颗石藻自动鉴定系统及其古海洋学应用——以南海MD05-2901柱状样研究为例

APPLICATION OF AUTOMATIC RECOGNITION SYSTEM FOR COCCOLITHS IN PALEOCEANOGRAPHY:A CASE STUDY AT SITE MD05-2901,SOUTH CHINA SEA

颗石藻自动鉴定系统SYRACO(Système de Reconnaissance Automatique de Coccolithes)是一种通过人工智能神经网络自动识别颗石藻种类并统计数量的工具软件。该软件能够快速可靠识别观察视域中的所有颗石个体,大幅度提高工作效率并弥补了在颗石藻属种鉴定过程中的人为误差。经过训练的SYRACO系统可以鉴定第四纪以来14个主要颗石藻属种,适用于第四纪以来的古环境研究。作者使用SYRACO自动鉴定系统对南海MD05-2901柱状样进行颗石藻属种鉴定,同时获得了Florisphaera profunda,Emiliania huxleyi,Gephyrocapsa oceani-ca等主要属种含量信息。将结果与先前人工统计数据对比表明二者具有很好的一致性,证明其在古海洋学研究中的应用价值。

Automatic recognition system for coccoliths （Systeme de Reconnaissance Automatique de Coceolithes, SYRACO） is a tool for automatic identification and count of coccolith by artificial neural networks. This software is able to identify individual coccoliths reliably at the same time as it is able to recognize almost all coccoliths present in a field of view, which would improve work efficiency significantly and prevent manual errors in routine work. The trained SYRACO system could identify 14 Quaternary coccolith taxa. We employ SYRACO to analyze coccolith specimens at Site MD05-2901 in the South China Sea and obtain the percentages of Florisphaera profunda, Emiliania huxleyi and Gephyrocapsa oceanica. The incorporation of the automatic recognition results and previous manual counts proves the utility of SYRACO on paleoceanographic research.