Image analysis as a tool for quantitative phycology: a computational approach to cyanobacterial taxa identification

摘要: In the following work we discuss application of image processing and pattern recognition to field quantitative phycology. We overview area review previously published literature pertaining analysis phycological images and, in particular, cyanobacterial processing. then main operations used process quantify data contained within them. To demonstrate utility cyanobacteria classification, present details an system for automatically detecting classifying several taxa Lake Biwa, Japan. Specifically, initially target genus Microcystis detection classification from among species Anabaena. subsequently extend classify a total six species. High-resolution microscope containing mix above other nontargeted objects are analyzed, any detected removed further analysis. Following enhancement, measure object properties compare them compiled database characteristics. Classification as belonging particular class membership (e.g., “Microcystis,”“A. smithii,”“Other,” etc.) is performed using parametric statistical methods. Leave-one-out results suggest error rate approximately 3%.