Instant identification of hydrocarbon spill types using laser-induced fluorescence and associated hyperspectral imaging

Marine hydrocarbon (HC) spills seriously harm the marine ecosystem and necessitate the creation of effective cleanup strategies. For efficient treatment, it is crucial to accurately identify the type of HC implicated in the spill. Diesel oil, gasoline, lubrication oil, and waste oil were chosen as the experimental starting materials since they are prevalent HC kinds. This study reports on the facile differentiation of diesel fuel and gasoline to engine oil via induced fluorescence by a laser. The tested material was stimulated with a 450-nm blue laser source. The fluorescence characteristic of each tested HC spill sample was captured via a hyperspectral imaging camera. Diesel fuel could be identified via their unique fluorescence signatures at 470 nm and 515 nm. Gasoline was found to be a less photoluminescent active material, with characteristic signals at 465 nm and 585 nm. At 505 nm, lubricating oil had the typical fluorescence signal. On the whole study spectrum, there are no re-emitted frequencies for the waste oil. Fluorescence signals secured good discrimination between fuel and engine oil of HC chemical structures. Signal intensity can contribute further discrimination. Fluorescence intensity at selected photoluminescence wavelengths secured further grouping between fuel and oil. Our imaging approach would provide investigators with the ability to define the HC and choose a mitigation strategy for fast HC spill identification. Aerial systems with a customized common camera that only functions at a certain wavelength filter with our methodology based on laser-induced fluorescence may be effectively developed for HC spill detecting applications.


Introduction
Due to its ability to provide more than 40% of the world's energy demands, crude (cr.) oil rose to the top of the global mineral demand throughout these last decades (Holechek et al. 2022). As a result, despite the quantity of oil lost due to spills or well discharges during the exporting phase, the globe had raised the exporting pace in order to collect more oil. Liquid petroleum hydrocarbons (HC), which may leak 4.5 million tonnes of oil were released into the oceans or seas, cause oil spills when they are released into the natural environment (Nkem et al. 2022;Kalia et al. 2022).
HC spill pollution can result in a number of natural disasters, including the inability of enough sunlight to reach the ocean's surface, a decrease in the oxygen's ability to dissolve, and an increase in the threat of extinction for various animal species in a maritime area because their long-term recuperation durations might be longer than usual and their reproduction rates could be minimal (Qixing et al. 2022;Zhu et al. 2022). In furthermore, the damage it does to aquatic biological life and the effects related to the enormous oil loss, which causes a huge economic downturn (Silva et al. 2022;Thakur and Koul 2022).
Real calamities have in fact occurred in recent decades as a result of HC spills. A significant leak resulted from the Deepwater Horizon oil blast drilling platform in 2010. Millions of barrels of oil contaminated the Gulf of Mexico, covering 10,000 km 2 or so (Li et al. 2020;Jolli et al. 2019). The marine ecosystems, including fish and seabirds, have been severely damaged as a result of this catastrophe. A severe HC spill event involving the Penglai 19-3 oil field in Bohai Bay, Northeast China, occurred in 2011 when more than 7000 gallons of oil leaked, filling the ocean with tons of oil (Yang et al. 2019). There were over 6200 km 2 of contaminated land. Another occurrence is the SANCHI tanker, which in 2018 clashed with a tanker in the East China Sea while transporting roughly 136,000 tons of condensate oil, exploded, and subsequently sunk (Bhangale et al. 2017). The oil slick might be carried onto the coast by the sea waves if the oil was not promptly monitored after the release. The health of individuals and coastal aquaculture fish supplies would be seriously threatened by this condition. Therefore, it is crucial to accurately identify HC spills on the water's surface in order to track their dispersion and effects (Cally 2018).
The mixes of crude oil and petroleum products, which are mostly hydrocarbon molecules, are very complicated (Vetere et al. 2021;Badamshin et al. 2021). Petroleum products' intricate composition necessitates the use of the most tactful methods conceivable. Moreover, a correct classification of HC spill types can aid in identifying the leak's origin, estimating possible damage, and choosing an appropriate course of action. Our research focused on spills of fuels or motor oil, which are more frequently generated by ship emissions or crashes. A combination of liquid aliphatic and aromatic HC with carbon atom counts between C 4 and C 12 and boiling temperatures between 86 and 437°F make up gasoline which is the first to be studied (Vempatapu and Kanaujia 2017). Diesel, the typical liquid fuel for stationary gas engines, is the second element. It is a byproduct of cr. oil that is distilled between 200 and 350 °C. It contains around 25% aromatic HC and roughly 75% aliphatic HC (C 10 H 20 -C 15 H 28 ) (Jansohn 2013). The chosen diesel for our investigation was with sulfur limit 5000 ppm. Lubricating (lub.) oils are made up of 10-20% additives to provide the oil certain qualities and 80-90% petroleum HC distillate for lubrication (Wexler 2005). Waste lub. oil is a dangerous substance because it contains impurities like metals and aromatic HC that are created when a motor burns fuel. When it is intentionally disposed of or accidently spills into the environment, it becomes an environmental concern, thus it will finally be included in this research.
These selected sorts of HC spills that occur at marine surface include traceable damage and this necessity the need to the creation of an HC spill spotless strategy. For the spill's pollutants to be treated quickly and effectively, these spill types involved in it must be accurately identified. The current standards for identifying different types of oil are mostly based on laboratory chemical identification techniques like mass spectrometry and chromatography (Bayona et al. 2015;Texeira et al. 2014), which can analyze different types and components of oil with accuracy but have low-rates detection speeds. The determination of oil type via remote sensing spectral analysis has gained popularity recently. Synthetic Aperture Radar (SAR), which offers the benefit of daylong imagery and is now thought of as a traditional method for remote sensing for spill spotting, is not constrained by weather conditions. The drawbacks of this method are that oil type identification cannot be done with SAR (Alpers et al. 2017;Marghany 2014). More information about HC spills is provided by optical sensors than by SAR signals, and the reaction to HC spills has long been a source of regard (Pisano et al. 2015;Lu et al. 2016). HC films' spectral properties are connected to their chemical properties and molecular vibrations. The absorption spectrum properties of different oil kinds' HC bonds are a significant basis for differentiating oil variants (Lu et al. 2019). But since only standard optical remotely sensed data includes just few bands, spectral absorption properties can easily be lost. The water level has a poor reflectivity due to Rayleigh scattering of the atmosphere, and the usable data from water only make up a small portion of the overall signal. Recovering information about the oil covering from the water's surface becomes more challenging if the spectral absorption information is lost (Kokaly et al. 2013). HC spill detection and type identification both benefit greatly from hyperspectral remote sensing. High spectral resolution and conspicuous detail information in hyperspectral data allow for accurate oil film separation from the marine background and HC identification based on specific oil film spectral signatures (Hu et al. 2010;Lu et al. 2013). The hyperspectral imaging (HI) technique, nevertheless, might be constrained to being ineffective at night without supporting light sources. Numerous types of materials have been studied using fluorescence spectroscopy over the last few decades and have shown advantages over other light-based approaches in regards to safety, quickness, and responsiveness. This efficient technique for analyzing oil samples uses laser-excited oil samples to produce a fluorescence signal that may be used to determine the kind of HC (Wigger et al. 2016;Fingas and Brown 2017;Hou et al. 2018). In Babichenko et al. (2016) and Baszanowska and Otremba (2014), fluorescence induced by laser light detection and ranging (LIDAR) technology was created to detect oil in open and coastal waters. The waters were raster scanned using a pulsed laser with a wavelength of 308 nm. As effective as the method was, utilizing a compact form to achieve rapid, greater detection was difficult.
For imaging the entire selected HC spills in our research trials, we used a special, simpler, and trustworthy approach that was low noise-resistant and dependable in terms of data interpretation. Laser-induced photoluminescence spectroscopy was adopted in an attempt to generate a novel fluorescence signature of HC spills due to ship leaks or incidents, which was stimulated via a blue laser source of the proper wavelength and power. In the meantime, measurements of the emission spectra of HC spills were precisely captured via a HI camera. The HI camera was chosen to capture fluorescence signatures for enhanced resolution, fast scan time, and quick information extraction in such a way that they are sufficiently informative to provide water surface discrimination in rapid optical screening of HC spills. A hyperspectral camera can provide spectral and spatial information over a wide spectrum band (350-1000 nm), including ultraviolet, visible, and near-infrared. It has a unique ability to collect vital spectral information (Aboughaleb et al. 2020; Aref et al. 2020). Our pilot study in this work is to get the chemical footprint for the four studied HC samples. The findings of our research revealed that the greatest change in the fluorescence signature for the studied HC spills occurred at one distinct common wavelength with the ability to make grouping between fuel and lub. oil. We were able to differentiate within each group based on the chemical footprint results obtained for each HC studied sample. Our noninvasive, non-destructive, and relatively compact imaging setup with the emission spectra from the HI camera supports current attempts to identify HC spills more objectively. With our method, specialists will be able to detect and identify spills more efficiently and precisely in a timely manner.

Theory
The process of emitting light physically from any material not heated in any way occurs in electrically excited states and is known as photoluminescence. This spectroscopy includes two main emission classes, including fluorescence and phosphorescence (Lakowicz 2006). Such emissions take place at a longer wavelength than the exciting beam (Harvey 2000). It is widely accepted that fluorescence is the emission between two identically spinning energy levels. Fluorescence lifetime is only 10 −5 : 10 −8 S (Jameson 2014). Thus, fluorescence emission often originates from the lowest-energy vibrational state that is thermally stable. Typically, a higher vibrational ground state is reached during decay to the ground state, which subsequently relaxes to a lower ground state and reaches thermal equilibrium. The emission transition could be also dispersed among several vibrational energy levels for a given excitation wavelength. A fluorescence emission spectrum may be measured for a single excitation wavelength. Phosphorescence is described as the emission of light between two energy levels that vary in their spin states. Different relaxation mechanisms, including radiationless ones, could occur via intermolecular transitions such as vibrational relaxation (vr), internal conversion (ic), external conversion (ec), and intersystem crossing (isc) (Skoog et al. 2018). Consequently, as illustrated in the "Jablonski diagram" in Fig. 1, phosphorescence happens at longer wavelengths and lower energy than the stimulating beam.
In the Jablonski diagram, the light absorbed, vibrational relaxation energy, intersystem crossover, and spontaneous low-intensity transition are all given the names "fluorescence," "phosphorescence," and "delayed fluorescence," respectively. The quantity of photons emitted as a percentage of photons absorbed is known as the photoluminescence quantum yield, or PhQY, of a molecule or substance. Understanding molecular activity and interactions for many crucial materials depends on this unique quality of a fluorescent molecule as illustrated in Eq. (1) (Harvey 2000).
where K fl is the fluorescence rate and K nr is the total of all radiation-free deexcitation routes' rates. When ratiationless deactivation process rates are slower than K fl , the PhQY is higher. However, PhQY is low if these other processes' rates are high in comparison to K fl .
HC spills exhibits autofluorescence, which is caused by the aromatic HC fraction of cr. oil, and the chemical composition (such as the concentrations of quencher and fluorophore) and physical properties (such as optical density and viscosity) of the oil (Teffens et al. 2011). Laser-induced fluorescence methodology is a non-linear phenomenon and is (1) PhQY = k fl k fl + k nr typically observed only at very high light intensities based on the molecular energy gap, which is unique for each HC spill material. Our starting point for this research was to get the emission spectra by observing the emission intensity across the relevant spectral range for the given energy of excitation.

Materials and methods
The methods and supplies used in our experimental study to identify and group the studied HC spills are covered in this section, along with how the various components were merged to produce the desired results. Our imaging system setup consists of four separate elements. The schematic diagram of our upcoming optical imaging experiment is shown in Fig. 2. It includes the oil samples inside a cuvette, as well as the HI camera (SOC710, Surface Optics Corporation), a 50 mW commercial blue laser with a beam diameter of 1.1 mm and a beam divergence of 0.6 m rad is used, and a laptop with its analysis software to extract nonlinear polarization density effect especially the fluorescent signal on the white reference settled all on the optical board bench. The wavelength dispersion component of the HI system is essentially a grating with a CCD detector. Based on wavelength, the spectrograph splits the diffused spectral reflected light beam into numerous directions while preserving its spatial information. After the dispersed light has been mapped onto the FPA detector, with one dimension containing the spectral data and the other the spatial data for the scanning line (520 pixels per row). The sensor is moved behind the lens on the utilized SOC710 camera. A complete hyperspectral

Sample preparation
In our study, most HC spills found abundantly in aquatic environments caused by ship collisions or ship emissions were chosen for this research work. We chose two types of ship fuel (diesel and gasoline) and two different samples of lub. oil. The two samples of oil divided to an unused sample and the other is a wasted sample to be studied and compared with respect to the normal lub. oil. The types of the selected four materials was available based on collecting from public ship supplies and added to four different cuvettes of 7 cm depth as shown in Fig. 3. All studied items got new and unused from the supplier except the waste oil was engine running for 5000 miles. All information and analysis regarding the classification of these studied oil samples can be found in the work's supplementary paragraphs. The HC spills database signature categorization could be utilized to develop a novel way for detecting different oils in a number of marine remote sensing applications.

HI optical arrangement for HC spill identification
We used a line-scanning HI camera with 520 pixels per line and gave typically 696 lines per cube to get hyperspectral cube image information for HC spill samples. The used camera encompasses the visible to near-infrared range (0.35-1.0 microns) and has the capability to be used in normal lighting circumstances. Our visible to NIR study used the SOC710 camera (Surface Optics Corporation, USA) with a resolution of nearly 5 nm spectrally. There were 128 frames in the scanned spectral cube image, each of which represented a different wavelength. The field of view of the used optical lens was 10° (Schneider Xenoplan, 35 mm focal length f/1.9 CCTV-LENS), which is suitable for a clear focus image between the HI camera and the studied sample. Before starting our experimental work, we should calibrate the HI camera to achieve the highest signal-to-noise ratio (S/N) output. A wideranging light source (a lamp with a wavelength spectrum range of 400-1000 nm, Derungs, 20P SX, 20 Watt) matched with the HI camera's spectral band is used for the calibration process. The background response is obtained from a spectral picture taken from a white reference sheet with a high reflectivity standard. The camera lens is totally covered with a non-reflective dark lid to get the black effect. The relative reflectance for the captured images is then calculated by these two obtained reference images using the formula below where I is the corrected spectral response captured image, I o is the raw spectral response captured image, I D is the dark captured image, and I B is the white captured image. When the tested specimen was triggered by a high energy steady blue laser source of light; it generated both nonlinear polarization effect, fluorescence, and phosphorescence signals depending on its molecular structure. The samples were lit by the HI camera at a distance of roughly 48 cm from the optical bench and the used laser source stimulated the samples at a distance of approximately 100 cm from the optical bench and about 90° off-axis from the HI camera as shown in Fig. 4a, b. The 10° FOV lens could capture a scene of dimensions 6 cm × 8 cm at this distance. At the focus line capture of the tested camera, the focus of the light source was in alignment to achieve a uniform distribution of light. During the experiment, the four investigated HC samples were irradiated by a blue laser at the same time to achieve also a uniform distribution of the light for the studied samples and assure the same optical paths. To prevent the analogue to digital (A/D) converter from becoming Fig. 3 The used oil samples for our HC detection study based on their fluorescence saturated, the laser intensity and detector integration time must be adjusted together. The nadir angle of camera output images was 0°, perpendicular to the optical bench. Data was collected, exposure was controlled, and the linear scanner's motors were operated by a computer running software (SOC710 operating software and HSAnalysis TM Data Analysis and Calibration Software). The SOC710 HI line scan camera integrated with a visible-NIR F/1.9 lens was initially centered on scanning the entire scene for each shot. The black and white references were recorded after the last line-scanning was completed. The data-capturing process for our four HC spill samples measured the intensity of the Fig. 4 a The benchtop HI system setup used for HC spill identification; b The studied HC samples re-emission after stimulated by the blue laser photons, or how it varies with wavelength. dataset from the used camera that measured 520 pixels per line, typically 696 lines per cube, 128 bands in the range of 0.35-1 micron, less than 40 microns spatial resolution, and about 4.69 nm spectral resolution. To increase the signal-to-noise ratio and reduce light overexposure, the integration time was adjusted to a nominal value 100 ms after adjusting the orientation of the laser light source on the objects. The captured scanning process with the white and black reference took about 30 s and nearly 10 s between each capture to return to the starting position. Blue laser engages with HC spill samples under investigation as the HI camera measures the emitted reflectance imagery to choose the appropriate fluorescent signal to determine initially whether we can recognize between the four studied samples.

Results and analysis
In this portion, we will discuss the profound achievements. Investigated HC spill samples were stimulated to the higher-energy state via 50 mW blue laser sourace of 450 nm wavelength. The stimulated material could relax to the ground state via the re-emission of flouresence signature. The emitted photons could correspond to all possible energy transitions of excited material. A SOC710 hyperspectral camera was employed to record and capture photoluminescence signature. When the tested samples were illuminated with blue laser source; it demonstrated high photoluminescence signature at longer wavelengths. Laser-induced photoluminescence signature of the investigated HC spill samples are represented in Fig. 5.
The four different HC spill samples demonstrated its distinctive fluorescence signature over the band 350-1000 nm. When the sample was stimulated with a blue laser source of 450 nm wavelength and 50 mW power, the fluorescence signature, as shown in Fig. 5, varies depending on the tested HC sample. At the perfect wavelength of the fluorescence signal, the most accurate detection and identification could take place. Diesel fuel demonstrated characteristic fluorescence signals at 470 and 515 nm, respectively, as shown in Fig. 5a. At 465 nm and 585 nm, respectively, gasoline generated single characteristic fluorescence signals, as deciphered in Fig. 5b. For various HC fuels, an identifying signature may be achieved by the variance in photoluminescence signature. lub. oil demonstrated a fluorescence signature that is different from both diesel fuel and gasoline. Figure 5c shows that lub. oil demonstrated characteristic one fluorescence signal at 505 nm. The waste oil curve shown in Fig. 5d shows that there is no re-emitted frequency on the entire studied spectrum, which means that blue laser is absorbed and this waste oil will appear dark relative to the unused lub. oil. Contaminants included in waste oil affect the fluorescence operation. Laser-induced fluorescence may be used to identify fuel from engine oil, which are all members of the HC family. Tabulation of main fluorescence signature of investigated HC spill materials is represented in Table 1. Figure 6 depicts a photoluminescence spectrum comparison for all examined HC spill materials.
Each material clearly exhibits a unique photoluminescence signature. We may use these characteristics to quickly identify HC spills. Using laser-induced photoluminescence, it may be possible to distinguish between engine fuel and oil. Because of the strong photon absorption in comparison to unused Lub. oil, we can observe from Fig. 6 that the impurities impact the oil, and the PhQY is the lowest value. The signal intensity may allow for additional differentiation; the identification of materials was not limited to the photoluminescence signature alone. The fluorescent spectra for the four diagrams  Fig. 6 show that the fluorescence frequencies at 470 nm and 515 nm are the ones with the highest intensity for diesel fuel. Based on the intensity differences at these proposed wavelengths, we can distinguish between diesel fuel and the other studied HC spills. We could also determine a peak intensity signal related to gasoline at 585 nm. This re-emitted spectral is the optimum peak frequency to identify gasoline relative to diesel fuel or oil.
Finally, we verified our data results on the selected optimum emitted wavelength (470-515 nm) for diesel oil and 585 nm related to gasoline for our HC spill identification purpose. The high photoluminescence signature was clearly observed by naked eye as shown Fig. 7a for the four studied HC spill samples. Figure 7b-d display the shows the selective 470 nm, 515 nm, and 585 nm spectral images captured by the SOC710 HI camera to achieve reorganization and grouping between diesel oil, gasoline, and lub. oil. Figure 7a shows that the 4 studied HC samples has unique re-emission signatures for each one and not interferes with the cuvette blue effect. We successfully characterized and grouped, as shown in Fig. 7b-d, the engine fuel with respect to oil for the HC spills using our fluorescent stimulation experimental setup with a blue light laser source, and assessed our imaging results using the 470 and 515 nm emitted spectral images for diesel fuel identification and the 585 nm emitted spectral image to characterize the gasoline, which is selected according to our highest intensity results to be a marker to cluster between the HC spills. The variation of signal intensity can secure another discrimination mean between different HC materials. Our techniques could aid examiners in making their initial estimations to determine the leakage source and take the right response.

Conclusion
There still is currently no novel instant identification of HC spill. Here, we effectively presented a new integrated technique for classifying emitted hyperspectral images that can be used to detect HC leakage. To categorize HC, a customized laser-induced photoluminescence was developed. Blue laser excitation is used to perform fluorescence measurements on four different HC spill samples. The hyperspectral classification technique for the stimulated HC samples by the blue laser combined with our HI imaging setup has shown that approaches for fluorescence spectra analysis enabled precise differentiation of the HC spill. Not only fluorescence signature was employed for HC spill distinguish; but also signal intensity can secure further discrimination. We effectively demonstrate that the intensity of the fuel represented by diesel and gasoline is higher than that of engine oil, which serves as our optical marker to group them. This prospective noninvasive technique would be made available by HI and enable examiners to characterize the HC spill, determine the source of spillage and resulting harm, and decide on a reaction strategy for early HC spill recognition. For HC spill identification applications, aerial systems with a custom ordinary camera that works only at a certain wavelength filter matched with the fluorescence signal outcomes from the stimulated blue laser source and an optical telescope for distance adaptation can be successfully created.
Author contribution AM executing data analysis, manuscript writing, draw figures and manuscript moderation and revision; YHE-S executing data analysis, manuscript writing, draw figures and manuscript moderation and revision.
Funding Open access funding provided by The Science, Technology & Innovation Funding Authority (STDF) in cooperation with The Egyptian Knowledge Bank (EKB).

Data availability
The authors stated and declare that all data is exist and available. The authors stated and declare that all codes are exist and available.

Competing interests
The authors stated and declare that competing of interests. All the co-authors are agreed for the research study publication.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.