Automated image-based quality control of molecularly imprinted polymer films

Way, Andrew R. (2019) Automated image-based quality control of molecularly imprinted polymer films. Masters thesis, Memorial University of Newfoundland.

[English] PDF - Accepted Version Available under License - The author retains copyright ownership and moral rights in this thesis. Neither the thesis nor substantial extracts from it may be printed or otherwise reproduced without the author's permission. Download (14MB)

Abstract

We present results of applying a feature extraction process to images of coatings of molecularly imprinted polymers (MIPs) coatings on glass substrates for defect detec- tion. Geometric features such as MIP side lengths, aspect ratio, internal angles, edge regularity, and edge strength are obtained by using Hough transforms, and Canny edge detection. A Self Organizing Map (SOM) is used for classification of texture of MIP surfaces. The SOM is trained on a data set comprised of images of manufactured MIPs. The raw images are first processed using Hough transforms and Canny edge detection to extract just the MIP-coated portion of the surface, allowing for surface area estimation and reduction of training set size. The training data set is comprised of 20-dimensional feature vectors, each of which is calculated from a single section of a gray scale image of a MIP. Haralick textures are among the quantifiers used as feature vector components. The training data is then processed using principal component analysis to reduce the number of dimensions of the data set. After training, the SOM is capable of classifying texture, including defects.

Actions (login required)

View Item