Discrimination of Visible Impurity in Bottled Oral Liquid Based on Mixed Characteristic

Abstract

On the issues of the various kinds of foreign substances in bottled oral liquid, simplex method of decision and low speed on visible foreign matter, we presents a fast determination algorithm based on machine vision. The algorithm uses image graying and GAUSSIAN filtering to get the target to be detected by frame difference method. The images were then processed by Laplace sharpening and binarization, and the SURF algorithm is used to match the foreign matters in multi frames. Finally, the matching suspicious points are judged by weighted feature. The experiment shows, this algorithm can overcome the deficiency of single feature, it can detect all kinds of impurities accurately. Meanwhile, this algorithm deals with fast speed, it can meet the requirements of online production.