An adaptive offshore oil spill detection method

Abstract

The present application relates to the field of offshore oil spill detection, especially to an adaptive offshore oil spill detection method, the present application extracts the marine environment features in the inspection image, and then analyzes the time domain changes of the marine environment features, constructs the time domain change curve of marine texture distribution features and the time domain change curve of image attribute features, and then verifies the index dispersion constraint and index mutation constraint of each curve segment, sets a label for the inspection image, determines the optimal verification perspective for the inspection image with the set label, and captures the drift feature to set a potential oil spill label under the optimal verification perspective, the present application saves the computing power for image analysis by using the change of marine environment features when large-area sea area is inspected at night, finds the optimal verification perspective of the highlighted features in the verification image under the environmental light interference by shifting the background through the optimal verification perspective, and improves the reliability of oil spill detection in large-area sea area at night through dynamic perspective selection and drift verification.

Description

Technical Field

[0001] This invention relates to the field of marine oil spill detection, and more particularly to an adaptive marine oil spill detection method. Background Technology

[0002] Marine oil spills are a serious marine environmental disaster, posing a severe threat to marine ecosystems. There is an urgent need to develop rapid and accurate monitoring technologies. Currently, drones, with their advantages of maneuverability, flexibility, and low cost, have become an important means of maritime inspection. Drones equipped with optical cameras (visible light or multispectral) can acquire high spatiotemporal resolution images of the ocean surface, providing a rich data source for oil spill detection. Furthermore, combined with relevant image processing technologies, they can identify oil spill characteristics.

[0003] For example, Chinese Patent Publication No. CN115331122A discloses a marine oil spill detection method, a drone, and a storage medium based on image segmentation. The method includes: determining an inspection area comprising multiple inspection nodes; determining an initial inspection path based on the multiple inspection nodes, the initial inspection path passing through each inspection node once; using the initial inspection path as the initial population for an artificial bee colony algorithm, and determining the energy-optimal target path using the artificial bee colony algorithm; conducting inspections according to the target path, and capturing images to be detected at each inspection node using a SAR image imager; inputting the images to be detected into a pre-trained oil spill dark spot segmentation model for image segmentation processing to obtain segmented images; and determining the marine oil spill detection result of the inspection node based on the segmented images. This application's embodiment can combine a drone equipped with a SAR image imager and an oil spill dark spot segmentation model to perform marine oil spill detection in an inspection area, thereby effectively improving the efficiency of marine oil spill detection.

[0004] However, the following problems still exist in the existing technology.

[0005] 1. Due to the wide monitoring area, the generated image data is massive, and due to the characteristics of the sea area, the image data has strong repetition. Performing unified standard traversal recognition processing on massive image data consumes a lot of computing power.

[0006] 2. At night, due to environmental factors and the varying effects of ambient light on the sea surface from different perspectives, the boundaries of oil spills become blurred and difficult to identify, affecting the accuracy of oil spill monitoring. Summary of the Invention

[0007] To address these issues, the present invention provides an adaptive marine oil spill detection method to overcome the problems in the prior art, such as the computational cost of performing standardized traversal recognition processing on massive image data, and the blurring of oil spill boundaries due to environmental factors at night, which makes them difficult to identify and affects the accuracy of oil spill monitoring.

[0008] To achieve the above objectives, the present invention provides an adaptive marine oil spill detection method, comprising:

[0009] The mobile acquisition unit is controlled to move along a predetermined inspection path in a predetermined sea area to acquire inspection images.

[0010] Extracting marine environmental features from inspection images includes: identifying several texture contours in the inspection images; filtering specific pixel bands from the texture contours based on contour attributes; determining marine texture distribution features based on the spatial distribution of specific pixel bands; and simultaneously extracting image attribute features from the inspection images.

[0011] Temporal variation analysis of marine environmental features based on inspection images includes temporally arranging the inspection images, constructing temporal variation curves of marine texture distribution features and image attribute features;

[0012] Based on the temporal variation curves of ocean texture distribution features and image attribute features, the discrete constraints and abrupt change constraints of each curve segment are verified, and labels are set for the inspection images.

[0013] In response to the label being set on the inspection image, the visual axis is determined based on the position of the inspection image to construct several verification viewpoints. The mobile acquisition unit is controlled to acquire verification images under several verification viewpoints to identify the differences in local areas in the verification images and highlight features in order to determine the optimal verification viewpoint.

[0014] The mobile acquisition unit is controlled to continuously acquire several verification images under the optimal verification perspective. After arranging them in time sequence, the fuzzy clustering boundary is determined. Based on the drift characteristics of the fuzzy clustering boundary, it is determined whether to set a potential oil spill label for the inspection image.

[0015] Furthermore, the process of identifying specific pixel bands and determining the distribution features of ocean textures based on the spatial distribution of these specific pixel bands includes:

[0016] Determine the contour attributes of several texture contours in the inspection image, including contour length and in-contour chromaticity;

[0017] If the contour attributes of the texture contour satisfy the band clustering condition, then the texture contour is determined to be a special pixel band.

[0018] Determine the shortest distance between each of the aforementioned special pixel bands and its nearest neighboring special pixel band, and calculate the mean of the shortest distance as the distribution feature of the ocean texture;

[0019] The band clustering conditions include that the contour length is greater than a predetermined length threshold and the chromaticity within the contour is within a predetermined ocean texture constraint range.

[0020] Furthermore, the range of the ocean texture constraint is predetermined, including,

[0021] Acquire several historical inspection images under corresponding lighting conditions as samples, and label the texture contours corresponding to the wave edges;

[0022] Record the average chromaticity of the pixels within each texture contour, and determine the upper and lower limits of the constraint range for the ocean texture based on the average values.

[0023] Furthermore, the image attribute feature is the average value of the image features corresponding to each pixel in the inspection image, and the image feature is either chroma or luminance.

[0024] Furthermore, the process of verifying the discrete constraints and abrupt change constraints of the indicators for each curve segment based on the temporal variation curves of ocean texture distribution features and image attribute features includes:

[0025] The amplitude variance of the peaks in each segment of the temporal variation curve of ocean texture distribution features and the temporal variation curve of image attribute features was extracted to verify the discrete constraints of the index.

[0026] Extract the aberration peaks from each segment of the temporal variation curve of ocean texture distribution features and the temporal variation curve of image attribute features, determine the proportion of aberration peaks, and verify the indicator mutation constraint.

[0027] If the amplitude variance of the peak falls within the corresponding variance constraint range, it meets the index discrete constraint. If the proportion of abnormal peaks is less than the predetermined abrupt change threshold, it meets the index abrupt change constraint. The abnormal peak is a peak in the curve segment where the difference ratio between the amplitude and the average amplitude is greater than the corresponding difference ratio threshold.

[0028] Furthermore, the process of labeling inspection images includes,

[0029] If any curve segment does not meet the discrete constraint of the index and / or does not meet the sudden change constraint of the index, then trace back the time domain segment corresponding to the curve segment;

[0030] Label all inspection images acquired within the time domain segment.

[0031] Furthermore, the process of determining the visual axis based on the position of the inspection image to construct several verification viewpoints, and controlling the mobile acquisition unit to acquire inspection images from these verification viewpoints includes,

[0032] Determine the sea area location when the inspection image is acquired, and construct an observation circle with the sea area location as the center and a predetermined distance as the radius;

[0033] Select an observation point on the observation circle, control the mobile acquisition unit to maintain the acquisition height, move it above the corresponding observation point, determine the virtual connection between the mobile acquisition unit and the sea area as the visual axis, and use the shooting angle corresponding to the visual axis as the verification angle.

[0034] Furthermore, the process of identifying distinguishing features in different local regions of the verification image to determine the optimal verification viewpoint includes,

[0035] Determine the image attribute features in each local region of the verification image, and determine the maximum difference between the image attribute features of each local region and the remaining local regions.

[0036] The mean value of the maximum difference in each local region is calculated as the feature highlighting the difference.

[0037] The verification perspective corresponding to the feature with the greatest difference is determined as the optimal verification perspective.

[0038] Furthermore, the process of determining the fuzzy cluster boundaries after temporal arrangement includes,

[0039] Perform image segmentation on the verification image and determine the segmentation boundaries;

[0040] Determine the difference ratio of image attribute features on both sides of the segmentation boundary;

[0041] If the image attribute feature difference ratio is greater than the predetermined image attribute difference ratio threshold, the segmentation boundary is determined as the fuzzy clustering boundary.

[0042] Furthermore, the process of determining whether to set a potential oil spill label for the inspection image based on the drift features of fuzzy clustering boundaries includes,

[0043] Calculate the average distance between corresponding fuzzy cluster boundaries in temporally adjacent verification images, and calculate the drift rate as a drift feature based on the average distance;

[0044] Determine whether the drift characteristics match the ocean current velocity; if they match, then set a potential oil spill label for the inspection image.

[0045] If the difference ratio between the drift feature and the ocean current velocity is less than a predetermined drift difference ratio threshold, then a match is determined.

[0046] Compared with existing technologies, this invention acquires inspection images by controlling a mobile acquisition unit, extracts marine environmental features from the inspection images, including marine texture distribution features and image attribute features, and then performs temporal variation analysis on the marine environmental features to construct temporal variation curves for marine texture distribution features and image attribute features. Subsequently, it verifies the discrete constraints and abrupt change constraints of each curve segment, labels the inspection images, determines the optimal verification viewpoint for the labeled inspection images, and captures drift features at the optimal verification viewpoint to determine whether a potential oil spill label is set for the corresponding inspection image. This invention saves computational power for image analysis by utilizing changes in marine environmental features during nighttime inspections of large-area sea areas. By shifting the background through the optimal verification viewpoint, it finds the optimal verification viewpoint where features stand out in the verification image under ambient light interference. Through dynamic viewpoint selection and drift verification, it improves the reliability of oil spill detection in large-area sea areas at night.

[0047] In particular, this invention captures marine environmental features in inspection images. In reality, due to the unique characteristics of sea areas, large areas of the sea surface are similar. The color and wave patterns of the sea surface tend to stabilize over a certain period, leading to high repetition in inspection images in most cases. Furthermore, due to the unique characteristics of oil films, they visually differ from the sea surface itself. The tension of the oil film also affects wave formation in the covered area, resulting in differences between waves in the covered and uncovered areas. Crucially, under ambient light scattering at night, the texture of wave edges is more prominent and easier to capture in nighttime environments. Based on this, this invention selects specific... The ocean texture distribution features, determined by specific pixel bands, are extracted without consuming computational power and can reflect the morphological manifestation of sea surface ripples under nighttime conditions. Furthermore, this invention considers image attribute features, which are determined based on the image features of the inspection image itself. These image features are inherent attributes of the inspection image and can be directly extracted, saving computational power. Selecting specific ocean environmental features facilitates extraction and subsequent construction of corresponding temporal variation curves, and reflects the difference between oil slicks and ordinary sea surfaces, providing data support for subsequent temporal variation analysis. It also facilitates subsequent label setting and improves the reliability of oil spill detection in large-area sea areas at night through dynamic perspective selection and drift verification.

[0048] In particular, this invention performs temporal variation analysis on the marine environmental features of inspection images and verifies the discrete constraints and abrupt change constraints of the indicators for each curve segment. Labels are assigned to the inspection images. In reality, due to the repetition of sea areas and environmental influences at night, inspection images become blurred, blurring the oil film boundary and weakening the difference between the oil film surface and the ocean surface. Therefore, considering that inspection is a continuous process, this invention considers the changes in marine environmental features in the temporal dimension. At night, the oil film boundary is easier to capture than the oil film surface, resulting in stronger data representation. Furthermore, the distribution features of marine texture in the marine environmental features are relatively prominent at night, enhancing image attribute characteristics. The characteristics at the oil slick boundary also have certain data representation. During continuous inspection, when the oil slick boundary appears in the inspection image, there will be certain abrupt changes in the image attribute features and ocean texture distribution features. In addition, some oil slick boundaries have scattered local oil slicks, which leads to the discreteness of the image attribute features and ocean texture distribution features. Therefore, by using the discrete changes and abrupt changes in the ocean environment features in the time domain, the oil slick boundary can be initially identified at night, which is convenient for subsequent selective verification. It is not necessary to verify all inspection images. Under the premise of ensuring reliability, it saves computing power and improves the reliability and detection efficiency of oil spill detection in large areas of the sea at night.

[0049] In particular, this invention constructs several verification perspectives. In practice, the inspection images are acquired by the mobile acquisition unit during the inspection process, located at different positions on the sea surface. Under different verification perspectives, the interference of background ambient light on the sea surface is different. Based on this, the invention considers identifying the distinguishing features of each local area in the verification image, finding the optimal verification perspective where the features are most prominent. This makes it easier to capture boundary features under the optimal verification perspective in nighttime environments. After determining the fuzzy clustering boundary, the invention uses drift features to determine whether to set a potential oil spill label for the inspection image. While ensuring reliability, this invention saves computing power and improves the reliability and detection efficiency of oil spill detection in large areas of the sea at night. Attached Figure Description

[0050] Figure 1 This is a schematic diagram of the steps of the adaptive marine oil spill detection method according to an embodiment of the invention;

[0051] Figure 2 This is a logic block diagram for determining a special pixel band according to an embodiment of the invention;

[0052] Figure 3 A logic block diagram for setting tags on inspection images according to an embodiment of the invention;

[0053] Figure 4 A logic block diagram for determining whether to set potential oil spill labels for inspection images in an embodiment of the invention. Detailed Implementation

[0054] To make the objectives and advantages of the present invention clearer, the present invention will be further described below with reference to embodiments; it should be understood that the specific embodiments described herein are merely for explaining the present invention and are not intended to limit the present invention.

[0055] Preferred embodiments of the present invention will now be described with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are merely illustrative of the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0056] Please see Figure 1 As shown, Figure 1 This is a schematic diagram illustrating the steps of an adaptive marine oil spill detection method according to an embodiment of the invention. The adaptive marine oil spill detection method according to an embodiment of the invention includes:

[0057] Step S1: Control the mobile acquisition unit to move along a predetermined inspection path in a predetermined sea area to acquire inspection images;

[0058] Step S2, extracting marine environmental features from the inspection image, including, determining several texture contours in the inspection image, filtering specific pixel bands from the texture contours based on contour attributes, determining marine texture distribution features based on the spatial distribution of specific pixel bands, and simultaneously extracting image attribute features of the inspection image;

[0059] Step S3: Perform temporal variation analysis based on the marine environmental features of the inspection images, including arranging the acquired inspection images in time sequence, constructing temporal variation curves of marine texture distribution features and image attribute features;

[0060] Step S4: Based on the temporal variation curve of ocean texture distribution features and the temporal variation curve of image attribute features, verify the discrete constraints and abrupt change constraints of each curve segment, and set labels for the inspection images.

[0061] Step S5: In response to the label being set on the inspection image, a number of verification viewpoints are constructed based on the position of the inspection image to determine the visual axis. The mobile acquisition unit is controlled to acquire verification images under the number of verification viewpoints to identify the differences in local areas in the verification images and highlight features in order to determine the optimal verification viewpoint.

[0062] Step S6: Control the mobile acquisition unit to continuously acquire several verification images under the optimal verification perspective, arrange them in time sequence to determine the fuzzy cluster boundary, and determine whether to set a potential oil spill label for the inspection image based on the drift characteristics of the fuzzy cluster boundary.

[0063] Specifically, there is no limitation on the specific structure of the mobile acquisition unit. It can be a drone, as long as it can be remotely controlled and equipped with image acquisition equipment to acquire inspection images. Those skilled in the art can choose the model of the drone to suit the usage environment.

[0064] Specifically, there are no restrictions on the method of setting the predetermined inspection path. The predetermined inspection path is designed by those skilled in the art so that the inspection images acquired by the mobile acquisition unit can traverse the predetermined sea area. This will not be elaborated further.

[0065] Please see Figure 2 As shown, Figure 2 This is a logic block diagram for determining specific pixel bands according to an embodiment of the invention. The process of identifying specific pixel bands and determining the distribution features of ocean textures based on the spatial distribution of specific pixel bands includes...

[0066] Determine the contour attributes of several texture contours in the inspection image, including contour length and in-contour chromaticity;

[0067] If the contour attributes of the texture contour satisfy the band clustering condition, then the texture contour is determined to be a special pixel band.

[0068] Determine the shortest distance between each of the aforementioned special pixel bands and its nearest neighboring special pixel band, and calculate the mean of the shortest distance as the distribution feature of the ocean texture;

[0069] The band clustering conditions include that the contour length is greater than a predetermined length threshold and the chromaticity within the contour is within a predetermined ocean texture constraint range.

[0070] Specifically, there are no restrictions on the method of identifying texture contours. In reality, at night, the edges of waves have textures under ambient light. For example, image segmentation algorithms can be used to identify texture contours. Of course, other methods can also be used, which will not be elaborated here.

[0071] In practice, the distance between the two ends of the contour can be used to conveniently reflect the contour length. The chromaticity within the contour is the average chromaticity of each pixel within the contour, which will not be elaborated further.

[0072] In practice, the predetermined length threshold is determined in advance. The purpose of setting the predetermined length threshold is to filter out shorter texture contours. Specifically, inspection images are acquired in advance, texture contours are identified, texture contours belonging to the wavy edge are marked by those skilled in the art, the average length of the texture contours is recorded, and the length threshold is set to 0.5 times the average length to filter out shorter texture contours.

[0073] This invention captures marine environmental features in patrol images. In reality, due to the unique characteristics of sea areas, large areas of the sea surface are similar. The color and wave patterns of the sea surface tend to stabilize over a certain period, leading to high repetition in patrol images in most cases. Furthermore, oil films, due to their unique characteristics, visually differ from the sea surface itself. The tension of the oil film also affects wave formation in the covered area, resulting in differences between waves in the covered and uncovered areas. Crucially, under ambient light scattering at night, the texture of wave edges is more prominent and easier to capture in nighttime environments. Based on this, this invention selects specific marine environments... Texture distribution features, specifically marine texture distribution features determined by special pixel bands, are extracted without consuming computational power and can reflect the morphological manifestation of sea surface ripples under nighttime conditions. Furthermore, this invention considers image attribute features, which are determined based on the image features of the inspection image itself. These image features are inherent attributes of the inspection image and can be directly extracted, saving computational power. Selecting specific marine environmental features facilitates extraction and subsequent construction of corresponding temporal variation curves, and reflects the difference between oil slicks and ordinary sea surfaces, providing data support for subsequent temporal variation analysis. It also facilitates subsequent label setting and improves the reliability of oil spill detection in large-area sea areas at night through dynamic perspective selection and drift verification.

[0074] Specifically, the range of the ocean texture constraint is predetermined and includes,

[0075] Acquire several historical inspection images under corresponding lighting conditions as samples, and label the texture contours corresponding to the wave edges;

[0076] Record the average chromaticity of the pixels within each texture contour, and determine the upper and lower limits of the constraint range for the ocean texture based on the average values.

[0077] In implementation, several historical inspection images of the same predetermined sea area collected at night can be obtained, the texture contours corresponding to the wave edges can be marked, and the mean value of the chromaticity of the pixels within the texture contour can be calculated. In implementation, the purpose of setting the ocean texture constraint range is to reflect the standard range of the wave texture chromaticity. The upper limit of the constraint is set to 1.35 times the mean value of the chromaticity of the pixels within the texture contour, and the lower limit of the constraint is set to 0.65 times the mean value of the chromaticity of the pixels within the texture contour. The ocean texture constraint range is a closed interval formed by the upper and lower limits of the constraint.

[0078] Specifically, the image attribute features are the average values ​​of the image features corresponding to each pixel in the inspection image, and the image features are either chroma or luminance.

[0079] Specifically, when an oil film is present, various image features of the oil film surface differ from those of the seawater surface in the inspection images. In practice, the preferred image feature is chromaticity.

[0080] Specifically, the process of verifying the discrete constraints and abrupt change constraints of the indicators for each curve segment based on the temporal variation curves of ocean texture distribution features and image attribute features includes:

[0081] The amplitude variance of the peaks in each segment of the temporal variation curve of ocean texture distribution features and the temporal variation curve of image attribute features was extracted to verify the discrete constraints of the index.

[0082] Extract the aberration peaks from each segment of the temporal variation curve of ocean texture distribution features and the temporal variation curve of image attribute features, determine the proportion of aberration peaks, and verify the indicator mutation constraint.

[0083] If the amplitude variance of the peak falls within the corresponding variance constraint range, it meets the index discrete constraint. If the proportion of abnormal peaks is less than the predetermined abrupt change threshold, it meets the index abrupt change constraint. The abnormal peak is a peak in the curve segment where the difference ratio between the amplitude and the average amplitude is greater than the corresponding difference ratio threshold.

[0084] The horizontal axis of the temporal variation curve of ocean texture distribution characteristics represents the time of image capture, and the vertical axis represents the ocean texture distribution characteristics corresponding to the image capture.

[0085] The horizontal axis of the image attribute feature time-domain variation curve represents the time when the inspection image was captured, and the vertical axis represents the image attribute features corresponding to the inspection image.

[0086] In practice, the purpose of setting the variance constraint range is to reflect the normal fluctuation range of the time-domain variation curve of the marine texture distribution characteristics and the time-domain variation curve of the image attribute characteristics of the inspection image corresponding to the sea surface under the condition of no oil film. To determine whether it meets the discrete constraint of the index, it is necessary to compare it with the corresponding variance constraint range.

[0087] The corresponding variance constraint range is preset, in which the inspection images during the nighttime inspection process in the oil-free sea area are acquired in advance, and several time-domain variation curves of marine texture distribution features and image attribute features are constructed.

[0088] Solve for the amplitude variance of wave peaks corresponding to several curve segments in the time-domain variation curve of ocean texture distribution characteristics, and then solve for the first mean of the amplitude variance.

[0089] Solve for the amplitude variance of the peaks corresponding to several curve segments in the time-domain variation curve of image attribute features, and then solve for the second mean of the amplitude variance.

[0090] The upper limit of the variance constraint range corresponding to the curve segment of the time-domain variation curve of the ocean texture distribution characteristics is set to 1.35 times the first mean, and the lower limit is 0.65 times the first mean.

[0091] The upper limit of the variance constraint range corresponding to the curve segment of the image attribute feature time-domain variation curve is set to 1.35 times the second mean, and the lower limit is set to 0.65 times the second mean. The variance constraint range is a closed interval formed by the upper and lower limits.

[0092] In practice, when encountering the oil film boundary, continuous abrupt peaks usually appear. The purpose of setting a threshold value for abrupt peaks is to avoid the randomness of the occurrence of abrupt peaks. In practice, the threshold value for abrupt peaks is selected within the range [0.15, 0.3], and is preferably 0.2.

[0093] The purpose of setting a difference ratio threshold in implementation is to reflect significant anomalies in wave peaks. The difference ratio threshold is preset, and corresponding difference ratio thresholds are set for the curve segments corresponding to the temporal variation curves of ocean texture distribution features and image attribute features.

[0094] Among them, the average peak value in the time domain variation curve of the ocean texture distribution characteristics corresponding to the inspection image inside the oil film is recorded in advance, the average peak value in the time domain variation curve of the ocean texture distribution characteristics of the inspection image without oil film is recorded, and the first difference ratio corresponding to the two average peak values ​​is recorded. Then, the difference ratio threshold corresponding to the curve segment of the time domain variation curve of the ocean texture distribution characteristics is the product of the first difference ratio and the offset coefficient.

[0095] The average peak value in the time-domain variation curve of the oil film and image attribute features is recorded in advance. The time-domain variation curve of the image attribute features of the corresponding inspection image without oil film is recorded. The second difference ratio corresponding to the two average peak values ​​is recorded. Then, the difference ratio threshold corresponding to the curve segment of the time-domain variation curve of the ocean texture distribution features is the product of the second difference ratio and the offset coefficient. The offset coefficient is selected in the interval [0.75, 0.85], and is preferably 0.8 in practice.

[0096] Specifically, please refer to Figure 3 The diagram shows a logic block diagram of setting labels on inspection images according to an embodiment of the invention. The process of setting labels on inspection images includes:

[0097] If any curve segment does not meet the discrete constraint of the index and / or does not meet the sudden change constraint of the index, then trace back the time domain segment corresponding to the curve segment;

[0098] Label all inspection images acquired within the time domain segment.

[0099] Specifically, the process of determining the visual axis based on the position of the inspection image to construct several verification viewpoints, and controlling the mobile acquisition unit to acquire inspection images from these verification viewpoints includes the following steps:

[0100] Determine the sea area location when the inspection image is acquired, and construct an observation circle with the sea area location as the center and a predetermined distance as the radius;

[0101] Select an observation point on the observation circle, control the mobile acquisition unit to maintain the acquisition height, move it above the corresponding observation point, determine the virtual connection between the mobile acquisition unit and the sea area as the visual axis, and use the shooting angle corresponding to the visual axis as the verification angle.

[0102] During implementation, the mobile acquisition unit is equipped with a GPS positioning system to determine the location coordinates and corresponding shooting time when the inspection images are captured in real time, and the sea area location is the sea level corresponding to the location coordinates;

[0103] In practice, the observation circle is selected within the interval [20m, 30m], with 25m being the preferred value.

[0104] In practice, eight observation points can be selected and evenly distributed on the observation circle, with the plane of the observation circle parallel to the sea level.

[0105] This invention performs temporal variation analysis on marine environmental features from inspection images, and verifies the discrete constraints and abrupt change constraints of indicators for each curve segment. Labels are assigned to the inspection images. In reality, due to the repetition of sea areas and environmental influences at night, inspection images become blurred, resulting in blurred oil film boundaries and weakened differences between the oil film surface and the ocean surface. Therefore, considering that inspection is a continuous process, this invention considers the temporal variation of marine environmental features. At night, oil film boundaries are easier to capture than the oil film surface, resulting in stronger data representation. Furthermore, the distribution characteristics of marine texture are relatively prominent at night, enhancing image attribute features. The oil slick boundary also possesses certain data characterization. During continuous inspections, when an oil slick boundary appears in the inspection image, there will be certain abrupt changes in the image attribute features and ocean texture distribution features. In addition, some oil slick boundaries have scattered local oil slicks, resulting in discreteness in the image attribute features and ocean texture distribution features. Therefore, by using the discrete changes and abrupt changes in the ocean environment features in the time domain, the oil slick boundary can be initially identified at night, which is convenient for subsequent selective verification. It is not necessary to verify all inspection images. Under the premise of ensuring reliability, it saves computing power and improves the reliability and detection efficiency of oil spill detection in large areas of the sea at night.

[0106] Specifically, the process of identifying distinguishing features in different local regions of the verification image to determine the optimal verification viewpoint includes the following steps:

[0107] Determine the image attribute features in each local region of the verification image, and determine the maximum difference between the image attribute features of each local region and the remaining local regions.

[0108] The mean value of the maximum difference in each local region is calculated as the feature highlighting the difference.

[0109] The verification perspective corresponding to the feature with the greatest difference is determined as the optimal verification perspective.

[0110] Specifically, the process of determining fuzzy cluster boundaries after temporal arrangement includes:

[0111] Perform image segmentation on the verification image and determine the segmentation boundaries;

[0112] Determine the difference ratio of image attribute features on both sides of the segmentation boundary;

[0113] If the image attribute feature difference ratio is greater than the predetermined image attribute difference ratio threshold, the segmentation boundary is determined as the fuzzy clustering boundary.

[0114] There are no restrictions on the method of image segmentation for the verification image. For example, edge detection algorithms can be used to determine the edges in the image to obtain the segmentation boundary, which can be used to characterize the boundary of the oil spill. This will not be elaborated further.

[0115] The image attribute difference ratio threshold is predetermined. The inspection images with oil film at night are predetermined, and the mean value of the first image attribute feature corresponding to the oil film area is determined. The inspection images of the sea surface without oil film at night are predetermined, and the mean value of the second image attribute feature corresponding to the oil film area is determined. The difference ratio between the mean value of the first image attribute feature and the mean value of the second image attribute feature is calculated. The image attribute difference ratio threshold is set as the product of the difference ratio and the fuzziness coefficient. The fuzziness coefficient is selected between [0.65, 0.85], preferably 0.65.

[0116] The ratio of the differences between two values ​​is the ratio of the absolute value of the difference between the two values ​​to the mean of the two values;

[0117] Specifically, please refer to Figure 4 As shown, Figure 4 The following is a logical block diagram illustrating whether to set potential oil spill labels for inspection images according to an embodiment of the invention. The process of determining whether to set potential oil spill labels for the inspection images based on the drift characteristics of fuzzy clustering boundaries includes...

[0118] Calculate the average distance between corresponding fuzzy cluster boundaries in temporally adjacent verification images, and calculate the drift rate as a drift feature based on the average distance;

[0119] Determine whether the drift characteristics match the ocean current velocity; if they match, then set a potential oil spill label for the inspection image.

[0120] Among them, if the difference ratio between the drift feature and the ocean current velocity is less than a predetermined drift difference ratio threshold, a match is determined;

[0121] The purpose of setting the drift difference ratio threshold during implementation is to reflect the matching of drift characteristics with ocean current velocity. Several continuous inspection images containing oil film boundaries are collected in advance to determine the drift characteristics of the oil film boundaries and the average difference ratio between the drift characteristics and the actual ocean current velocity. In actual situations, a certain error is allowed. The drift difference ratio threshold is set as the product of the average difference ratio and the error coefficient. The error coefficient is selected in the interval [1.15, 1.35].

[0122] Ocean current speed can be obtained by acquiring ocean current speed information in a predetermined sea area in advance.

[0123] This invention constructs several verification perspectives. In practice, inspection images are acquired by mobile acquisition units during inspections, located at different positions on the sea surface. Under different verification perspectives, the interference of ambient light on the sea surface varies. Based on this, the invention considers identifying the distinctive features of different local areas in the verification images, finding the optimal verification perspective where the features are most prominent. This makes it easier to capture boundary features under the optimal verification perspective in nighttime environments. After determining the fuzzy clustering boundary, the invention uses drift features to determine whether to set potential oil spill labels for the inspection images. This saves computing power and improves the reliability and efficiency of oil spill detection in large-area sea areas at night while ensuring reliability.

[0124] If the adaptive marine oil spill detection method of the present invention is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in the various embodiments of the present invention. The aforementioned storage medium includes various media that can store program code, such as USB flash drives, mobile hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0125] The technical solution of the present invention has been described above with reference to the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will all fall within the scope of protection of the present invention.

Claims

1. An adaptive method for detecting marine oil spills, characterized in that, include: Control the mobile acquisition unit to move along a predetermined inspection path in a predetermined sea area to acquire inspection images; Extracting marine environmental features from inspection images includes: identifying several texture contours in the inspection images; filtering specific pixel bands from the texture contours based on contour attributes; determining marine texture distribution features based on the spatial distribution of specific pixel bands; and simultaneously extracting image attribute features from the inspection images. Temporal variation analysis of marine environmental features based on inspection images includes temporally arranging the inspection images, constructing temporal variation curves of marine texture distribution features and image attribute features; Based on the temporal variation curves of ocean texture distribution features and image attribute features, the discrete constraint and abrupt change constraint of each curve segment are verified. Labels are set for the inspected images. This includes extracting the amplitude variance of the peaks in each curve segment of the temporal variation curves of ocean texture distribution features and image attribute features to verify the discrete constraint of the index; extracting the variant peaks in each curve segment of the temporal variation curves of ocean texture distribution features and image attribute features, and determining the proportion of variant peaks to verify the abrupt change constraint of the index. If the amplitude variance of the peaks falls within the corresponding variance constraint range, the discrete constraint of the index is met. If the proportion of variant peaks is less than a predetermined abrupt change threshold, the abrupt change constraint of the index is met. The variant peaks are the peaks in the curve segment whose amplitude-to-average amplitude difference ratio is greater than the corresponding difference ratio threshold. In response to the label being set on the inspection image, the visual axis is determined based on the position of the inspection image to construct several verification viewpoints. The mobile acquisition unit is controlled to acquire verification images under several verification viewpoints to identify the differences in local areas in the verification images and highlight features in order to determine the optimal verification viewpoint. The mobile acquisition unit is controlled to continuously acquire several verification images under the optimal verification perspective. After arranging them in time sequence, the fuzzy clustering boundary is determined. Based on the drift characteristics of the fuzzy clustering boundary, it is determined whether to set a potential oil spill label for the inspection image.

2. The adaptive marine oil spill detection method according to claim 1, characterized in that, The process of identifying distinctive pixel bands and determining the distribution features of ocean textures based on the spatial distribution of these distinctive pixel bands includes: Determine the contour attributes of several texture contours in the inspection image, including contour length and in-contour chromaticity; If the contour attributes of the texture contour satisfy the band clustering condition, then the texture contour is determined to be a special pixel band. Determine the shortest distance between each of the aforementioned special pixel bands and its nearest special pixel band, and calculate the mean of the shortest distances as the distribution feature of the ocean texture; The band clustering conditions include that the contour length is greater than a predetermined length threshold and the chromaticity within the contour is within a predetermined ocean texture constraint range.

3. The adaptive marine oil spill detection method according to claim 2, characterized in that, The ocean texture constraint range is predetermined and includes, Acquire several historical inspection images under corresponding lighting conditions as samples, and label the texture contours corresponding to the wave edges; Record the mean value of the chromaticity of the pixels within each texture contour, and determine the upper and lower limits of the constraint range of the ocean texture based on the mean value.

4. The adaptive marine oil spill detection method according to claim 1, characterized in that, The image attribute features are the average values ​​of the image features corresponding to each pixel in the inspection image, and the image features are either chroma or luminance.

5. The adaptive marine oil spill detection method according to claim 4, characterized in that, The process of labeling inspection images includes, If any curve segment does not meet the discrete constraint of the index and / or does not meet the sudden change constraint of the index, then trace back the time domain segment corresponding to the curve segment; Label all inspection images acquired within the time domain segment.

6. The adaptive marine oil spill detection method according to claim 1, characterized in that, The process of determining the visual axis based on the position of the inspection image, constructing several verification viewpoints, and controlling the mobile acquisition unit to acquire inspection images from these verification viewpoints includes: Determine the sea area location when the inspection image is acquired, and construct an observation circle with the sea area location as the center and a predetermined distance as the radius; Select an observation point on the observation circle, control the mobile acquisition unit to maintain the acquisition height, move it above the corresponding observation point, determine the virtual connection between the mobile acquisition unit and the sea area as the visual axis, and use the shooting angle corresponding to the visual axis as the verification angle.

7. The adaptive marine oil spill detection method according to claim 1, characterized in that, The process of identifying distinctive features in different local regions of a verification image to determine the optimal verification viewpoint includes the following steps. Determine the image attribute features in each local region of the verification image, and determine the maximum difference between the image attribute features of each local region and the remaining local regions. The mean value of the maximum difference in each local region is calculated as the feature highlighting the difference. The verification perspective corresponding to the feature with the greatest difference is determined as the optimal verification perspective.

8. The adaptive marine oil spill detection method according to claim 1, characterized in that, The process of determining fuzzy cluster boundaries after temporal arrangement includes, Perform image segmentation on the verification image and determine the segmentation boundaries; Determine the difference ratio of image attribute features on both sides of the segmentation boundary; If the image attribute feature difference ratio is greater than the predetermined image attribute difference ratio threshold, the segmentation boundary is determined as the fuzzy clustering boundary.

9. The adaptive marine oil spill detection method according to claim 1, characterized in that, The process of determining whether to set a potential oil spill label for the inspection image based on the drift features of fuzzy clustering boundaries includes: Calculate the average distance between corresponding fuzzy cluster boundaries in temporally adjacent verification images, and calculate the drift rate as a drift feature based on the average distance; Determine whether the drift characteristics match the ocean current velocity; if they match, then set a potential oil spill label for the inspection image. If the difference ratio between the drift feature and the ocean current velocity is less than a predetermined drift difference ratio threshold, then a match is determined.

Images