Recognition on the working status of Acetes chinensis quota fishing vessels based on a 3D convolutional neural network

Abstract

In the present study, a method for identifying the status of Acetes chinensis fishing vessels based on a 3D convolutional neural network is proposed, so as to protect marine biodiversity, monitor the working status of Acetes chinensis fishing vessels and assist in the realization of quota fishing. The Vessel Monitoring System (VMS) was installed on the quota fishing vessels to collect work data from June 16, 2021 to July 13, 2021. According to the characteristics of the fishing vessels, the work status of the fishing vessels was divided into five statuses such as stopping, sailing, putting net, waiting and pulling net. The 3D convolutional neural network Acetes3DNet was designed to extract the multi-dimensional and multi-level features of the data and trained in the training set. Finally, the effectiveness of the model was verified in the validation set. The training results were combined with the Beidou ship position data to restore the working process of the fishing vessel. The experimental results reveal that after 150 epochs of training, the precision, recall, and f1 score of Acetes3DNet on the training set reached 99.02%, 99.19%, and 99.09%, respectively, while the precision, recall, and f1 score on the validation set reached 97.09%, 96.82%, and 96.68%. Research shows that Acetes3DNet can circumvent the limitations of traditional 2D neural networks in dynamic target detection, complete recognition of the working status of Acetes chinensis quota fishing vessels, and show the historical work process of the ship in an intuitive manner. The experimental results are conducive to standardizing the management of fishing vessels and protecting marine life.