Session: 16-02-01: Poster Session: NSF Research Experience for Undergraduates (REU), NSF Posters

Paper Number: 99712

99712 - Defect Detection and Printed Pattern Measurement for Roll-to-Roll Microcontact Printing 

This project focuses on the development of image processing techniques for defect detection and printed pattern measurement in roll-to-roll (R2R) microcontact printing (µCP) and the characterization of µCP stamp compression-deformation. The outcome of this project will offer researchers in R2R printing domain some insights into the sources and factors of defect formation. Moreover, it will also demonstrate the effects on deformations that using a rolling stamp have as well as the differences in elasticity that exist in the stamp, depending on its thickness. This work will improve the overall effectiveness and reliability of R2R µCP. 

We will acquire stamp deformation images using an in-situ imaging tool designed and established in the lab for a printing roller of the R2R µCP. Image processing algorithms were developed to extract patterns on the images coming from a camera mounted underneath and pointing at the printing roll. The processing consisted of using thresholding technique to segment the printed pattern from the background. Since the illumination is usually uneven resulting in the pattern in one spot having the same brightness value as the background in another spot, local adaptive segmentation techniques were introduced. However, many of the defects have similar pixel density values to the pattern itself, so the resulting segmented patterns are still mixed with defects in it. Therefore, we propose to separate the defects from printed patterns by recognizing the print pattern using the prior knowledge of repetitive and regular positioning in the R2R µCP product design and removing any pixels that don’t match the designed position. Another method could be to use a different threshold on the original image to segment out the defects. Defects are often darker than the rest of the pattern so can be easily segmented out. However, if there is an issue with the stamp, or uneven force was applied then there might be part of the pattern missing, in which case thresholding will not work. This can be remedied by using the known pattern we are printing to detect the missing pieces. Finally, the detected defects will be fed into an associative learning Artificial Neural Network (ANN) to further classify them into subcategories. Furthermore, detecting the deformation of the stamp can be done primarily through traditional image segmentation techniques with thresholding methods. 

Additionally, force data will be collected using an FMS RMGZ 922 Force Measuring Roller and FMS EMGZ310 Digital Microprocessor Controlled Tension Amplifier. The controller outputs an amperage that corresponds to the force acting on the Roller. This force data will be measured and synchronized to the stamp deformation data captured by the camera. As a result, the force and image segmentation data will be used to extract the stamp deformation-compression laws for in-situ quality control of µCP. 

Presenting Author: Ilya McCune-Pedit University of Massachusetts Amherst

Presenting Author Biography: Ilya McCune-Pedit is working toward his B.S. in mechanical engineering and B.A. in German studies at the University of Massachusetts Amherst. He is a teaching assistant, a club board member, and captain of his high school sports team. In his free time, he goes biking, skiing, and sailing, or repairs his computers. Ilya's research includes mechanical control and computer vision.

Authors:

Ilya McCune-Pedit University of Massachusetts Amherst
Isabella Lambros University of Massachusetts Amherst
Meysam Safarzadeh University of Massachusetts Amherst
Jixin Yin University of Massachusetts Amherst
Xian Du University of Massachusetts Amherst