Welcome to Overseas Expertise Introduction Center for Discipline Innovation of Watershed Ecological Security in the Water Source Area of the Mid-line Project of South-to-North Water Diversion!

学院主页      | 加入收藏 
首页 > 研究前沿 > 正文

研究前沿

【研究进展】基于深度学习的天然水体(活体)浮游生物的自动标识及计数

发布日期:2023-06-17  文章来源:   点击数:

Zhuo Chen, Meng Du, Xu-Dan Yang, Wei Chen, Yu-Sheng Li, Chen Qian,* and Han-Qing. Deep-Learning-Based Automated Tracking and Counting of Living

Plankton in Natural Aquatic Environments.Environ. Sci & Technol.,2023


【研究进展】基于深度学习的天然水体(活体)浮游生物的自动标识及计数[Environ. Sci & Technol.,2023]http://www.jlakes.org/ch/reader/view_news.aspx?id=20230612140929001


ABSTRACT: Plankton are widely distributed in the aquatic environment and serve as an indicator of water quality. Monitoring the spatiotemporal variation in plankton is an efficient approach to forewarning environmental risks. However, conventional microscopy counting is time-consuming and laborious, hindering the application of plankton statistics for environmental monitoring. In this work, an automated video-oriented plankton tracking workflow (AVPTW) based on deep learning is proposed for continuous monitoring of living plankton abundance in aquatic environments. With automatic video acquisition, background calibration, detection, tracking, correction, and statistics, various types of moving zooplankton and phytoplankton were counted at a time scale. The accuracy of AVPTW was validated with conventional counting via microscopy. Since AVPTW is only sensitive to mobile plankton, the temperature- and wastewater-discharge induced plankton population variations were monitored online, demonstrating the sensitivity of AVPTW to environmental changes. The robustness of AVPTW was also confirmed with natural water samples from a contaminated river and an uncontaminated lake. Notably, automated workflows are essential for generating large amounts of data, which are a prerequisite for available data set construction and subsequent data mining. Furthermore, data-driven approaches based on deep learning pave a novel way for long term online environmental monitoring and elucidating the correlation underlying environmental indicators. This work provides a replicable paradigm to combine imaging devices with deep-learning algorithms for environmental monitoring.

KEYWORDS: environmental monitoring, plankton abundance, deep learning, tracking, detection, aquatic environment