Rapid detection for optical micrograph of plankton in ballast water based on neural network

Abstract

There exist various types of organisms in ballast water, among which Escherichia coli and harmful plankton pose a great threat to the marine environment. The automatic and rapid detection of plankton in ballast water is crucial for protecting the marine environment. In this work, A convolutional neural network based on Mobilenetv3 was proposed to rapidly detect plankton. The model can detect five species plankton, Chattonella marina, Alexandrium tamarense, Prorocentrum donghaiense, Phaeocystis globosa, and Karlodinium veneficum. And four traditional neural networks were used for performance comparison. The experimental results show that, compared to the traditional neural network, the developed neural network has a higher speed and accuracy of recognition. The use of neural networks to detect plankton will greatly improve the protection of marine environments.