海上搜寻目标光谱探测追踪研究

Remote sensing technique provides a way for maritime search and rescue in the recent frequent crash. While so far people mostly make use of the remote sensing high spatial resolution images to widely and timely interpret and discriminate crashing targets during maritime searching and rescue.We will make use of the spectral character of Remote Sensing to identify these drifting targets. In order to build the maritime searching targets' spectral library, in this study we will measure the spectra of different typical searching targets, especially for the jet fuel. The remote sensing spectral characters and the stable isotope model for different stages of weathering oil will be built, which is convenient for estimating the weathering stage and the lasting time. In this way, we can make accurate assessment for the maritime search and rescue. Meanwhile, with the help of the hyper-spectral Remote sensing, the identification algorithm of wreckage and jet fuel and efficient spectral un-mixing idea will be put forward, so that we can build the identification model trough the spectral library and the hyperspecteal Remote Sensing to achieve the research's objective for maritime searching targets spectral detection and tracking. Furthermore, this study will use the stable isotope technique wich is based on the atom level to track the origin of the jet fuel to ensure the reliability of the detection results. We should take advantages of hyper-spectral remote sensing in order to break through the difficulty of quickly and exactly identifying maritime searching targets.

遥感为多发的海难事件海上搜寻目标提供一种方法。但目前利用遥感技术进行海上目标搜寻多应用于多时相高分辨率遥感影像进行大范围、目视判别和解译可疑目标的方法。本课题利用海上搜寻目标的实测光谱，建立海上搜寻目标光谱数据库，特别针对海上搜寻失事飞机泄露的航空煤油带，建立碳稳定同位素组成及遥感光谱特征的航空煤油风化规律模型。通过不同风化时间航空煤油的光谱特性及其油指纹的变化特征，反演油膜所经历的风化过程，推断油膜在海面存在时间，为海上搜寻提供可靠时间依据。通过对搜寻目标的端元提取、光谱分解及光谱匹配算法的优化，建立实测光谱库与高光谱遥感目标特征的探测模型，实现海上搜寻目标光谱的快速自动探测。再利用稳定同位素技术对油膜特征进行原子水平的油指纹信息模型构建，实现目标的鉴别与溯源，保证遥感海上搜寻目标的可靠性，协同完成精确的海上搜寻目标探测寻踪。以实现利用遥感光谱分辨率特征优势进行海上目标探测溯源的突破。

为探究海面常见船舶用油与失事飞机泄漏航空煤油的光谱特征细微差异，利用地物光谱仪，模拟一定的海面深水环境进行了航空煤油、轻柴油、重柴油及海水水体光谱特征的定量测量，并对测量结果进行光谱特征定性分析；利用遥感高光谱影像光谱特征及实测光谱库，对不同风化阶段的重要海上目标航空煤油，构建遥感光谱及稳定同位素的风化规律模型，以解决油膜在海上经历风化的物理化学变化所引起的探测难题；结合空间信息的视觉显著性及光谱信息的光谱相似性显著性，进行遥感影像显著性分析，实现待探测海上搜寻目标感兴趣区自动提取研究，将海量遥感影像缩小范围局部定位，大幅度提高海上搜救事件救援效率，弥补传统人工目视寻找失事目标的缺陷；利用稳定同位素技术对油膜特征进行原子水平的油指纹信息模型构建，实现了目标的鉴别与溯源，保证遥感技术对海上目标搜寻的可靠性，协同完成精确的海上搜寻目标探测溯源；利用高光谱遥感进行海上失事油膜提取研究. 提出的DSPP端元提取预处理算法提取端元精度较高，抗噪能力较强，消耗时间最短,可以满足应用于海上搜寻遥感探测的实际应用需要；根据油的光谱特征，大体估算油品所处的风化过程，进而推断油膜在海面存在的时间范围。同时，通过分析不同阶段的油膜光谱特征及其与水体光谱特征差异，可对油膜的可识别性进行评价；进行了NDWI及形态学开运算的水陆分离，利用视觉显著性的GBVS模型将提取的海面影像快速定位海上船舶目标，最终利用目标形状特征的水平集活动轮廓模型及纹理特征的马尔可夫（Markov random field, MRF）纹理模型进行影像分割，实现船舶目标的精确提取；对海上事故中漂浮于海面的飞机、船舶残骸等小目标地物进行高光谱探测研究，用于广阔海域海上事故中像元级乃至亚像元级残骸的搜寻探测