Software Development Of Multi-channel Airborne Laser Fluorescence System For Marine Oil Spill Detection

As the oil consumption with the development of national economy keeps increasing, oil pollution accompanied by exploration and transportation results in huge financial losses and makes a significant influence on marine ecological environment. Airborne laser fluorescence system, as one of the most effective and potential tool for oil detection, could provide oil fluorescence spectral data with high spatial resolution and high accuracy. The information including oil type and oil film thickness could be derived from the spectral data, and all these information can help the related department to make good decision about emergency treatment of oil spill.For monitoring oil pollution on sea surface, ORSI-OUC(the Ocean Remote Sensing Institute of Ocean University of China) started to develop an airborne LIDAR system from 2006 under the national project of "Research on marine oil detection technique by airborne fluorescence LIDAR system with multiple channels". To meet the needs of real-time oil detection in the ocean, the thesis describes the development of application software of laser fluorescence system for oil detection based on Visual C++. By analysis of software architecture, functional model partitioning and with some mature software technologies, the period of developing software is greatly shortened and the software quality is improved.We use T-Clock (TClk) technique, developed by National Instruments Company (NI), to synchronize several digitizer boards during data sampling, and we use the Windows multi-threaded programming to make the system working in real time. The techniques of ActiveX control and the hybrid programming of Matlab and Visual C++on the basis of Component Object Model (COM) were used to realize the data display, storage and analysis. The result shows that the software works well for oil detection, and it could be used in other similar software development.The thesis also studied the algorithm of oil fluorescence spectral recognition. The recognition process includes two steps. In the first step, the algorithm of Spectral Information Divergence-Spectral Angle Mapper(SID-SAM) is used to roughly recognize the light oil, medium oil and lubricant oil from others; In the second step,'Log_Relief_Lssvm'model is used to further classify between the crude and heavy oil. In the 'Log_Relief_Lssvm' model, the input data is normalized, standardized and calculated in log-scale firstly, and then eleven input features were selected to be the input of the classifier---Least Squares Support Vector Machine (LSSVM).The recognition model is selected by the Partial Swarm Model Selection (PSMS) algorithm based on training data sets and recognition algorithm library. The test results showed the selected model is working well.