基于深度学习技术的藻类智能监测系统开发

浮游藻类对水环境的变化非常敏感,是评价水环境质量的重要指示生物。传统的浮游藻类监测依靠人工采样分析,需要专业检测人员使用显微镜对藻细胞逐一鉴定并计数,耗时耗力且严重依赖检测人员的专业知识与鉴定经验,限制了浮游藻类监测工作的标准化推广和普及应用。利用神经网络模型建立了一套浮游藻类智能监测系统,该系统能够实现浮游藻类检测的自动进样、自动显微摄影,同时充分发挥深度学习技术在视觉分析领域的优势,自动进行浮游藻类智能识别和计数。使用大量浮游藻类样品开展了深度学习模型训练和结果验证,结果表明,该浮游藻类智能监测系统能够顺利完成浮游藻类样品自动进样、拍摄、鉴定和计数等一系列流程,且智能识别系统鉴定计数结果与人工镜检结果的误差较小。该系统还具有进一步的泛化和拓展能力,随着后续模型训练样品数量的增多,系统识别效率和精度可得到进一步提升,在浮游藻类监测及研究领域具有广阔的应用前景。

Research on Monitoring System of Algae Detection and Classification Based on Deep Learning

Phytoplankton is used as important indicators to assess water quality owing to its sensitivity to the changes of water environment.However,phytoplankton community analysis requires artificial identification.Manual identification process is very time-consuming and heavily relies on the professional knowledge and experience of the inspectors,which limits the standardization and popularization of the monitoring of planktonic algae.In this study,a set of intelligent monitoring system for planktonic algae was established using the neural network model.The system can realize automatic sampling and microscopic photography for planktonic algae detection,and at the same time give full play to the advantages of deep learning technology in the field of visual analysis,and automatically perform the intelligence of planktonic algae Identify and count.A large-scale microscopic algal sample was used for deep learning model training and result verification.The results showed that the planktonic algae intelligent monitoring system could successfully complete a series of processes such as automatic sampling,photographing,identification and counting of planktonic algae samples,and the error between the identification and counting results of the intelligent identification system and the results of manual microscopy was small.As more algal samples were used for deep learning training,the accuracy will be further improved.Our system showed promising performance on algal detection,classification and counting.