Fiber Length Measurement by Image Processing

Show full item record

Title: Fiber Length Measurement by Image Processing
Author: Ikiz, Yuksel
Advisors: Jon P. Rust, Chair
H. Joel Trussell, Member
Robert A. Barnhardt, Member
Warren J. Jasper, Member
Abstract: This research studied the accuracy and feasibility of cotton fiber length measurement by image processing as an alternative to existing systems. Current systems have some weaknesses especially in Short Fiber Content (SFC) determination, which is becoming an important length parameter in industry. Seventy-two treatments of five factors were analyzed for length and time measurements by our own computer program. The factors are: Sample preparation (without fiber crossover and with fiber crossover), lighting (backlighting and frontlighting), resolution (37-micron, 57-micron, 106-micron, and 185-micron), preprocessing (4-neighborhood and 8-neighborhood), and processing (outlining, thinning, and adding broken skeletons). The best results in terms of accuracy, precision and analysis time for images without fiber crossovers were: 106-micron resolution with frontlighting using an 8-neighborhood thresholding algorithm and using an outline algorithm for length determination. With fiber crossovers, 57-micron resolution with backlighting using an 8-neighborhood thresholding algorithm and using a thinning algorithm combined with an adding algorithm for combining broken skeletons. Using the above conditions, 1775 area can be analyzed using our current equipment in 15 seconds. In the case of images with crossovers, only 117 can be analyzed in 15 seconds. This research demonstrates that successful sample preparation without fiber crossovers would create the best fiber length measurement technique, however with fiber crossovers the system efficiency has been proven as well.
Date: 2000-08-10
Degree: PhD
Discipline: Fiber and Polymer Science
URI: http://www.lib.ncsu.edu/resolver/1840.16/4642


Files in this item

Files Size Format View
etd.pdf 1.335Mb PDF View/Open

This item appears in the following Collection(s)

Show full item record