Research On Key Technologies Of Smart Fishing Based On AIS Data

Fishery,an old and emerging industry,has been accompanied by human civilization.In recent years,serious issues of global food security have highlighted the importance of fishery.There are serious problems such as over-utilization of resources and illegal fishing in fishing industry alone.In this intelligent age of connectivity,smart fishing is essential to manage fisheries better and ensure the sustainability of fishery.Among them,the keys are understanding the fishing status of fishing vessels comprehensively,grasping the fishing effort macroscopically and analyzing the spacetime changes of fishing grounds accurately.In view of the key problems of smart fishing,this paper proposes methods to identify fishing status of fishing vessels,predict the main engine power of vessels and analyze fishing grounds based on the data of Automatic Identification System(AIS).The main work and innovation points of this paper are as follows:(1)Identifying fishing status of fishing vessels correctly is the basis of monitoring fishing behaviors.In this paper,a model used to identify fishing state based on integrated Convolutional Neural Networks(CNN)is constructed for different types of fishing vessels.Firstly,the original data were analyzed and preprocessed,including outlier processing,track segmentation,derived features and data standardization.Secondly,an integrated CNN model is built based on Inception-Res Net architecture,whose input is a combination of derived features and original features.The convolutional level uses feature integrated method to fuse the outputs of each layer,and the Sigmoid function is used for the final classification.Finally,the model was compared with the traditional models including Support Vector Machine(SVM)and CNN,which proves the effectiveness of the integrated CNN model in identifying fishing status of fishing vessels.(2)The main engine power is an important parameter for estimating fishing effort.A new method based on Gaussian Mixture Models(GMM)and Deep Neural Networks(DNN)is proposed for predicting the main engine power of large vessels.Firstly,the correlation analysis of vessel features was carried out,and the vessel features with larger correlation coefficient are selected as the input of the GMM-DNN model.Secondly,GMM is used for cluster analysis of vessel features,and the clustering results are used as labels,which are used as inputs to DNN with vessel features.At the same time,Adam-Dropout is used to optimize model.Finally,in order to explore the effectiveness of the method,we compare the prediction effects of multiple linear regression,nonlinear regression,DNN and GMM-DNN.The experiment shows that the GMM-DNN model has the best performance in the main engine power prediction.(3)The space-time distribution of fisheries is the data support for the realization of smart fishing.Based on the above algorithms,we propose a method for fishing ground analysis based on fishing effort,and exploit a smart fishing system independently.Firstly,select a moderate spatial resolution and analyze the fishing time and fishing effort.Secondly,we use global Moran index to judge the data correlation and clustering in the space.The method of kernel density estimation is used for drawing and analyzing fisheries thermodynamic chart based on fishing effort.Compared with the traditional fisheries analysis method based on the fishing time,this method is more accurate and effective.Finally,in order to realize smart fishing,this paper exploit a shipboard smart fishing system,which can be combined with shipboard intelligent sensor network system and intelligent shore-based big data center to form an intelligent ship networking platform,aiming to improve efficiency,reduce cost and ensure safety through big data and intelligent algorithms.