Marine oil slicks show brighter or darker than surrounding oil-free seawater under different sunglint, which can be observed by satellite optical sensors. Although this has been interpreted using a critical angle concept and simulated using the Cox-Munk model, it has not been demonstrated in high spatial resolution images from airborne sensors. In this study, an AISA (airborne imaging spectrometer for applications) image was used to determine the characteristics of non-emulsion oil slicks under sunglint in high spatial resolution images. Although a similar positive or negative contrast between non-emulsion oil slicks and oil-free seawater can be observed, it is difficult to directly model sunglint reflectance due to the different remote sensing scale effect. There are many sun glitter pepper noise produced by various micro-mirror facets of ocean surface in high spatial resolution images. Based on the optical image characteristics, a normalized noise index (ξ) was designed to evaluate the pepper noise in 1830 band-difference results. Then a level segmentation method was used to delineate the oil slicks under various sunglint from a minimum pepper noise image. Our study provides a preliminary reference for airborne optical remote sensing of oil slicks under various levels of sunglint.
© 2018 Optical Society of America under the terms of the OSA Open Access Publishing Agreement
Marine non-emulsion oil slicks result from natural seepage or oil spill accidents [1–3]. Remote detection of these oil slicks can be accomplished by satellite Synthetic Aperture Radar (SAR) and optical sensors [4–7]. An oil slick’s modulation of surface roughness reduces the backscattering signal, causing it appears darker than the oil-free seawater in an SAR image . In addition, oil slicks and seawater have different refractive index , and their difference in surface roughness and refractive index also result in oil slicks showing bright and dark features under certain viewing geometry [9–13].
On the ocean surface, the specular and near-specular reflection of direct sunlight is termed sunglint, and the angle (θm) between the viewing direction and the direction of specular reflection is often used to indicate the strength of sunglint reflectance (LGN) [9,14]. The actual ocean surface can be conceptualized as containing numerous “facets”, the reflection of sunlight by which follows Snell’s Law . The Cox-Munk model, a type of probability density function, depends on the statistical surface roughness and refractive index and has been used to calculate an oil slick’s LGN in coarse spatial resolution images [10,13,15]. The difference in surface roughness and refractive index between oil slicks and the surrounding oil-free seawater can cause a change in contrast between them from negative to positive, which varies with the strength of sunglint, from weak to strong [8–10]. The critical angle is often where the contrast changes from negative to positive under sunglint, which exhibits no or negligible contrast between oil slicks and the non-oiled seawater [8–10].
The critical angle concept and the Cox-Munk model can be used effectively to interpret and simulate oil-seawater contrast reversal under sunglint in multi-band coarse-resolution optical sensors [15–17]. However, it is unclear how an oil slick appears under sunglint when view through airborne optical sensors which often have often a high spatial resolution and wide field of view angle. Two important questions are: Is it possible to determine the critical angle between oil slicks and surrounding seawater under sunglint using high spatial resolution airborne imagery? Can we enhance the contrast of oil slicks and non-oiled seawater to improve the identification of oil slicks?
In this study, the airborne imaging spectrometer for applications (AISA) onboard Y-12 aircraft of the State Oceanic Administration (SOA) measured the radiance of spilled oil slicks and non-oiled seawater under sunglint around an oil platform in the coastal zone near the estuary of the Pearl River in China. The contrast reversal and critical angle between oil slicks and non-oiled seawater under sunglint can be determined in airborne optical images. Spectral difference is recommended to reduce the impact of sunglint on oil slicks or the surrounding non-oiled seawater. To choose the effective bands, a normalized noise index has been employed to assess the AISA band-difference to enhance their contrasts. This approach potentially provides an initial step in providing a solution to the problem of detecting oil slicks using airborne optical image under sunglint.
2. Study area and AISA imagery
The AISA onboard Y-12 aircraft of the SOA is often used for marine environmental remote monitoring in China. On 14 October 2005, oil spilled from a Chinese platform (114°54′41″ E, 21°22′46″ N) and spread over the ocean surface to form surrounding oil slicks (Fig. 1).
The platform located in the basin at the mouth of the Pearl River; it is an extensional basin on the passive continental margin of the northern South China Sea. The AISA was installed onboard the aircraft to observe the oil slicks on the same day (UTC time: 02:50). We obtained the spatial resampling data from SOA; it covered 400-970 nm, and the average spectral resolution of 61 bands was about 9.2 nm, and the spatial resolution was 5 m. The altitude of the aircraft was ~700 m, and the scan width and imaging length along the track line was ~430 m and ~2100 m, respectively. Atmospheric effect can be ignored at this low altitude and thus atmospheric correction was not performed in this study. Reflectance of 789 nm, 675 nm and 552 nm bands was used as the Red, Green, and Blue channels to compose the RGB image. The oil slicks under different levels of sunglint were not clearly resolved, even with the RGB image enhancement. For the high spatial resolution image, the essence of the contrast phenomenon between oil slicks and surrounding non-oiled seawater under sunglint was the probability of occurrence of dark or bright pixels.
3. Theoretical background
Classical statistical models, such as the Cox-Munk model (Eq. (1)), can be used to calculate the ocean surface sunglint reflectance (LGN) using images with coarse spatial resolution (i.e. MODIS and MERIS) . One component of this model is a probability distribution function (PDF) which can be derived from sea surface roughness and the viewing angles (Eq. (2) and Eq. (3)).
Although modeled LGN can be in accord with the observations collected from the coarse spatial-resolution image, it is difficult to simulate the sunglint reflectance of a high spatial resolution image due to the remote sensing scale effect (Fig. 2). In this study, the key step for detecting oil slicks was not to calculate sunglint reflectance accurately, but rather to eliminate the impact of sunglint and thus enhance the contrast between oil slicks and the surrounding non-oiled seawater. Bright oil slicks against oil-free seawater were observed close to the platform, and dark oil slicks were observed further away from the platform (Fig. 1). This indicated that contrast reversal between the oil slicks and the surrounding non-oiled seawater under sunglint can be also detected in airborne optical imagery.
Band-difference algorithms give higher accuracy and less errors than band-ratio algorithms in ocean color optical remote sensing . Thus, it was employed to enhance the contrast between oil slicks and the surrounding oil-free seawater in this study. There is a total of 61 bands of AISA data resulting in 1830 combinations (C(61, 2) = 1830); however, a key issue was how to select the optimal band difference. Here, we used the Normalized noise index (ξ) to evaluate the random noise between oil slicks and surrounding oil-free seawater of the 1830 combinations of band-difference images. This parameter (ξ) can be calculated as:Fig. 3). The lower the value of ξ, the greater the stable brightness contrast between oil slicks and seawater.
The morphology of oil slicks is a distinctive characteristic of band-difference results, and a level set segmentation method is used to detect them. The principal idea is to obtain contours of a zero level set of an implicit function defined in a higher dimension, and then to develop the level set function according to a partial differential equation. In the extraction of oil slicks, active contours are dynamic curves that approach the boundaries of the area of oil slick distribution . The edge indicator function is defined by
4. Results and discussion
4.1 Brightness contrast reversal under sunglint
It is difficult to display simultaneously the contrast reversal between oil slicks and the surrounding oil-free seawater under sunglint. Two different bands (443 nm and 885 nm) are used to show these various contrasts. The gray scale image of the 443 nm band has been stretched to illustrate that the oil slick is brighter than the surrounding oil-free seawater under strong sunglint, and the 885 nm image has been stretched to illustrate that the oil slick is darker than the background seawater under weak sunglint (Fig. 4(a) and 4(b)). As the contrast reversal occurs under various sunglint, a critical angle line will be defined between the dark and bright oil slicks which is parallel to the flight track. The positive or negative contrast between the spectra of non-oiled seawater and bright oil slicks, or of seawater and dark oil slicks, had been determined Fig. 4(c) and Fig. 4(d).
4.2 Refining band-difference
The obvious contrast between oil slicks and oil-free seawater implies that band difference is a simple and practical method for eliminating sunglint and enhancing contrast. 61 bands of the AISA data generate 1830 combinations of band-difference. To choose the two optimal bands, a triangular matrix of ξ is established to evaluate the band-difference images (Fig. 5). A transect in negative contrast (Fig. 5, inset) is used to calculate the value of ξ for each combination. ξ ranges from 0.06 to 0.7, with low ξ indicating that there is less pepper noise in the band-difference image. In other words, oil slicks are more clearly resolved. Four combinations of band difference, with ξ = 0.1, ξ = 0.3, ξ = 0.5, and ξ = 0.7 (indicated by the numbers I-IV in Fig. 5) can be used to illustrate the characteristics of the images (Fig. 6).
Oil slicks in four band-difference images with ξ = 0.1, ξ = 0.3, ξ = 0.5, and ξ = 0.7 indicate that the oil slick is clearly resolved in the image with low ξ, and is difficult to detect in the image with high ξ due to the latter contains more pepper noise (Fig. 6). The images (ξ = 0.1, ξ = 0.3, ξ = 0.5, and ξ = 0.7) are obtained from the respective band differences between band 29 (656 nm) and band 12 (498 nm), band 48 (837 nm) and band 27 (637 nm), band 50 (856 nm) and band 37 (731 nm), band 50 (507 nm) and band 12 (417 nm).
4.3 Identification of marine oil slicks
All of the bright and dark oil slicks in the images under various levels of sunglint exhibit a negative contrast in the band-difference images. According to the minimum value of ξ in Fig. 5 and 6, the optimal band-difference between the 498 nm and 609 nm band for oil slick identification can be specified (Fig. 6(a)). This provides the clearest brightness contrast for differentiating oil slicks from seawater under various levels of sunglint. In this study, the level segmentation method (which considers changes in the property of local grayness) is employed to delineate the oil slicks. After more than 2000 iterations, marine oil slicks is finally delineated. Sea surface wind speed was about 2-3 ms−1, blowing from NE to SW (https://manati.star.nesdis.noaa.gov/), so that the oil slicks drifted and spread in a southwesterly direction from the platform. The length of oil slicks is about 1.74 km and the total area is about 0.087 km2 (Fig. 7).
Marine oil slicks have been observed in images obtained by satellite coarse resolution optical sensors in which contrast reversal under various levels of sunglint, and sunglint reflectance can be calculated using a statistical model. The oil slicks spilled from a platform near to the estuary of the Pearl River in China were observed using airborne optical sensors with a high resolution (AISA). Although it is difficult to identify contrast reversal or the critical angle under various levels of sunglint in the RGB composite image of AISA, the enhanced single band shows dark or bright oil slicks. This demonstrate that the visual characteristics of oil slicks observed through airborne sensors are consistent with those of satellite sensors. Sunglint reflection has an uncertain effect on the identification and spatial quantification of oil slicks using high-resolution airborne images due to scale effect. In this study, the band-difference method is used to enhance the contrast between dark or bright oil slicks and background seawater, under various levels of sunglint, to facilitate the identification of oil slicks. The normalized noise index is designed to effectively evaluate the noise level of the oil-seawater contrast to enable us to select the optimal image from 1830 difference images. Subsequently, dark and bright oil slicks can be delineated accurately using a type of identification algorithm. These results show that the contrast reversal of oil slicks can be observed in airborne images, and that it is an efficient method for identifying dark and bright oil slicks under various levels of sunglint using the contrast enhancement of optimal band-difference and an evaluation parameter (the Normalized noise index). Thus, this approach provides a preliminary reference for airborne optical remote sensing of oil slicks under sunglint.
National Natural Science Foundation of China (Grant Nos. 41771376 and 61675099), the National Key Research and Development Program of China (Grant Nos. 2016YFB0501502, 2016YFC1400901).
The authors thank State Oceanic Administration of China for providing AISA image.
1. K. Kvenvolden and C. Cooper, “Natural seepage of crude oil into the marine environment,” Geo-Mar. Lett. 23(3), 140–146 (2003). [CrossRef]
2. I. R. Macdonald, N. L. Guinasso Jr., S. G. Ackleson, J. F. Amos, R. Duckworth, R. Sassen, and J. M. Brooks, “Natural oil slicks in the Gulf of Mexico visible from space,” J. Geophys. Res. 98(C9), 16351–16364 (1993). [CrossRef]
3. Y. Lu, X. Li, Q. Tian, G. Zheng, S. Sun, Y. Liu, and Q. Yang, “Progress in Marine Oil Spill Optical Remote Sensing: Detected Targets, Spectral Response Characteristics, and Theories,” Mar. Geod. 36(3), 334–346 (2013). [CrossRef]
4. B. Zhang, W. Perrie, X. Li, and W. G. Pichel, “Mapping sea surface oil slicks using radarsat-2 quad-polarization sar image,” Geophys. Res. Lett. 38(10), 415–421 (2011). [CrossRef]
5. J. Zhao, M. Temimi, H. Ghedira, and C. Hu, “Exploring the potential of optical remote sensing for oil spill detection in shallow coastal waters--a case study in the Arabian Gulf,” Opt. Express 22(11), 13755–13772 (2014). [CrossRef] [PubMed]
6. I. Leifer, W. J. Lehr, D. S. Beatty, E. Bradley, R. Clark, P. Dennison, Y. Hu, S. Matheson, C. E. Jones, B. Holt, M. Reif, D. A. Roberts, J. Svejkovsky, G. Swayze, and J. Wozencraft, “State of the art satellite and airborne marine oil spill remote sensing: Application to the BP Deepwater Horizon oil spill,” Remote Sens. Environ. 124, 185–209 (2012). [CrossRef]
7. A. Pisano, F. Bignami, and R. Santoleri, “Oil spill detection in glint-contaminated near-infrared MODIS imagery,” Remote Sens. 7(1), 1112–1134 (2015). [CrossRef]
8. Y. Lu, S. Sun, M. W. Zhang, B. Murch, and C. Hu, “Refinement of the critical angle calculation for the contrast of oil slicks under sunglint,” J. Geophys. Res. 121(1), 148–161 (2016). [CrossRef]
9. C. Hu, X. Li, W. G. Pichel, and F. E. Muller-Karger, “Detection of natural oil slicks in the NW Gulf of Mexico using MODIS imagery,” Geophys. Res. Lett. 36(1), 58–69 (2009). [CrossRef]
10. C. Jackson and W. Alpers, “The role of the critical angle in brightness reversals on sunglint images of the sea surface,” J. Geophys. Res. 115(C9), C09019 (2010). [CrossRef]
11. T. Kuser, E. Vahtmäe, and J. Praks, “A sun glint correction method for hyperspectral imagery containing areas with non-negligible water leaving NIR singal,” Remote Sens. Environ. 113(10), 2267–2274 (2009). [CrossRef]
13. C. Cox and W. Munk, “Measurement of the roughness of the sea surface from photographs of the Sun’s glitter,” J. Opt. Soc. Am. 44(11), 838–850 (1954). [CrossRef]
14. Y. Wen, M. Wang, Y. Lu, S. Sun, M. Zhang, Z. Mao, J. Shi, and Y. Liu, “An alternative approach to determine critical angle of contrast reversal and surface roughness of oil slicks under sunglint,” Int. J. Digit. Earth 11(9), 972–979 (2018). [CrossRef]
15. H. Zhang and M. Wang, “Evaluation of sun glint models using MODIS measurements,” J. Quant. Spectrosc. Ra. 111(3), 492–506 (2010). [CrossRef]
16. S. Sun and C. Hu, “Sun glint requirement for the remote detection of surface oil film,” Geophys. Res. Lett. 43(1), 309–316 (2016). [CrossRef]
17. H. Zhang, K. Yang, X. Lou, Y. Li, G. Zheng, J. Wang, X. Wang, L. Ren, D. Li, and A. Shi, “Observation of sea surface roughness at a pixel scale using multi-angle sun glitter images acquired by the ASTER sensor,” Remote Sens. Environ. 208, 97–108 (2018). [CrossRef]
18. C. Hu, Z. Lee, and B. Franz, “Chlorophyll a algorithms for oligotrophic oceans: A novel approach based on three-band reflectance difference,” J. Geophys. Res. 117(C1), 7395 (2012). [CrossRef]
19. C. Li, C. Xu, C. Gui, and M. Fox, “Level set evolution without re-initialization: a new variational formulation,” In Computer Vision and Pattern Recognition, 2005. CVPR 2005. IEEE Computer Society Conference on (Vol. 1, pp. 430–436).