classified according to the combined features and distance formulas. The experiment results show that the new method is more convenient and effective to classify typical defects.