Discrimination of Oil Slicks and Lookalikes in Polarimetric SAR Images Using CNN.

Oil slicks and lookalikes (e.g., plant oil and oil emulsion) all appear as dark areas in polarimetric Synthetic Aperture Radar (SAR) images and are highly heterogeneous, so it is very difficult to use a single feature that can allow classification of dark objects in polarimetric SAR images as oil slicks or lookalikes. We established multi-feature fusion to support the discrimination of oil slicks and lookalikes. In the paper, simple discrimination analysis is used to rationalize a preferred features subset. The features analyzed include entropy, alpha, and Single-bounce Eigenvalue Relative Difference (SERD) in the C-band polarimetric mode. We also propose a novel SAR image discrimination method for oil slicks and lookalikes based on Convolutional Neural Network (CNN). The regions of interest are selected as the training and testing samples for CNN on the three kinds of polarimetric feature images. The proposed method is applied to a training data set of 5400 samples, including 1800 crude oil, 1800 plant oil, and 1800 oil emulsion samples. In the end, the effectiveness of the method is demonstrated through the analysis of some experimental results. The classification accuracy obtained using 900 samples of test data is 91.33%. It is here observed that the proposed method not only can accurately identify the dark spots on SAR images but also verify the ability of the proposed algorithm to classify unstructured features.

A New Approach of Oil Spill Detection Using Time-Resolved LIF Combined with Parallel Factors Analysis for Laser Remote Sensing.

In hope of developing a method for oil spill detection in laser remote sensing, a series of refined and crude oil samples were investigated using time-resolved fluorescence in conjunction with parallel factors analysis (PARAFAC). The time resolved emission spectra of those investigated samples were taken by a laser remote sensing system on a laboratory basis with a detection distance of 5 m. Based on the intensity-normalized spectra, both refined and crude oil samples were well classified without overlapping, by the approach of PARAFAC with four parallel factors. Principle component analysis (PCA) has also been operated as a comparison. It turned out that PCA operated well in classification of broad oil type categories, but with severe overlapping among the crude oil samples from different oil wells. Apart from the high correct identification rate, PARAFAC has also real-time capabilities, which is an obvious advantage especially in field applications. The obtained results suggested that the approach of time-resolved fluorescence combined with PARAFAC would be potentially applicable in oil spill field detection and identification.

[Orthogonal Vector Projection Algorithm for Spectral Unmixing].

Spectrum unmixing is an important part of hyperspectral technologies, which is essential for material quantity analysis in hyperspectral imagery. Most linear unmixing algorithms require computations of matrix multiplication and matrix inversion or matrix determination. These are difficult for programming, especially hard for realization on hardware. At the same time, the computation costs of the algorithms increase significantly as the number of endmembers grows. Here, based on the traditional algorithm Orthogonal Subspace Projection, a new method called. Orthogonal Vector Projection is prompted using orthogonal principle. It simplifies this process by avoiding matrix multiplication and inversion. It firstly computes the final orthogonal vector via Gram-Schmidt process for each endmember spectrum. And then, these orthogonal vectors are used as projection vector for the pixel signature. The unconstrained abundance can be obtained directly by projecting the signature to the projection vectors, and computing the ratio of projected vector length and orthogonal vector length. Compared to the Orthogonal Subspace Projection and Least Squares Error algorithms, this method does not need matrix inversion, which is much computation costing and hard to implement on hardware. It just completes the orthogonalization process by repeated vector operations, easy for application on both parallel computation and hardware. The reasonability of the algorithm is proved by its relationship with Orthogonal Sub-space Projection and Least Squares Error algorithms. And its computational complexity is also compared with the other two algorithms', which is the lowest one. At last, the experimental results on synthetic image and real image are also provided, giving another evidence for effectiveness of the method.

[Characterization of Time-Resolved Laser-Induced Fluorescence from Crude Oil Samples].

To evaluate the feasibility of laser induced time-resolved fluorescence technique for in-situ detection of underwater suspended oil spill, extensive investigations have been carried out with different densities of crude oil samples from six different wells of Shengli Oilfield in this work. It was found that the fluorescence emission durations of these crude oil samples were almost the same, the Gate Pulse Delay of DDG (Digital Delay Generator) in the ICCD started at 52ns and ended at 82ns with a width (FWHM) of 10 ns. It appears that the peak location and lifetime of fluorescence for different crude oil samples varied with their densities, and those with similar densities shared a similar lifespan with the closer peak locations of fluorescence. It is also observed that the peak of fluorescence remained the same location before reaching the maximum intensity, subsequently shift to longer wavelength as fluorescence attenuated from maximum intensity with a red shift among 17-30 nm varied with samples. This demonstrated that the decay rate of fluorescent components in the crude oils was different, and energy transfer between these components might exist. It is hoped that those obtained results and characteristics could be the useful information for identification of suspended spilled-oil underwater.

[Effective endmembers based bilinear unmixing model].

An effective endmembers based bilinear unmixing algorithm is prompted in the present paper together with an end-member subset selection algorithm as well. Firstly, the endmembers are ranked according to their distance to the mixed pixel, involving the Euclidean distance and spectral angle. And then, an effective subset of the endmembers is abstracted considering both the ranking result and the change of error. The algorithm reduces the influence of endmembers which are not component of the mixed pixel, decrease the number of endmembers involved in unmixing and improve the accuracy of abundance. The test results for simulation image prove that the algorithm would provide a lower reconstructing error. And the analysis results of actual airborne hyperspectral oil spill image further illustrate its effectiveness.

[Study of prediction models for oil thickness based on spectral curve].

Nowdays, oil spill accidents on sea occur frequently. It is a practical topic to estimate the amount of spilled oil, which is helpful for the subsequent processing and loss assessment. With the rapid development of hyperspectral remote sensing technology, estimating the oil thickness becomes possible. Firstly, a series of oil thicknesses are tested with the AvaSpec Spectrometer to get their corresponding spectral curves. And then the characteristics of the spectral curve are extracted to analyze their relationship with the oil thickness. The study shows that the oil thickness has large correlation with variables based on hyperspectral positions such as R(g), R(o), and vegetation indexes such as RDVI, TVI and Haboudane. Curve fitting, BP neural network and SVD iteration method were chosen to build the prediction models for oil thicknesses. Finally, the analysis and evaluation of each estimating model are provided.