Oil spill quantity estimation method based on aerial survey

Abstract

The invention discloses an oil spill quantity estimation method based on aerial survey, and relates to the technical field of oil spill monitoring, and the method comprises the steps: firstly carrying out the sampling shooting of an oil spill region, setting a corresponding pixel value range according to the colors of different types of oil films in a color image, then carrying out the overall shooting, carrying out the region division of the oil films based on the pixel value range, and carrying out the calculation of the oil spill quantity. The thickness of the thin oil film is determined according to the pixel value, the thickness of the emulsified oil film and the thickness of the primary color oil film are measured through a radar, the overlapping part of the primary color oil film and the emulsified oil film is extracted, and further adjustment is carried out in a mode of setting a matching weight; and finally, superposing the oil spill amounts corresponding to all pixel points in the image to obtain the total oil spill amount. According to the technical scheme, the thicknesses of different types of oil films are detected in multiple measurement modes, the accuracy of estimation of the overall oil spill amount is improved, the complexity of later calculation is reduced, local adjustment is carried out on the boundary area of the different types of oil films, and the calculation precision is improved.

Description

Technical Field

[0001] The present application relates to the technical field of oil spill monitoring, and in particular to a method for estimating oil spill volume based on aerial survey. Background Art

[0002] In the fields of environmental monitoring and marine pollution control, oil slick detection technology is crucial for emergency response and pollution assessment in oil spills. Existing oil slick detection methods primarily rely on optical and spectral techniques, such as multispectral and hyperspectral imaging, and fluorescence imaging. While these methods can accurately capture the spectral characteristics and thickness of oil slicks under laboratory conditions, they exhibit numerous shortcomings in practical marine applications, particularly when collecting heavy oil over large areas.

[0003] When traditional methods are used to detect heavy oil in large sea areas, a single sampling method is often insufficient to detect different types of oil films due to complex sea surface conditions (such as waves, reflections, and wind interference) and the dynamic and uneven distribution of oil films. For example, for the primary color of heavy oil films, changes in their thickness have little effect on the intensity of reflected light, which makes the relationship between the spectral reflectance signal and the film thickness more complex, limiting the accuracy of the spectrometer when processing such oil films. For emulsified oil films, the scattering effect of the emulsified oil film means that there is no direct linear relationship between the intensity of the reflected light and the actual thickness of the oil film. Especially when a large number of small oil droplets and water are present on the surface of the emulsified oil film, the spectral reflectance signal may become blurred due to light scattering, making it difficult to accurately reflect the thickness of the oil film. In addition, the high-precision spectral equipment used by traditional methods is expensive, the amount of data obtained is huge, and the subsequent analysis process is complex, affecting the efficiency of real-time monitoring.

[0004] In oil spills, in addition to heavy oil, emulsified oil is also formed due to the mixing of oil and water. This oil film has significant visual and spectral characteristics that differ from heavy oil. However, due to its thin thickness and color changes that are significantly affected by ambient lighting and seawater background, traditional detection methods often have difficulty effectively capturing its characteristics, and in some cases it is even directly ignored. This selective detection not only leads to an incomplete estimate of the oil volume in the entire oil spill area, but also may underestimate the actual impact of the accident on the environment. In addition, there are clear boundaries or transition zones between different types of oil films (such as primary oil films, emulsified oil films, and thin oil films) within the oil spill area. Due to the mixing of the oil films and the gradual thickness changes, the physical properties of these regions show continuous changes. When dealing with these boundary regions, traditional methods generally do not introduce corresponding correction coefficients or directly ignore these transition zones, resulting in large errors in thickness and area calculations. When these errors accumulate in large-scale oil film estimates, they significantly affect the accuracy of the final oil spill volume. Summary of the Invention

[0005] The purpose of this application is to provide an oil spill estimation method based on aerial survey to address the shortcomings of the existing technology, so as to solve the current problems in large-scale heavy oil collection and oil film detection, such as high equipment investment cost, complex post-data processing, single sampling method is difficult to adapt to the detection of different types of oil films, insufficient detection of emulsified oil and thin oil, and insufficient processing of oil film boundary areas.

[0006] The technical solution of this application is to provide an oil spill volume estimation method based on aerial survey, which includes:

[0007] Step 1: Use an onboard camera to sample and photograph the target oil spill area to obtain a pre-collected image, manually calibrate the pre-collected image, and classify the oil film in the pre-collected image into a primary oil film, an emulsified oil film, and a thin oil film;

[0008] Step 2: Set the pixel value range corresponding to each type of oil film based on the color characteristics of different types of oil films in the pre-collected image. If the pixel value ranges of different oil films overlap, the following processing is performed: if the pixel value range of the thin oil film overlaps with the pixel value range of the primary color or emulsified oil film, the middle value of the overlapping part is used as the boundary pixel value, and the pixel value range of the thin oil film and other oil films are re-divided. If the pixel value range of the primary color oil film overlaps with the pixel value range of the emulsified oil film, the overlapping part is used as the mixed pixel value range.

[0009] Step 3: Use the onboard camera to capture a vertical color image of the target oil spill area. Perform brightness correction on the color image based on the average brightness of the pre-collected image. Then, divide the oil film in the corrected color image into regions based on the pixel value range of each oil film and the mixed pixel value range. Simultaneously, use the onboard ultrasonic radar to measure the same region and obtain the oil film thickness in the same region.

[0010] Step 4: For thin oil films, oil spills in the oil spill area are sampled and experimentally simulated to measure the relationship between different oil film thicknesses and image pixel values ​​under the same external light intensity. The thickness of the thin oil film in the corresponding area is determined based on the pixel values ​​of the thin oil film area in the color image.

[0011] Step 5: For each pixel in the mixed pixel area, a matching weight is set based on its color similarity with the original oil film and the emulsified oil film. The initial oil film thickness of the mixed pixel area measured by the airborne ultrasonic radar is retrieved, and the oil film thickness corresponding to each pixel is updated based on the matching weight.

[0012] Step 6: For different types of oil films, the oil spill volume of each pixel is calculated based on the actual area and thickness of each pixel corresponding to the type of oil film in the color image, and the oil spill components of each pixel in all color images are superimposed to obtain the total oil spill volume.

[0013] Furthermore, in step 2, the corresponding pixel value range is set according to the color features of different types of oil films in the pre-collected image, specifically including:

[0014] According to the classification results of the oil film, the original color oil film area, emulsified oil film area and thin oil film area are marked in the pre-collected image, and all pixel values ​​corresponding to each oil film area are extracted. For any area, the initial three-channel pixel value range is first set according to the extreme values ​​of the three-channel pixel values ​​in the area, and then the three-channel pixel value range of each category of oil film is adjusted according to the overlap of the pixel value range. The initial three-channel pixel value range of the original color oil film is [P Min ,P Max ], the initial three-channel pixel value range of the emulsified oil film is [E Min ,E Max ], the initial three-channel pixel value range of the thin oil film is [T Min ,T Max ], where P represents primary oil film, E represents emulsified oil film, and T represents thin oil film.

[0015] Furthermore, the pixel value range of the primary color oil film [P Min ,P Max ] is represented as:

[0016]

[0017] Where, Represents the minimum intensity value of the red channel, green channel, and blue channel in the primary color oil film area obtained by averaging the minimum values ​​of the primary color oil film pixels in different pre-collected images. Represents the maximum intensity values ​​of the red channel, green channel, and blue channel in the primary color oil film area obtained by averaging the maximum values ​​of the primary color oil film pixels in different pre-acquisition images. N is the number of pre-acquisition images, and n is the index of the pre-acquisition image.

[0018] Furthermore, in step 2, the three-channel pixel value ranges of each type of oil film are adjusted according to the overlap of the pixel value ranges, specifically including:

[0019] For any two oil films, when the pixel value ranges of the three channels intersect, it is determined that the three-channel pixel value ranges of the two oil films overlap;

[0020] The three-channel pixel value range of the thin oil film is compared with the three-channel pixel values ​​of the other two oil films. When the three-channel pixel value ranges of the two oil films being compared overlap, the channel with the smallest overlapping range is selected for adjustment, while the pixel value ranges of the other two channels remain unchanged. If the minimum overlapping range is not unique, a channel is randomly selected. For the selected channel, the middle value of the overlapping part is used as the boundary pixel value that divides the two pixel value ranges to distinguish the two categories;

[0021] The three-channel pixel value range of the original oil film is compared with the three-channel pixel value of the emulsified oil film. When the three-channel pixel value ranges of the two oil films overlap, the overlapping part is used as the mixed pixel value range, and a channel with the smallest overlapping range and no inclusion relationship in the pixel value range is selected as the judgment channel for distinguishing the original oil film, emulsified oil film and mixed pixels. If the minimum overlapping range is not unique, a channel is randomly selected as the judgment channel. For the selected channel, the pixels in the overlapping part are used as mixed pixels.

[0022] Furthermore, step 4 specifically includes:

[0023] The relationship between pixel value and thickness is expressed by a multivariable nonlinear function, and the expression of pixel value and thickness of thin oil film is obtained:

[0024] h(R,G,B)=a·exp(b R R+b G G+b B B)

[0025] Where h(R,G,B) is the thickness of the thin oil film, a is a constant coefficient, b R is the weight coefficient of the red channel, b G is the weight coefficient of the green channel, b B is the weight coefficient of the blue channel;

[0026] The experimental data on the relationship between the pixel value and thickness of the thin oil film are obtained, and the relationship data are brought into the error function expression. The optimal parameters {a, b R ,b G ,b B}, the error function is expressed as:

[0027]

[0028] In the formula, M is the number of samples of the change relationship data, m is the index of the sample, and h m is the actual oil film thickness corresponding to the mth sample, R m is the intensity value corresponding to the red channel of the mth sample, G m is the intensity value corresponding to the green channel of the mth sample, B m is the intensity value corresponding to the blue channel of the mth sample.

[0029] Furthermore, setting the corresponding matching weight in step 5 specifically includes:

[0030] For any area in the primary color oil film area and the emulsified oil film area, the mean value of each channel is taken as the central RGB value of the oil film. The central RGB value of the primary color oil film is The central RGB value of the emulsified oil film is The first Euclidean distance between a single pixel in the mixed pixel area and different oil films is calculated, and the second Euclidean distance between the original color oil film and the emulsified oil film is calculated. The color similarity between the pixel and the different oil films is determined based on the ratio of the first Euclidean distance to the second Euclidean distance. The similarity is normalized to obtain the matching weight between the pixel and the different oil films.

[0031] The color similarity between the pixel points in the mixed pixel area and different oil films is:

[0032]

[0033] Where, L P is the color similarity between the pixel and the original color oil film, L E The color similarity between the pixel and the emulsified oil film, ∈ is a constant, D P is the first Euclidean distance between the pixel and the primary color oil film, D E is the first Euclidean distance between the pixel and the emulsified oil film, D PE is the second Euclidean distance; normalizing the similarity yields:

[0034]

[0035] Where, ω P is the matching weight between the pixel point in the mixed pixel area and the original color oil film, ω E is the matching weight between the pixel point in the mixed pixel area and the emulsified oil film.

[0036] Furthermore, step 5 further includes: for the mixed pixel area, calculating the oil spill amount of the pixel based on the oil film thickness of the pixel and the matching weights corresponding to different oil films, specifically including:

[0037] The measurement data of the airborne ultrasonic radar is retrieved to obtain the initial oil film thickness of each pixel in the mixed pixel area. The oil spill volume of the pixel is calculated based on the matching weight between the pixel and different oil films:

[0038]

[0039] Where, O mix is the oil spill component of the mixed pixel area, s is the actual area corresponding to the pixel point, h mix,i is the initial oil film thickness of the i-th pixel in the mixed pixel area, c E is the oil concentration of the emulsified oil film, and I is the total number of pixels in the mixed pixel area.

[0040] Furthermore, in step 6, for the original color oil film, emulsified oil film, and thin oil film, the oil spill amount of each pixel is calculated based on the actual area and oil film thickness corresponding to each pixel in the color image, specifically including:

[0041] The measurement data of the airborne ultrasonic radar is retrieved to obtain the oil film thickness of each pixel in the original oil film area and the emulsified oil film. The oil spill component is calculated based on the actual area corresponding to the pixel and the oil film thickness of each pixel. The oil spill component of the original oil film area is:

[0042]

[0043] Where K is the number of pixels in the primary color oil film area, h P,k is the oil film thickness at the kth pixel in the primary color oil film area, and the oil spill component of the emulsified oil film is:

[0044]

[0045] Where V is the number of pixels in the emulsified oil film area, h E,d is the oil film thickness at the vth pixel in the emulsified oil film area;

[0046] Retrieve the oil film thickness data of each pixel in the thin oil film area, and calculate the oil spill component of the thin oil film based on the actual area corresponding to the pixel and the oil film thickness of each pixel:

[0047]

[0048] Where Y is the number of pixels in the thin oil film area, h T,y is the oil film thickness at the y-th pixel in the thin oil film area.

[0049] Furthermore, step 3 includes performing brightness correction on the color image according to the average brightness of the pre-collected image, specifically including: calculating the brightness normalization factor according to the average brightness of the pre-collected image in advance: L s is the average brightness of the pre-collected image, L f The average brightness of the color image is obtained by multiplying the RGB value of each pixel in the color image by scale to perform brightness correction. The corrected RGB value is expressed as R′=Rscale, G'=G·scale, and B'=B·scale.

[0050] The beneficial effects of this application are:

[0051] First, the technical solution in the present application classifies the oil films on the sea surface according to characteristics such as color and thickness, and adopts a combination of optical images and radar measurements to realize the detection of oil films of different categories, which solves the problem that a single sampling method is difficult to adapt to the detection of different types of oil films, making the overall oil spill estimation more accurate; the technical solution of the present application obtained a thickness calculation model of thin oil films through experiments in the early stage. After obtaining the color image, the thickness can be directly calculated according to the RGB value of the thin oil film area in the image. The area covered by the thin oil film on the sea surface is larger than that covered by other oil films. Compared with the method of calculating all pixels one by one when using a spectrometer, the method of directly calculating the thickness using this model in the present application simplifies the subsequent data processing process, reduces the complexity of the later analysis, and improves real-time performance; compared with the methods in the prior art, the technical solution in the present application is more suitable for large-scale heavy oil collection at sea.

[0052] Second, the technical solution in this application divides different oil films by setting a three-channel pixel value range. In order to solve the problem of inaccurate distinction of oil film boundaries due to inaccurate human labeling during the division process, this application extracts the overlapping parts between the oil films for further division to avoid division errors. In addition, this application also sets a mixed pixel value range to extract the pixels falling in the overlapping part between the primary color oil film and the emulsified oil film for further adjustment. The technical solution in this application fully considers the discontinuity and mixing effect of the oil film, compensates for the problem of blurred boundary and unclear transition between the primary color oil film and the emulsified oil film, reduces the statistical error of the mixed state oil film, and improves the accuracy of the calculation results of the total oil spill volume. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] The advantages of the above and / or additional aspects of the present application will become apparent and readily understood from the description of the embodiments in conjunction with the following drawings, in which:

[0054] Figure 1 It is a schematic flow chart of an oil spill volume estimation method based on aerial survey according to an embodiment of the present application. DETAILED DESCRIPTION

[0055] In order to more clearly understand the above-mentioned objectives, features and advantages of the present application, the present application is further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be noted that the embodiments of the present application and the features therein can be combined with each other in the absence of conflict.

[0056] In the following description, many specific details are set forth to facilitate a full understanding of the present application. However, the present application may also be implemented in other ways different from those described herein. Therefore, the scope of protection of the present application is not limited to the specific embodiments disclosed below.

[0057] like Figure 1 As shown, this embodiment provides an oil spill estimation method based on aerial survey, including:

[0058] Step 1: Use the onboard camera to sample and shoot the target oil spill area to obtain a pre-collected image, manually calibrate the pre-collected image, and classify the oil film in the pre-collected image into original color oil film, emulsified oil film and thin oil film.

[0059] It should be noted that leaked oil typically floats on the water surface, forming an oil film. Due to the action of water flow, the oil film can form film shapes of varying thicknesses. Especially in offshore areas, the surging of seawater can accelerate the emulsification of the oil film, making the oil film shape more complex. Oil films of different types and thicknesses have significant differences in physical properties, environmental impacts, and monitoring methods. The same oil film detection method cannot be applied to all oil films. Since most tanker accidents involve heavy oil, this embodiment mainly describes heavy oil, such as crude oil and HFO (Heavy Fuel Oil).

[0060] Specifically, oil films of different types or thicknesses can show different color characteristics. Thick oil films will be concentrated together and distributed in blocks, which usually appear in darker colors such as dark brown, black, and the original color of crude oil; thin oil films are formed by the diffusion of thick oil films, surrounding the thick oil films with a larger range. Thin oil films will slightly change the color of the background water, usually appearing in lighter colors such as metallic and light brown; the emulsified oil film is an emulsion-like substance mixed with oil and water, usually appearing in colors such as light yellow, orange-red, and brown, and may also have irregular ripples or turbid visual effects. Among them, the emulsified oil film is usually distributed in positions where the oil film thickness is thinner and at the edge of the oil film. The emulsified oil film at the edge is distributed in a narrow band. Therefore, oil film classification is required during monitoring. Testing based on the classification results can more accurately assess the impact of the oil film on the environment and provide a basis for subsequent oil spill accident handling.

[0061] The camera and positioning system are installed on the carrier drone. Based on the position data fed back by the positioning system, the drone is controlled to reach multiple sampling shooting locations above the target oil spill area in sequence. At each sampling shooting location, the oil spill area is photographed using the camera to obtain a pre-collected image, which is then uploaded to the host computer.

[0062] Step 2: Set the pixel value range corresponding to each type of oil film according to the color characteristics of different types of oil films in the pre-collected image. If the pixel value ranges of the oil films overlap, the following processing is performed: if the pixel value range of the thin oil film overlaps with the pixel value range of the primary color or emulsified oil film, the middle value of the overlapping part is used as the boundary pixel value, and the pixel value range of the thin oil film and other oil films is re-divided; if the pixel value range of the primary color oil film overlaps with the pixel value range of the emulsified oil film, the overlapping part is used as the mixed pixel value range.

[0063] In the host computer, the oil film is artificially divided into primary color oil film, emulsified oil film and thin oil film. According to the classification results of the oil film, the primary color oil film area, emulsified oil film area and thin oil film area are marked respectively in the pre-acquisition image, and all pixel values ​​of the primary color oil film area, emulsified oil film area and thin oil film area are extracted respectively. For any area, first, the initial three-channel pixel value range is set according to the extreme values ​​of the three-channel pixel values ​​in the area, and then the three-channel pixel value range of each category of oil film is adjusted according to the overlap of the pixel value range.

[0064] Specifically, the initial three-channel pixel value range corresponding to the primary color oil film is [P Min ,P Max ], the initial three-channel pixel value range corresponding to the emulsified oil film is [E Min ,E Max ], the initial three-channel pixel value range corresponding to the thin oil film is [T Min ,T Max ], where P represents primary oil film, E represents emulsified oil film, and T represents thin oil film.

[0065] The initial three-channel pixel value range corresponding to the primary color oil film [P Min ,P Max ] is represented as:

[0066]

[0067] In the formula, R represents the red channel, G represents the green channel, and B represents the blue channel. Represents the minimum intensity value of the red channel, green channel, and blue channel in the primary color oil film area obtained by averaging the minimum values ​​of the primary color oil film pixels in different pre-collected images. represents the maximum intensity value of the red channel, green channel, and blue channel in the primary color oil film area obtained by averaging the maximum values ​​of the primary color oil film pixels in different pre-collected images. N is the number of pre-collected images, and n is the index of the pre-collected image (i.e., n is the nth pre-collected image). Represents the minimum intensity value of the red channel of the original color oil film area in the nth pre-collected image, Represents the maximum intensity value of the red channel of the original color oil film area in the nth pre-collected image, Represents the minimum intensity value of the green channel of the original color oil film area in the nth pre-collected image, Represents the maximum intensity value of the green channel of the original color oil film area in the nth pre-collected image, Represents the minimum intensity value of the blue channel of the original color oil film area in the nth pre-collected image, Represents the maximum intensity value of the green channel of the original color oil film area in the nth pre-collected image.

[0068] The initial three-channel pixel value range corresponding to the emulsified oil film [E Min ,E Max ] is represented as:

[0069]

[0070] Where, Represents the minimum intensity value of the red channel, green channel, and blue channel in the emulsified oil film area obtained by averaging the minimum values ​​of the emulsified oil film pixels in different pre-collected images. represents the maximum intensity values ​​of the red channel, green channel, and blue channel in the emulsified oil film area obtained by averaging the maximum values ​​of the emulsified oil film pixels in different pre-collected images; in this embodiment, the calculation process of the initial three-channel pixel value range of the emulsified oil film is the same as that of the original color oil film, and will not be repeated here.

[0071] The initial three-channel pixel value range corresponding to the thin oil film [T Min ,T Max ] is represented as:

[0072]

[0073] Where, represents the minimum intensity value of the red channel, green channel, and blue channel in the thin oil film area obtained by averaging the minimum values ​​of the thin oil film pixels in different pre-acquisition images. represents the maximum intensity value of the red channel, green channel and blue channel in the thin oil film area obtained by averaging the maximum values ​​of the thin oil film pixels in different pre-collected images; in this embodiment, the calculation process of the initial three-channel pixel value range of the thin oil film is the same as that of the original color oil film, and will not be repeated here.

[0074] For any two oil films, when the pixel value ranges of the three channels intersect, the three-channel pixel value ranges of the two oil films are determined to overlap. Based on the overlap of the initial three-channel pixel value ranges of the three oil films, the three-channel pixel value ranges of each type of oil film are adjusted, specifically including:

[0075] The three-channel pixel value range of the thin oil film is compared with the three-channel pixel values ​​of the other two oil films (including the original color oil film and the emulsified oil film). When the three-channel pixel value ranges of the two compared oil films overlap, the channel with the smallest overlapping range is selected for adjustment, and the pixel value ranges of the other two channels remain unchanged. If the minimum overlapping range is not unique, a channel is randomly selected for adjustment. For the selected channel, the middle value of the overlapping part is used as the boundary pixel value for dividing the two pixel value ranges to distinguish the two categories. Among them, if the two pixel value ranges are the same, the other channel is selected for adjustment.

[0076] For example, and Overlap, the middle value R of the overlapping part mid As the boundary pixel value, if Then As the pixel value range of the thin oil film, The pixel value range used as the primary color oil film.

[0077] The three-channel pixel value range of the original color oil film is compared with the three-channel pixel value of the emulsified oil film. When the three-channel pixel value ranges of the two oil films overlap, the overlapping part is used as the mixed pixel value range. At the same time, a channel with the smallest overlapping range and no inclusion relationship in the pixel value range is selected as the judgment channel for distinguishing the original color oil film, emulsified oil film and mixed pixels. If the minimum overlapping range is not unique, a channel is randomly selected as the judgment channel. For the selected channel, the pixels in the overlapping part are mixed pixels, the remaining pixels in the original color oil film pixel value range excluding the overlapping part are original color oil film pixels, and the remaining pixels in the emulsified oil film pixel value range excluding the overlapping part are emulsified oil film pixels.

[0078] In this embodiment, after obtaining the initial three-channel pixel value range corresponding to each oil film, the three-channel pixel value ranges of any two oil film areas are compared to avoid the situation where the three-channel pixel value ranges completely overlap and the three-channel pixel value ranges are all in an inclusion relationship (that is, the pixel value ranges of each channel are in an inclusion relationship, and the inclusion relationship means that a small range is included in another large range). If the three-channel pixel value ranges corresponding to the two oil films completely overlap or are all in an inclusion relationship, the initial three-channel pixel value ranges need to be recalculated until an incompletely overlapping three-channel pixel value range appears, so as to prevent the included pixel value range from being completely removed when removing the overlapping part, resulting in the oil film losing the pixel value range of one channel and a calculation error (here it is mainly to avoid the situation where the pixel ranges of the two oil films cannot be distinguished).

[0079] In this embodiment, the type of oil in an oil spill is uncertain, and each type of oil has distinct color characteristics. Therefore, before officially capturing a color image of the oil spill area, it is necessary to pre-collect multiple images for different oil film types. These images serve as pre-collected images. After an oil spill occurs, the pre-collected images are used to obtain color feature information of the oil film. Based on this color feature information, the colors corresponding to different types of oil film are determined. Based on the colors of different oil film types, the pixel value ranges for the respective oil film types are determined. This allows the final captured color image to be quickly segmented into regions based on the preset pixel value ranges, thereby determining the location distribution characteristics of different oil film types.

[0080] It should be noted that when labeling various types of oil films in the pre-collected images based on human experience, some details will be ignored, resulting in inaccurate labeling problems. For example, the boundaries of various types of oil films are not accurately distinguished, resulting in overlap between the three-channel pixel value ranges corresponding to different types of oil films, especially for primary color oil films and emulsified oil films. In the target oil spill area, the edge of the primary color oil film or the location where the oil layer is thinner has a large contact area with the seawater, which is more likely to cause slight emulsification. These locations usually show a mixed state of primary color oil film and emulsified oil film, with a small area and no obvious color change. When manually labeled, they are usually directly classified as primary color oil film. In specific image processing, if the pixel values ​​of the initial three channels corresponding to each oil film are directly used for division, not only will the division be incorrect, but the influence of the mixed state of primary color oil film and emulsified oil film on the total oil spill will also be ignored, resulting in inaccurate calculation results. In addition, because primary color oil films on the sea are usually irregularly distributed and their edges are long, if no adjustment is made when calculating the oil spill volume, the calculation error will increase. Therefore, this application extracts the overlapping parts between each thin oil film and makes further divisions to avoid the problem of division errors. This application also sets a mixed pixel value range to extract the pixels falling in the overlapping parts between the primary oil film and the emulsified oil film for further processing to weaken the influence of the mixed state of the primary oil film and the emulsified oil film on the total oil spill volume and improve the accuracy of the calculation results.

[0081] Step 3: Use the airborne camera to capture a vertical color image of the target oil spill area. The color image is then brightness-corrected based on the average brightness of the pre-collected image. The interval between pre-collection and formal collection should not be too long to avoid significant changes in ambient light. These images can also be performed simultaneously, with the pre-collected image first sent back to the host computer for processing. The oil film in the corrected color image is divided into regions based on the pixel value range and mixed pixel value range corresponding to each oil film, resulting in primary color oil film regions, emulsified oil film regions, thin oil film regions, and mixed pixel regions. Simultaneously, an airborne ultrasonic radar is used to measure the same region on the sea surface (i.e., the target oil spill region captured by the airborne camera) to obtain oil film thickness data for the same region. The camera is positioned at roughly the same height above the sea surface for each capture.

[0082] The target oil spill area is divided into multiple shooting areas of equal size. The positioning system and camera are installed on a carrier drone. Based on the position data fed back by the positioning system, the drone is controlled to reach the shooting points above each shooting area in sequence along a pre-set route. At each shooting point, the camera is used to shoot in a vertical direction to obtain a color image from a vertical perspective. The color images collected by the camera in each shooting area together constitute the overall color image of the target oil spill area, that is, the shooting range covers the entire oil spill area, wherein each shooting point has the same height from the sea surface. The collected color images are uploaded to the host computer for brightness correction and oil film area division (that is, the original color oil film, emulsified oil film, thin oil film and mixed pixel area are divided in the color image according to the pixel value range and mixed pixel value range corresponding to each oil film).

[0083] While taking color images, the same area is scanned and measured using an airborne ultrasonic radar to obtain the oil film thickness data of the same area, and the measured image data is positionally aligned with the ultrasonic radar data. The specific method is as follows: the pixel points in the color image are matched one-to-one with the coordinate points in the positioning system, and then the measurement points of the airborne ultrasonic radar are matched one-to-one with the coordinate points in the positioning system, so that the color image pixel points and the radar measurement points can be spatially aligned in a unified positioning coordinate system. The color image is then divided into a grid, where the grids are squares of equal area, so that a single grid is smaller than a preset area and a single grid contains more than two radar measurement points. The average of the measurement results of each radar measurement point is taken as the oil film thickness of the pixel points corresponding to the primary color oil film, emulsified oil film, and mixed pixels in the grid.

[0084] In this embodiment, the airborne ultrasonic radar needs to perform large-scale measurements, and an array-type ultrasonic radar can be used, such as a MIMO imaging radar, phased array radar, or SAR synthetic aperture radar. In this embodiment, the camera and radar are both mounted on the drone and rigidly connected, preventing relative motion. The positioning system provides the drone's position and attitude in the ground coordinate system. Each pixel captured by the camera can be calculated using camera intrinsic and extrinsic parameters to determine its position in the ground coordinate system. Each measurement point captured by the radar is a three-dimensional point with direction and distance information, which can be directly transformed to the ground coordinate system using the radar's extrinsic parameters.

[0085] In order to quickly complete brightness correction and oil film area division, before capturing the overall color image of the target oil spill area, the brightness normalization factor is calculated based on the average brightness of the pre-collected image. The brightness normalization factor is expressed as:

[0086]

[0087] Where, L s is the average brightness of the pre-collected image, Lf is the average brightness of the color image collected in step 3. The calculation formula of average brightness is as follows:

[0088]

[0089] Where, L includes L s and L f , X is the number of pixels in a single image; the RGB value of each pixel in the color image is multiplied by scale to perform brightness correction (when the corrected value exceeds the range [0,255], the excess part will be clipped. For example, when the corrected value is 257, the excess part will be clipped to obtain a value of 155). The corrected RGB value is expressed as:

[0090] R′=R·scale, G′=G·scale, B′=B·scale

[0091] In this embodiment, pixels in the color image that are not within the pixel value range corresponding to the primary color oil film, the emulsified oil film, and the thin oil film are used as background pixels. These background pixels correspond to the sea surface without oil film.

[0092] In this embodiment, since the target oil spill area is divided into multiple shooting areas of equal size, after completing shooting in a single shooting area, the camera will upload the color image to the host computer, and at the same time, the drone equipped with ultrasonic radar will be controlled to the same area to measure the oil film thickness. The host computer uses existing image processing software platforms (such as MATLAB, OpenCV and other software tools) to divide the oil film in the color image into regions in real time based on the pixel value ranges corresponding to different types of oil film, and obtain the regions corresponding to different types of oil film.

[0093] Step 4: For thin oil films, the oil spill area is sampled and experimentally simulated to measure the relationship between different oil film thicknesses and image pixel values ​​under the same external light intensity. The thickness of the thin oil film in the corresponding area is determined based on the size of the pixel value of the thin oil film area in the color image.

[0094] It should be noted that before determining the thickness of the thin oil film based on the size of the pixel value of the thin oil film area in the color image, it is necessary to obtain the change relationship data of the thin oil film pixel value and thickness through experimental measurement in advance, and then fit the mathematical model of pixel value and thickness based on these corresponding data through nonlinear fitting. In subsequent calculations, the corresponding thickness is directly calculated through this model, that is, the pixel value of the thin oil film area in the color image.

[0095] The relationship between pixel value and thickness is expressed by a multivariable nonlinear function, and the expression of pixel value and thickness of thin oil film is obtained as follows:

[0096] h(R,G,B)=a·exp(b R R+bG G+b B B)

[0097] Where h(R,G,B) is the thickness of the thin oil film, a is a constant coefficient, representing the scale factor of the model, and b R is the weight coefficient of the red channel, which indicates the influence of the red channel on the oil film thickness, b G is the weight coefficient of the green channel, which indicates the influence of the green channel on the oil film thickness, b B is the weight coefficient of the blue channel, which represents the influence of the blue channel on the oil film thickness.

[0098] The oil spilled in the oil spill area is sampled. Under the same external light intensity, the oil spill sample is gradually added to the experimental tank according to the predetermined amount. After each addition of the oil spill sample, the oil film image is vertically photographed at a fixed height using a camera, and the oil film thickness is measured using a thickness measuring device. The pixel value of the oil film at each thickness is recorded to obtain the relationship data between the pixel value and thickness of the thin oil film. The relationship data is substituted into the error function (representing the difference between the predicted value and the actual value) expression, and the optimal parameters {a, b} are obtained by minimizing the error function. R ,b G ,b B}, the error function is expressed as:

[0099]

[0100] In the formula, M is the number of samples of the change relationship data, m is the index of the sample, and h m is the actual oil film thickness corresponding to the mth sample (i.e., the thickness value measured in the experiment), R m is the intensity value corresponding to the red channel of the mth sample, G m is the intensity value corresponding to the green channel of the mth sample, B m is the intensity value corresponding to the blue channel of the mth sample.

[0101] In this embodiment, thin oil films are typically thin and uniform, exhibiting distinct color characteristics during optical inspection. The color of thin oil films (i.e., changes in RGB values) vary significantly with thickness, allowing the film thickness to be inferred using a pre-established mathematical model of pixel value and thickness. However, primary oil films are thicker and less transparent, and emulsified oil films have strong light scattering properties. These two types of oil films do not vary significantly in color with thickness, making it impossible to use a pixel value and thickness model to determine thickness.

[0102] Step 5: For the mixed pixel area, for each pixel point therein, set the corresponding matching weight (i.e., the oil contribution weight corresponding to different oil films) according to its color similarity with the original color oil film and the emulsified oil film, retrieve the initial oil film thickness of the mixed pixel area measured by the airborne ultrasonic radar, and update the oil film thickness corresponding to each pixel point based on the matching weight.

[0103] For any area in the primary color oil film area and the emulsified oil film area in a single color image, the mean value of each channel is taken as the central RGB value of the oil film. The central RGB value of the primary color oil film is The central RGB value of the emulsified oil film is Among them, the central RGB value of the primary color oil film Expressed as:

[0104]

[0105] Where M is the total number of pixel values ​​in the primary color oil film area, i is the index of each pixel point in the primary color oil film area, is the red channel intensity value of the i-th pixel in the primary color oil film area, is the green channel intensity value of the i-th pixel in the primary color oil film area, is the blue channel intensity value of the i-th pixel in the primary color oil film area; the calculation method of the central RGB value of the emulsified oil film is the same as that of the primary color oil film, which will not be repeated here.

[0106] For a single pixel in the mixed pixel area, the first Euclidean distance between the pixel and different oil films is calculated based on its three-channel pixel values. The second Euclidean distance between the original color oil film and the emulsified oil film is calculated. The color similarity between the pixel and the different oil films is determined based on the ratio of the first Euclidean distance to the second Euclidean distance. The similarity is normalized to obtain the matching weights of the pixel and the original color oil film and the matching weights of the pixel and the emulsified oil film. Specifically, the first Euclidean distance between the pixel and different oil films is expressed as:

[0107]

[0108] Where D P is the first Euclidean distance between the pixel and the primary color oil film, D E is the first Euclidean distance between the pixel and the emulsified oil film, j is the index of the pixel in the mixed pixel area; the second Euclidean distance between the original color oil film and the emulsified oil film is expressed as:

[0109]

[0110] The color similarity between the pixel points in the mixed pixel area and different oil films is expressed as:

[0111]

[0112] Where, L P is the color similarity between the pixel and the original color oil film, L E The color similarity between the pixel and the emulsified oil film, ∈, is a small constant to prevent division by zero errors; the similarity is normalized to obtain the matching weight:

[0113]

[0114] Where, ω P is the matching weight between the pixel point and the original color oil film in the mixed pixel area, ω E is the matching weight between the pixel point and the emulsified oil film in the mixed pixel area.

[0115] The oil concentration of each point of the offshore emulsified oil film is obtained by sampling detection. The average oil concentration of each point is taken as the oil concentration of the emulsified oil film. The measurement data of the airborne ultrasonic radar is retrieved to obtain the initial oil film thickness of each pixel in the mixed pixel area. The oil film thickness corresponding to each pixel is updated based on the initial oil film thickness and matching weight. The oil spill volume of the pixel is then calculated based on the oil film thickness corresponding to each pixel and the actual area corresponding to the pixel:

[0116]

[0117] Where, O mix is the oil spill component of the mixed pixel area, s is the actual area corresponding to the pixel point, h mix,i is the initial oil film thickness of the i-th pixel in the mixed pixel area, c E is the oil concentration of the emulsified oil film, I is the total number of pixels in the mixed pixel area, h mix,i (ω P +ω E c E ) represents the updated oil film thickness of the i-th pixel in the mixed pixel area.

[0118] Step 6: For different types of oil films, the oil spill volume of each pixel is calculated based on the actual area and thickness of each pixel corresponding to the type of oil film in the color image, and the oil spill volume of each pixel in all color images is superimposed to obtain the total oil spill volume.

[0119] For the primary color oil film in the color image, the measurement data of the airborne ultrasonic radar is retrieved to obtain the oil film thickness of each pixel in the primary color oil film area. Based on the actual area and oil film thickness corresponding to the pixel point, the oil spill component of the primary color oil film is calculated:

[0120]

[0121] Where, OP is the oil spill component of the primary color oil film area, K is the number of pixels in the primary color oil film area, k is the index of each pixel in the primary color oil film area, h P,k is the oil film thickness at the kth pixel in the primary color oil film area.

[0122] For the emulsified oil film in the color image, the measurement data of the airborne ultrasonic radar is retrieved to obtain the oil film thickness of each pixel in the emulsified oil film area. Based on the actual area and oil film thickness corresponding to the pixel point, the oil spill component of the emulsified oil film is calculated:

[0123]

[0124] Where, O E is the oil spill component in the emulsified oil film area, V is the number of pixels in the emulsified oil film area, v is the index of each pixel in the emulsified oil film area, and h E,v is the oil film thickness at the vth pixel in the emulsified oil film area.

[0125] Retrieve the oil film thickness data corresponding to each pixel in the thin oil film area in the color image, and calculate the oil spill component of the thin oil film based on the actual area and oil film thickness corresponding to the pixel:

[0126]

[0127] Where, O T is the oil spill component in the thin oil film area, Y is the number of pixels in the thin oil film area, y is the index of each pixel in the thin oil film area, and h T,y is the oil film thickness at the y-th pixel in the thin oil film area.

[0128] The amount of oil spill corresponding to a single color image is:

[0129] O Sec =O P +O E +O T +O mix

[0130] The oil spill volumes corresponding to all color images are superimposed to obtain the total oil spill volume.

[0131] In this embodiment, the oil spill area captured in a single color image is only a part of the seawater oil spill area. When calculating the total oil spill volume, the calculated oil spill volumes corresponding to all color images need to be superimposed.

[0132] The technical solution of the present application uses radar to detect thicker oil films and determines the thickness of thin oil films based on color. It is suitable for darker heavy oils, such as crude oil, HFO, etc. In addition, the technical solution of the present application can divide the target oil spill area in a grid manner, calculate the oil spill volume based on the color image of a single area, and finally perform superposition calculations. It is suitable for detecting a large range of target oil spill areas.

[0133] The steps in this application can be adjusted in order, combined, and deleted according to actual needs.

[0134] The units in the device of the present application can be combined, divided and deleted according to actual needs.

[0135] Although the present application is disclosed in detail with reference to the accompanying drawings, it should be understood that these descriptions are merely exemplary and are not intended to limit the application of the present application. The scope of protection of the present application is defined by the appended claims and may include various modifications, alterations and equivalents made to the invention without departing from the scope and spirit of the present application.

Claims

1. A method for estimating oil spill volume based on aerial survey, characterized in that: The method includes: Step 1: Use an onboard camera to sample and photograph the target oil spill area to obtain a pre-collected image, manually calibrate the pre-collected image, and classify the oil film in the pre-collected image into a primary oil film, an emulsified oil film, and a thin oil film; Step 2: Set the pixel value range corresponding to each type of oil film based on the color characteristics of different types of oil films in the pre-collected image. If the pixel value ranges of different oil films overlap, the following processing is performed: if the pixel value range of the thin oil film overlaps with the pixel value range of the primary color or emulsified oil film, the middle value of the overlapping part is used as the boundary pixel value, and the pixel value range of the thin oil film and other oil films are re-divided. If the pixel value range of the primary color oil film overlaps with the pixel value range of the emulsified oil film, the overlapping part is used as the mixed pixel value range. Step 3: Use the onboard camera to capture a vertical color image of the target oil spill area. Perform brightness correction on the color image based on the average brightness of the pre-collected image. Then, divide the oil film in the corrected color image into regions based on the pixel value range of each oil film and the mixed pixel value range. Simultaneously, use the onboard ultrasonic radar to measure the same region and obtain the oil film thickness in the same region. Step 4: For thin oil films, oil spills in the oil spill area are sampled and experimentally simulated to measure the relationship between different oil film thicknesses and image pixel values ​​under the same external light intensity. The thickness of the thin oil film in the corresponding area is determined based on the pixel values ​​of the thin oil film area in the color image. Step 5: For each pixel in the mixed pixel area, a matching weight is set based on its color similarity with the original oil film and the emulsified oil film. The initial oil film thickness of the mixed pixel area measured by the airborne ultrasonic radar is retrieved, and the oil film thickness corresponding to each pixel is updated based on the matching weight. Step 6: For different types of oil films, the oil spill volume of each pixel is calculated based on the actual area and thickness of each pixel corresponding to the type of oil film in the color image, and the oil spill components of each pixel in all color images are superimposed to obtain the total oil spill volume.

2. The oil spill estimation method based on aerial survey according to claim 1, characterized in that: In step 2, the corresponding pixel value range is set according to the color features of different types of oil films in the pre-collected image, specifically including: According to the classification results of the oil film, the original color oil film area, emulsified oil film area and thin oil film area are marked in the pre-collected image, and all pixel values ​​corresponding to each oil film area are extracted. For any area, the initial three-channel pixel value range is first set according to the extreme values ​​of the three-channel pixel values ​​in the area, and then the three-channel pixel value range of each category of oil film is adjusted according to the overlap of the pixel value range. The initial three-channel pixel value range of the original color oil film is [P Min ,P Max ], the initial three-channel pixel value range of the emulsified oil film is [E Min ,E Max ], the initial three-channel pixel value range of the thin oil film is [T Min ,T Max ], where P represents primary oil film, E represents emulsified oil film, and T represents thin oil film.

3. The oil spill estimation method based on aerial survey as claimed in claim 2, characterized in that: The pixel value range of the primary color oil film [P Min ,P Max ] is represented as: Where, Represents the minimum intensity value of the red channel, green channel, and blue channel in the primary color oil film area obtained by averaging the minimum values ​​of the primary color oil film pixels in different pre-collected images. Represents the maximum intensity values ​​of the red channel, green channel, and blue channel in the primary color oil film area obtained by averaging the maximum values ​​of the primary color oil film pixels in different pre-acquisition images. N is the number of pre-acquisition images, and n is the index of the pre-acquisition image.

4. The oil spill estimation method based on aerial survey according to claim 3, characterized in that: In step 2, the three-channel pixel value ranges of each type of oil film are adjusted according to the overlap of the pixel value ranges, specifically including: For any two oil films, when the pixel value ranges of the three channels intersect, it is determined that the three-channel pixel value ranges of the two oil films overlap; The three-channel pixel value range of the thin oil film is compared with the three-channel pixel values ​​of the other two oil films. When the three-channel pixel value ranges of the two oil films being compared overlap, the channel with the smallest overlapping range is selected for adjustment, while the pixel value ranges of the other two channels remain unchanged. If the minimum overlapping range is not unique, a channel is randomly selected. For the selected channel, the middle value of the overlapping part is used as the boundary pixel value that divides the two pixel value ranges to distinguish the two categories; The three-channel pixel value range of the original oil film is compared with the three-channel pixel value of the emulsified oil film. When the three-channel pixel value ranges of the two oil films overlap, the overlapping part is used as the mixed pixel value range, and a channel with the smallest overlapping range and no inclusion relationship in the pixel value range is selected as the judgment channel for distinguishing the original oil film, emulsified oil film and mixed pixels. If the minimum overlapping range is not unique, a channel is randomly selected as the judgment channel. For the selected channel, the pixels in the overlapping part are used as mixed pixels.

5. The oil spill estimation method based on aerial survey as claimed in claim 2, characterized in that: The step 4 specifically includes: The relationship between pixel value and thickness is expressed by a multivariable nonlinear function, and the expression of pixel value and thickness of thin oil film is obtained: h(R,G,B)=a·exp(b R R+b G G+b B B) Where h(R,G,B) is the thickness of the thin oil film, a is a constant coefficient, b R is the weight coefficient of the red channel, b G is the weight coefficient of the green channel, b B is the weight coefficient of the blue channel; The experimental data on the relationship between the pixel value and thickness of the thin oil film are obtained, and the relationship data are brought into the error function expression. The optimal parameters {a, b R ,b G ,b B }, the error function is expressed as: In the formula, M is the number of samples of the change relationship data, m is the index of the sample, and h m is the actual oil film thickness corresponding to the mth sample, R m is the intensity value corresponding to the red channel of the mth sample, G m is the intensity value corresponding to the green channel of the mth sample, B m is the intensity value corresponding to the blue channel of the mth sample.

6. The oil spill estimation method based on aerial survey according to claim 3, characterized in that: Setting the corresponding matching weight in step 5 specifically includes: For any area in the primary color oil film area and the emulsified oil film area, the mean value of each channel is taken as the central RGB value of the oil film. The central RGB value of the primary color oil film is The central RGB value of the emulsified oil film is The first Euclidean distance between a single pixel in the mixed pixel area and different oil films is calculated, and the second Euclidean distance between the original color oil film and the emulsified oil film is calculated. The color similarity between the pixel and the different oil films is determined based on the ratio of the first Euclidean distance to the second Euclidean distance. The similarity is normalized to obtain the matching weight between the pixel and the different oil films. The color similarity between the pixel points in the mixed pixel area and different oil films is: Where, L P is the color similarity between the pixel and the original color oil film, L E The color similarity between the pixel and the emulsified oil film, ∈ is a constant, D P is the first Euclidean distance between the pixel and the primary color oil film, D E is the first Euclidean distance between the pixel and the emulsified oil film, D PE is the second Euclidean distance; normalizing the similarity yields: Where, ω P is the matching weight between the pixel point in the mixed pixel area and the original color oil film, ω E is the matching weight between the pixel point in the mixed pixel area and the emulsified oil film.

7. The oil spill estimation method based on aerial survey according to claim 6, characterized in that: The step 5 further includes: for the mixed pixel area, calculating the oil spill amount of the pixel based on the oil film thickness of the pixel and the matching weights corresponding to different oil films, specifically including: The measurement data of the airborne ultrasonic radar is retrieved to obtain the initial oil film thickness of each pixel in the mixed pixel area. The oil spill volume of the pixel is calculated based on the matching weight between the pixel and different oil films: Where, O mix is the oil spill component of the mixed pixel area, s is the actual area corresponding to the pixel point, h mix,i is the initial oil film thickness of the i-th pixel in the mixed pixel area, c E is the oil concentration of the emulsified oil film, and I is the total number of pixels in the mixed pixel area.

8. The oil spill estimation method based on aerial survey according to claim 5, characterized in that: In step 6, for the primary oil film, emulsified oil film, and thin oil film, the oil spill amount of each pixel is calculated based on the actual area and oil film thickness corresponding to each pixel in the color image, specifically including: The measurement data of the airborne ultrasonic radar is retrieved to obtain the oil film thickness of each pixel in the original oil film area and the emulsified oil film. The oil spill component is calculated based on the actual area corresponding to the pixel and the oil film thickness of each pixel. The oil spill component of the original oil film area is: Where K is the number of pixels in the primary color oil film area, h P,k is the oil film thickness at the kth pixel in the primary color oil film area, and the oil spill component of the emulsified oil film is: Where V is the number of pixels in the emulsified oil film area, h E,d is the oil film thickness at the vth pixel in the emulsified oil film area; Retrieve the oil film thickness data of each pixel in the thin oil film area, and calculate the oil spill component of the thin oil film based on the actual area corresponding to the pixel and the oil film thickness of each pixel: Where Y is the number of pixels in the thin oil film area, h T,y is the oil film thickness at the y-th pixel in the thin oil film area.

9. The oil spill estimation method based on aerial survey according to claim 1, characterized in that: The step 3 includes performing brightness correction on the color image according to the average brightness of the pre-collected image, specifically including: calculating the brightness normalization factor according to the average brightness of the pre-collected image in advance: L s is the average brightness of the pre-collected image, L f The average brightness of the color image is obtained by multiplying the RGB value of each pixel in the color image by scale to perform brightness correction. The corrected RGB value is expressed as R′=R·scale, G′=G·scale, and B′=B·scale.

Images