Fishing Vessel Behaviors Classification And Visualization

Fishing vessel operation plays an extremely important role in offshore operations in general.The effective classification and identification of the fishing vessel operation behavior is beneficial to the maritime traffic scheduling and fishery safety production in the future.However,a series of problems,such as the losses of data in the process of information collection or uploading,as well as the spatio-temporal data characteristics,result in inaccurate behavior identification of the fishing vessels,which may not only endangers the safety of offshore operators and lead to property losses,but also increases the probability of maritime traffic hazards.Taking the spatio-temporal data recorded by Beidou devices of a large number of fishing vessels as research data,this thesis classifies the operation behavior of each fishing vessel by analyzing the historical track information,so as to assist the Fishery Management Department to identify the specific type of fishing vessel operation.Meanwhile,through the visual fishing vessel operation,this thesis also aims to analyze,predict and judge the fishing vessel operation behavior,thus guiding the safe production of fishery.Basically,the main research contents of this thesis are as follows:(1)Construction of the feature engineering.In view of the less features of the original data collected,and the problem of data missing,error data and data redundancy in the process of data collection or upload,the feature engineering is constructed after analyzing the original data variables.On the whole,it mainly includes two parts.The first part is the feature engineering based on latitude,longitude,speed and direction,which mainly uses statistical methods to obtain new features.The second part is the feature engineering based on trajectory information,which mainly uses Geo Hash and natural language processing methods such as Word2 Vec,TFIDF and Count Vectorizer to construct new features.(2)The classification model training of fishing vessel operation behavior.Feature engineering is constructed and fused with original data to obtain 305-dimensional features which can be used as the model input.The 5-fold machine learning method Light GBM is used to classify the fishing vessel operation behavior to judge the operation behavior of fishing vessel.The construction of the model makes full use of the input feature information.By setting the parameters of the model,a series of factors including precision,recall,F1-Score,accuracy,macro average and weighted average are taken as evaluation indexes of the model effect.Combining the built-in function plot＿importance of Light GBM with the third-party library Matplotlib of Python,the feature importance bar graph of this model is drawn,and then the top 10 features that contribute the most to the experimental results are obtained.Meanwhile,the model achieves the goal of accurately identifying fishing vessel operation behaviors and types at sea.In addition,compared with XGBoost,SVM,GBDT and Random Forest algorithms already used in literature,the accuracy of Light GBM adopted by this model is up to 0.9390,which indicates that it has the best classification effect.(3)The visualization of fishing vessel operation.Based on Vue.js and Spring Boot framework,this model not only realizes the separate development of foreground and background,but also realizes the data interaction with My SQL database by JPA.At the same time,the ECharts tool is used to import available charts,and Baidu map API is used to import Baidu map.In short,the model has realized the following four modules: a)the group operation module including fishing vessel operation behavior types and maritime status;b)the sea area distribution module including purse seine operation,trawl operation and gill net fishing operation,which is convenient to analyze the distribution of these three operations in the offshore;c)the ship positioning monitoring module including fishing vessel positioning monitoring and fishing vessel trajectory;d)the training model is applied in the classification module of fishing vessel operation,thus realizing the classification calculation.