Prediction Of Fishing Boat Operation Mode Based On Beidou Trajectory Sequence Data

The Beidou system is a global satellite navigation system developed by our country.It has been widely used on fishing vessels and can be used to provide fishermen with location information.With the advent of the era of big data,we can make full use of the large amount of trajectory data of fishing boats provided by the Beidou system to provide more valuable services.At the same time,there are three main ways to operate fishing boats in the ocean: trawling,purse seine and drift net.Different methods of operation are used to catch different kinds of fish.The thesis uses the fishing vessel trajectory data provided by the Beidou fishing vessel monitoring system to judge the operation mode of the fishing vessel.Through the study of the operation mode of the fishing vessel,the relevant fishery department can further understand the number of various fish caught and various fish.The distribution of similar living areas can make a fishery plan that is more in line with sustainable development and ensure the development of fisheries and the ocean in a healthier direction.The work of this thesis is as follows:(1)The trajectory data of the fishing vessel collected by the Beidou fishing vessel monitoring system is the initial data set.Eliminating abnormal data such as speed,and processing the data set.Perforing feature engineering on the initial data set to extract key information about the trajectory of the fishing boat,including speed and heading related features,latitude and longitude related features,and the interactive features between speed and longitude and latitude.An integrated feature selection method is designed to perform feature selection for more than 100 features of the structure.Using the lightgbm algorithm to train the model on the selected features,so as to recognize the type of fishing operations of the fishing boat.This method has reached 0.915,0.885,and 0.908 in accuracy,recall and precision.(2)Constructed fishery based on Long Short-Term Memory(LSTM)neural network(Long Short-Term Memory)Boat behavior recognition method,this method recognizes the fishing boat operation mode from the time series characteristics.The input,output and structure of the LSTM model are determined,and the data collected by the Beidou equipment is fully utilized.The input to the model includes four features:speed,longitude,latitude,and heading.The parameters of the model are selected through experiments to improve the accuracy and precision.Rate,recall rate and F1-score are used as the evaluation indicators of experimental effects,and the influence of hyperparameters and data processing methods on the model is analyzed.This method has reached 0.893,0.852,0.890 in accuracy,recall and precision.(3)A deep learning fishing boat behavior recognition method based on the attention mechanism is constructed,and the input features are longitude,latitude,speed,and heading.This method uses two LSTM structures.One LSTM receives the speed and heading information of the fishing boat,and is used to extract the speed of the fishing boat and the higher-level information contained in the heading,such as returning to the port,finding a fishing spot,and sailing to a fishing spot.,And then another LSTM structure receives the latitude and longitude information of the fishing boat,which is used to extract the higher-order information contained in the latitude and longitude of the fishing boat,such as the main working area of the fishing boat and some spatial information of the fishing boat.Then the attention mechanism is used to integrate the information extracted from the two structures.Experiments have verified the superiority of this method in the recognition of fishing boat operation mode,and compared it with the single LSTM receiving all the information.Furthermore,the two LSTM structures are replaced with the currently very popular pure attention structure transformer,and the speed information and the latitude and longitude information are input into the encoder and decoder of the transformer,respectively.The article designes a transformer-based pure attention mechanism fishing boat operation mode recognition model SAFBC: Self-Attent Ive Neural Fishing Boats Classfication.The model can achieve 0.918,0.883,0.919 in accuracy,recall,and precision through simple information input.