Research And Implementation Of An Intelligent Analysis System For Ocean Buoy Safety Based On Machine Learning

Marine buoys are a kind of important marine monitoring equipment.which can carry various hydrological,meteorological,and biochemical sensors to enabling real-timely observe the ocean and collect marine environmental data all around the year.However,due to complex and volatile sea conditions,harsh marine environments,electromagnetic interference,and aging equipment,it is very challenging to ensure the safety,stability,and reliability of buoys.Ocean buoy is often damaged and impacted,and its internal sensor failures often occur.Therefore,it is of great significance to automatically detect potential dangerous targets around the buoy platform,and diagnose sensor status in real time.It is important to fully guarantee the safety of the buoy platform,enabling it to conduct long-term and stable-online observation,and ensuring the authenticity and reliability of the observation data of various sensors on the buoy platform.Therefore,this thesis focuses on researching and developing an intelligent analysis system for marine buoy safety in two aspects,including fault diagnosis of marine buoy sensors and detection of external dangerous targets.Firstly,this thesis designs an intelligent data analysis system to diagnosis sensor failures inside the buoy.Statistical methods and artificial intelligence neural network models are used to analyze and train time series data observed by sensors,automatically detect abnormal data from long-term sequential data samples,calculate data accuracy,and evaluate sensor failure rates.On one side,wavelet packet decomposition and reconstruction technology be used to decompose and extract information from the time series data observed by sensors.On the other side,an improved radial basis function neural network model and K-nearest neighbor classification model are established to train the characteristics of sensor data,identify abnormal data,identify sensor fault types,and diagnose 6 types of wave sensor faults a.for improving the quality and reliability of monitoring data of ocean buoys.Secondly,in order to solve the problem that marine buoys are damaged by external dangerous targets this thesis proposes a lightweight intelligent target recognition model based on YOLOv5.By adding a channel attention mechanism(ECA),adjusting the feature pyramid pooling layer(SPPF),and creating a small target detection layer,automatic recognition of people,ships Automatic identification and capture of potential hazards such as large marine life buoys,an improved YOLOv5 model is developed to accomplish high-accurately identification of potential danger around buoys.Furthermore,the lightweight intelligent target recognition model is hardware-cured and packaged.Finally,ocean sensor fault diagnosis results and video target detecting monitoring results,as well as electronic chart data,buoy position data,the position of ships around the buoy,and buoy status information,are integrated into a marine buoy monitoring and management platform system.Using multi-source data fusion technology,this thesis combines technologies of software and hardware design,to develop the marine buoy monitoring and management platform system based on B/S software development architecture and the raspberry pie hardware platform.This system establishes an integrated data monitoring and display platform to monitor and analyze the internal sensor status and external dangerous targets of the buoy in real time.