Create your own conference schedule! Click here for full instructions

Abstract Detail



Bioinformatic and Biometric Methods in Plant Morphology

Han, Jing Ginger [1], Cao, Hongfei [1], Punyasena, Surangi [2], Restrepo, Alejandra [3], Jaramillo, Carlos [4], Shyu, Chi-Ren [1].

Shape-Based Search of Neotropical Pollen Images using Morphological Characteristics.

A large degree of manual sorting of morphological types is required in traditional pollen analysis. Computer automation is needed to improve the throughput of the research, streamline the laborious manual process, and identify synonymies. In this collaborative research, palynologists collect microscopy images of pollen and spores with largely diverse taxa and computer scientists develop a shape-based search engine to efficiently annotate and retrieve the microscopy images containing pollen grains with most similar morphological characteristics and statistics of distribution within specimen. The palynologists are presented with two options of search – search by pollen grain image examples or search by morphological semantics that are of most interest. The system is developed around the standard reference images and annotations made by the palynological expert, which are currently only used to qualitatively categorize and describe pollen morphotypes.
Visual features, such as shape, size, and texture, are extracted from individual pollen grain images to represent the content of an image and are stored in a customized high-dimensional database indexing structure. The “search by example” engine takes a query image uploaded by palynologists for a new study, annotates extracted grains, and retrieves database images that are most similar to the new image. When there are multiple pollen grains that are of interest in a microscopy image, we comprise the query as a “bag of grains” form to search against the database that are comprised of ‘bags of grains’ from all other microscopy images. The top retrieved images contain similar distribution statistics of pollen grain types. The “search by semantics” engine searches the database based on visual features that best describe the morphological semantics provided by the palynologists. A mathematical model has been built to use possibility functions to automatically retrieve images that are relevant to the query semantics (a collection of morphospecies). The result images are sorted based on the similarity in the semantic space.
This shape-based search engine is applied to a set of Miocene Neotropical images as a case study. This is a growing dataset of images, with ~450 morphological types. This database search engine could be applied in the practice of pollen and spore counting, providing palynologists an efficient tool to automate the annotation and reduce the work load and, in the meantime, potentially assist the discovery of novel paleoecological knowledge.
This project is supported by the National Science Foundation under grant numbers DBI-1053024 and DBI-1052997

Broader Impacts:


Log in to add this item to your schedule

1 - University of Missouri, Department of Computer Science, 241 Engineering Building West, Columbia, MO, 65211, USA
2 - University of Illinois, 505 S Goodwin Ave, 139 Morrill Hall, Urbana, IL, 61801, USA
3 - University of Illinois, Plant Biology, 505 S Goodwin Avenue, Urbana, IL, 61801, USA
4 - Smithsonian Tropical Research Institute, Center for Tropical Paleoecology and Archaeology, Ancon, Panama, Panama

Keywords:
machine vision
palynology
database
Miocene
search engine.

Presentation Type: Symposium or Colloquium Presentation
Session: C2
Location: Prince of Wales/Riverside Hilton
Date: Monday, July 29th, 2013
Time: 3:45 PM
Number: C2009
Abstract ID:896
Candidate for Awards:None


Copyright 2000-2012, Botanical Society of America. All rights reserved