IDENTIFICATION OF CELL TYPE OF MICROALGA Haematococcus lacustris BY IMAGE RECOGNITION

Abstract

 Microalga Haematococcus lacustris produces astaxanthin, a carotenoid, under stressful conditions. The alga changes its cells into three main types (flagellate, palmelloid, and cyst) depending on environmental conditions. In the course of studying the method of H. lacustris dominant culture to realize a resource recovery wastewater treatment process that removes and recovers nutrients while producing astaxanthin, image recognition techniques was investigated to understand the cell types. With OpenCV, a classifier that differentiated between spherical cells (palmelloids and cysts) and flagellates with 84% of accuracy. However, a two-step classification was required to further classify palmelloids and cysts which have red color derived from astaxanthin by cell color. Simultaneous recognition of the three cell types by YOLO resulted in a high accuracy of over 90% for all types. The effect of image quality on recognition accuracy was examined. The accuracy decreased significantly when lower quality images were used during recognition than during training.