Files in this item
|(no description provided)|
|Title:||Machine vision microscopy as an on-line sensor for bioprocesses|
|Author(s):||Richburg, Brent Allen|
|Doctoral Committee Chair(s):||Reid, John F.|
|Department / Program:||Agricultural and Biological Engineering|
|Discipline:||Agricultural and Biological Engineering|
|Degree Granting Institution:||University of Illinois at Urbana-Champaign|
|Abstract:||A machine vision microscopy system (combination of video hardware and software and microscope for automated identification) was developed for classifying and counting microscopic objects in a Bacillus thuringiensis (Bt) fermentation. Image morphology properties consisting of area, perimeter, major length, minor length, number of holes, area of holes, and perimeter of holes were collected for vegetative cells, spores and protein crystals, and cells with included spores and crystals. Spatial resolution was 0.1333 $\mu$m/pixel in the horizontal direction and 0.1667 $\mu$m/pixel in the vertical direction.
A distance classifier and a neural network classifier were evaluated for accuracy using a limited data set. The neural network classifier was selected, refined, and implemented into software to classify, count and identify objects on the vision system display. The classifier accurately identified 92.0 percent of vegetative cells and 91.0 percent of spores when compared to human classification. Vegetative cell counts were 97.3 percent accurate and spore counts were 77.9 percent accurate. Identification of cells with included spores/crystals was 16.7 percent accurate and count accuracy for this class was 27.8 percent. The system did not accurately identify cells with inclusions due to their imaging characteristics. However, the system classified spores and vegetative cells with enough precision for determination of changes in cell and spore population over time. Spore counts from the machine vision system were plotted over time and showed that the system could identify when spore population began to increase and when it reached a maximum.
Object counts were performed in approximately 1.7 seconds for live images. The count results from five images were averaged together to get object counts. Averaging more images did not lower variability or change the mean count. The counts obtained from the machine vision microscopy system accurately followed trends in vegetative cell growth when compared to optical density measurements. The machine vision microscopy system predicted beginning and end of log growth within an hour of those predicted by optical density measurements.
|Rights Information:||Copyright 1992 Richburg, Brent Allen|
|Date Available in IDEALS:||2011-05-07|
|Identifier in Online Catalog:||AAI9236578|
This item appears in the following Collection(s)
Graduate Dissertations and Theses at Illinois
Graduate Theses and Dissertations at Illinois
Dissertations and Theses - Agricultural and Biological Engineering