辅以纹理特征的HJ-CCD海上溢油信息提取——以PL19-3溢油为例

Oil spill information extraction combined with texture features from HJ-CCD Sensors-A case study in PL19-3 oil spill incident

鉴于依赖光谱特征的传统溢油信息提取方法面临信息提取精度低的困境,提出采用光谱特征与纹理分析结合的方法应用于溢油监测.选择位于渤海的蓬莱19-3油田溢油事故为研究对象,基于覆盖溢油事故阶段的8景30m分辨率HJ-CCD数据,在溢油目标提取过程中,引入了方向性纹理特征分析,将主成分光谱降维、方向梯度边缘检测等技术相结合,形成了基于光谱与纹理特征的溢油信息提取技术.所述方法经8组数据检验后,用类间分歧度方法进行了对比评价.结果表明:将纹理分析方法应用于溢油信息提取,类间分歧度提高到1.9999,提高了油膜影响边界和油膜厚度分区识别能力.

Traditional information extraction technique of optical satellite remote sensing constitutes an important component of oil spills monitoring system,but it subjects to monitoring accuracy and ability dependence on spectral features.Based on CCD（30m spatial resolution） data from operational HJ instruments,we taking Penglai 19-3 oil spill incident was took as an example to discuss the method of combining spectrum with directional textural information to improve the accuracy of extracted information.A principal components-based algorithm first extracted all spectrum information of oil-on-water.Then a directional gradient algorithm acquired the edge distribution of oil-contaminated area.Finally,the proposal method were tested with 8 scenes of HJ-CCD data and compared with conventional method based on singe spectrum using Jeffries-Matusita separability index.The results show that the introduction of the directional texture analysis is effective to the edge detection of contaminated zone and the identification of thick-thin oil distribution,which is feasible in the oil spill monitoring based on HJ-CCD data.