Oil Identification And Intelligent Detection Using Multi-sensors For Marine Oil Spill

Marine oil spill is one of the most serious problems in the world.Oil spill is a frequent occurrence,and a series of problems are waiting to be answered: What time is it? Where does Spilled oil take place? What kind of oil is spilled? How much oil is spilled? Who spilled the oil? When the oil spill occurs,the time and place is easy to determine,but the next three questions are difficult to answer.This paper attempts to answer these three questions based on multi sensor and intelligent algorithms.Based on the experimental data collected by spectrometer and multi spectral camera and portable spectrometer,this paper studies the feature extraction and recognition model in the process of oil classification..Further from the viewpoint of spilled oil recovery spilled oil emulsion detection problem is concerned.Based on Airborne Hyperspectral Imaging Instrument,oil spill coverage and precise measurement of the area of the oil spill,we proposed the end of the quadratic extraction combined with non negative matrix factorization andabundance estimation method based on independent component analysis(ICA).Based on the near infrared spectrum data,the analysis of the oil spill source tracing is studied by using octane number as an example,the model is optimized and the analytic method based on support vector machine is also proposed.The specific research work of this paper is as follows:According to the classification of oil,the oil spill simulation experiment using ocean spectrometer,multispectral camera,hyperspectral camera.oil identificationis studied under the visible light,ultraviolet light and other modes of gasoline,diesel,kerosene,oil,crude oil and peanut oil and other oils.Based on Oil spill identification model based support vector machine is obtained by optimizing the spectral data.Then we build a multi spectral imaging system based on multi spectral and study oil spill feature recognition method based on independent component analysis.For hyperspectral imaging,oil spill is identified based on radiation index and model consensus and proves the feasibilityof the methods by experiments.As oil spill events tend to occur at night cruel environment,this paper provides a solution based on UV LED for the case of night spilling oil emulsion detection,.Accurate detection of the emulsion oil spill and area,has a positive significance in evaluating the damage.Through comparing the characteristics of combinatorial optimization,we optimize the feature selection and extraction model,and pointed out that Fisher method can choose specific band and has more practical significance than the recognition model based on PCA transform.We put forward a band selection method for spilled oil emulsion detection by band index optimal bands.This band selection method can be directly applied to the subsequent real-time optical detection device development,and provides an alternative model for night emulsion detection.Accurate detection oil spill volume is one of the major difficulties faced in spilled oil exploration,identification of oil mass need to accurately measure the area and thickness of spilled oil.Due to the limitation of the resolution of airborne and spaceborne sensors,to accurately determine the area still have difficulty in overflowing,and sensor for spilled oil thickness still is not applicable.This article attempts to estimate oil spill coverage from respectively two respects,and propose estimating method of spilled oil coverage based on endmember extraction and non negative matrix factorization with secondary endmember exploration budgets,and based on independent component analysis of a endmembers and abundances decomposition and estimation of spill oil covered.We verify verify the feasibility of the method through the simulation data and real spilled oil Airborne hyperspectral imaging data.Finally,according to the problem of oil spill tracing,this paper takes octane number as an example to study the preliminary analysis of gasoline components.Although the octane number prediction is an old problem,this paper based on the gasoline near infrared spectroscopy,for the first time study the octane number of a variety of models of combinatorial optimization problems,and finally the model has a more extensive application value.Also Lambert Beer law and spectral mixture model is proposed by independent component analysis of oil group(octane number prediction),the established regression model of support vector machine can be widely used in industrial grade of octane value prediction in production practice.Based on remote sensing,precise detection and analysis of ocean oil spill is important for detecting marine oil spil.The working method and working mode proposed in paper has positive significance.for oil spill and for the application of remote sensing of oil spill Although this paper is focused on the research of oil spill detection method for formation of the ocean,but it can be effectively transplanted to the detection and identification of petroleum industry,food industry and other industries.