Research On Data Mining Of Fishing Vessel Sailing Based On Artificial Intelligence

Marine ships use the Automatic Identification System(AIS)terminal to transmit massive amounts of ship positioning data to the AIS data center through satellite communications,ground base stations,etc.Among them,the AIS terminal can realize data exchange between marine vessels.After the AIS data center processes the fishing vessel positioning data,it is stored in the database in a structured manner.Ship trajectory data can realize ship trajectory prediction,collision detection and navigation safety in and out of ports through Vessel Monitoring System(VMS)and Vessel Traffic Services(VTS).In addition,the ship positioning data processed by big data and machine learning technology can be used to analyze the temporal and spatial distribution of fishery resources,the identification of fishing boat operations,and the illegal operations of fishing boats.In this paper,through data preprocessing,machine learning,deep neural network and other technologies,the massive AIS data of fishing vessels are studied,and the algorithm also is verified to ensure the significance of the research in the actual production activities.The main work content and innovations of this article are as follows:1.Design a preprocessing algorithm for fishing boat AIS data.The data used in this article comes from Zhejiang Ocean and Fishery Bureau and Ali Tianchi 2020 Digital China Innovation Competition.Among them,the data provided by the Zhejiang Ocean and Fisheries Bureau are the AIS data of fishing vessels from March to May of2016 and 2015,and the fishing vessels that provided the data are all registered in Zhejiang.In this paper,using the distributed computing method of Hadoop,the AIS data of all fishing boats are divided into all trajectory data of a single fishing boat sorted by time.In order to find the hot spots in the fishing area and the regularity of their changes,the AIS data of fishing vessels are divided separately according to year and month.Secondly,analyze and clean up data that is abnormal,wrong,or does not meet the range of latitude and longitude.Finally,for the extracted fishing boat trajectory data,the method of linear interpolation is used to fill in the missing data.This work laid the foundation for the subsequent analysis and mining of fishing vessel positioning data.2.Design a fishing vessel trajectory prediction model.In view of the multi-feature characteristics of fishing boat trajectory data and the requirements for accuracy and dynamics of trajectory prediction,first visualize all fishing boat trajectory data,and then pass the feature detection of ORB and BF(Brute-Force)Matching is used to calculate image similarity to classify the types of fishing boat trajectories,and then construct multi-feature time series trajectory data.Finally,a variant model of Bi-LSTM based on recurrent neural network is used.The processed trajectory data of Zhejiang fishing boats are used for model training,and the model is used for trajectory prediction.In this part,the prediction results are discussed according to the parameter selection of the neural network,and the experiment is compared with other methods.3.Design the recognition algorithm of fishing boat operation mode.Fishing vessel operation mode recognition is the use of big data and machine learning technology to analyze massive fishing vessel positioning data,statistics of speed,direction,operating space,operating time and other data of the conventional characteristics,mode characteristics,special conditions,and construct a multi-dimensional feature project.Then use Light GBM,a machine learning algorithm,to determine how it works.Using machine learning and other technologies to automatically identify the operation mode of the fishing vessel from the massive fishing vessel trajectory,it is possible to find the fishing vessel that violates the regulations and change the operation mode.At the same time,it can further understand the operation of a certain type of fishing vessel at a specific time and in a specific sea area,and then calculate the fishery resources of the sea area.Change the situation and strengthen the management of fishery resources.4.Design a hot spot discovery algorithm for fishing areas.First of all,the heat of fishing area hot spots in a week,a month and a year under the time interval,analysis of the changes in fishing area hot spots.Then,an algorithm for twice clustering to identify fishing hot spots is proposed,which uses K-Means clustering algorithm to select the clustering center of each fishing vessel,and then uses K-Means clustering again to obtain the final fishing hot spot,which will effectively obtain the fishing hot spot at a specific time.