Pivotal Technology for the Detection of Lens’ Defects Based on Image Recognition

ABSTRACT

With flaws such as scratches, dirt and pitting on the surface of the lens as the research object in this paper, the key technology of machine vision inspection based on in-depth analysis of image recognition was studied as well as the programs of extraction and quantification based on image recognition on the surface of the lens. First, for the insufficient edge destruction of the traditional de-noising method, an improved filtering algorithm has been proposed. Then, the interference of the lens frame is removed when calculating the size of the defects by using the circular region detection algorithm. And a new improved calculation method of small connectivity target based on mathematical morphology by pixels’ scanning and marking has been proposed. Finally, artificial neural networks were introduced and by using BP neural network classifier, the classification of scratches and pitting was realized. Experimental results showed certain guiding significance for detecting the surface quality of the lens based on image processing technology.

Keywords: