Automation of pollen analysis using a computer microscope : a thesis presented in partial fulfilment of the requirements for the degree of Master of Engineering in Computer Systems Engineering at Massey University
Loading...
Date
2004
DOI
Open Access Location
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Massey University
Rights
The Author
Abstract
The classification and counting of pollen is an important tool in the understanding of processes in agriculture, forestry, medicine and ecology. Current pollen analysis methods are manual, require expert operators, and are time consuming. Significant research has been carried out into the automation of pollen analysis, however that work has mostly been limited to the classification of pollen. This thesis considers the problem of automating the classification and counting of pollen from the image capture stage. Current pollen analysis methods use expensive and bulky conventional optical microscopes. Using a solid-state image sensor instead of the human eye removes many of the constraints on the design of an optical microscope. Initially the goal was to develop a single lens microscope for imaging pollen. In-depth investigation and experimentation has shown that this is not possible. Instead a computer microscope has been developed which uses only a standard microscope objective and an image sensor to image pollen. The prototype computer microscope produces images of comparable quality to an expensive compound microscope at a tenth of the cost. A segmentation system has been developed for transforming images of a pollen slide, which contain both pollen and detritus, into images of individual pollen suitable for classification. The segmentation system uses adaptive thresholds and edge detection to isolate the pollen in the images. The automated pollen analysis system illustrated in this thesis has been used to capture and analyse four pollen taxa with a 96% success rate in identification. Since the image capture and segmentation stages described here do not affect the classification stage it is anticipated that the system is capable of classifying 16 pollen taxa, as demonstrated in earlier research.
Description
Keywords
Optical data processing, Image processing -- Data processing, Palynology -- Computer programs, Palynology -- Data processing