A method for oil film detection based on CEO and RCOA

Abstract

The application relates to the technical field of oil spill monitoring, in particular to an oil film detection method based on CEO and RCOA, which comprises data preprocessing; dividing a population into an elite subgroup and a common subgroup according to fitness values; adopting a CEO algorithm to perform large-scale search in the whole solution space through a multi-center gravity mechanism and a Levy flight strategy for the elite subgroup; adopting an RCOA algorithm to perform fine search around elite individuals through a spiral following mechanism and a Gaussian disturbance strategy for the common subgroup; triggering a bidirectional interaction mechanism every fixed iteration interval to realize information sharing between the two subgroups and enhance population diversity; performing post-processing operations to obtain oil spill detection results. The application constructs a weighted function fusing regional statistical features and edge structure features through a fitness function, and introduces a low threshold dynamic promotion factor and multiple constraint penalty terms to constrain the search space, so that convergence to a mathematically optimal but unreasonable solution is avoided.

Description

Technical Field

[0001] This invention relates to the field of oil spill monitoring technology, and in particular to an oil film detection method based on CEO and RCOA. Background Technology

[0002] Currently, spaceborne and airborne oil spill detection technologies have made significant progress, but these two methods cannot fully meet the requirements for real-time oil spill tracking. Marine radar oil spill detection technology aims to provide effective detection data in emergency response to oil spill incidents. Subsequently, with the continuous advancement of marine radar technology, both detection range and accuracy have been significantly improved.

[0003] Oil spill monitoring technology is crucial for timely oil spill cleanup and accurate assessment of ecological and environmental damage. It includes dynamic monitoring of oil spills, tracking oil film drift and spread, monitoring oil spill weathering, and assessing the impact on ecosystems.

[0004] However, existing technologies are not efficient and accurate enough for oil spill monitoring. Summary of the Invention

[0005] To address this, the present invention provides an oil film detection method based on CEO (Cosmic Evolution Optimization) and RCOA (Reindeer Cyclone Optimization Algorithm) to overcome the problems of insufficient efficiency and accuracy in existing oil spill monitoring technologies.

[0006] To achieve the above objectives, the present invention provides an oil film detection method based on CEO and RCOA, comprising: Data preprocessing; The population is divided into elite and ordinary subpopulations based on fitness values. The elite subgroup uses the CEO algorithm to perform a large-scale search in the entire solution space through a multi-center gravity mechanism and a Lévy flight strategy. The ordinary subgroup uses the RCOA algorithm to perform a fine search around the elite individuals through a spiral following mechanism and a Gaussian perturbation strategy. A bidirectional interaction mechanism is triggered at fixed intervals to achieve information sharing and enhanced population diversity between the two subgroups; Perform post-processing operations to obtain oil spill detection results.

[0007] Furthermore, the data preprocessing employs a contrast-limited adaptive histogram equalization algorithm to enhance edge details of the oil spill area and suppress excessive amplification of background noise.

[0008] Furthermore, the fitness value is determined according to the fitness function. calculate, in, The basic weights for entropy components; , is the normalized entropy score; As a low-threshold dynamic boosting factor, when Effective at time ; The basic weights for the inter-class variance components; Normalized inter-class variance score; The basic weights for edge consistency components; This is the normalized edge consistency score; Penalty for the percentage of dark areas; An absolute penalty is imposed for values ​​below the specified value. Punishment for regional balance.

[0009] Furthermore, the normalized entropy score Used to measure the uniformity of gray-level distribution within each segmented region. in, For the first Pixel percentage of each region; For the first Shannon entropy of each region; Indicates grayscale value The percentage of pixels in the entire image; The normalized entropy score The smaller the value, the more uniform the area.

[0010] Furthermore, the normalized inter-class variance score Used to measure the degree of grayscale difference between different regions after segmentation. in, The inter-class variance for a three-category classification; The global grayscale variance; For the first The average gray level of each region; The global grayscale mean; The normalized inter-class variance score The larger the value, the higher the regional differentiation.

[0011] Furthermore, the normalized edge consistency score Used to measure the degree of overlap between the segmentation boundary and the actual edge of the image. in, This is the ratio of the boundary average gradient to the global average gradient. This represents the average value of the global gradient magnitude; Indicates only in the boundary area The average gradient magnitude over the gradient; The normalized edge consistency score A higher value indicates a more accurate segmentation.

[0012] Furthermore, the dark area proportion penalty Absolute penalty for values ​​below Regional balance penalty Used to optimize the fitness function. in, For the first The percentage of pixels in each region.

[0013] Furthermore, the CEO simulates the gravitational interactions and random motions of celestial bodies in the universe, providing two update modes for each elite individual, resulting in new candidate thresholds after the update. , in, For the first An elite individual; This is the current globally optimal solution; Another elite individual selected randomly; This represents the current iteration number; This represents the maximum number of iterations. Represents Lévy's flight step size vector; , is a random number; , which is the upper bound of the threshold; , which is the lower bound of the threshold; when At that time, individuals are updated under the dual attraction of the global optimal solution and random elite individuals, realizing multi-center collaborative search; when At that time, the body performs a Lévy flight random walk, jumping out of local optima by long-distance jumps, and the search range gradually shrinks as the number of iterations increases.

[0014] Furthermore, the RCOA simulation of the spiral movement pattern of a reindeer herd around its leader guides ordinary individuals to converge towards elite individuals. in, Shrinkage factor; The spiral angle decreases in range with iteration; , is a standard normally distributed random vector; when At that time, individuals move in a spiral motion centered on random elite individuals to achieve a precise local search; when At this time, individuals perform random walks guided by the global optimal solution, enhancing the diversity of the population.

[0015] Furthermore, retain The results of the interval are processed and post-processing operations such as small-area noise filtering, hole filling, and shielding of ship interference areas are performed to obtain the oil spill detection results.

[0016] Compared with existing technologies, the advantages of this invention lie in its use of a hierarchical parallel algorithm that integrates Cosmic Evolution Optimization (CEO) and Reindeer Cyclone Optimization Algorithm (RCOA) to achieve dual-threshold segmentation. The population is divided into an elite subgroup and a general subgroup based on fitness values. The elite subgroup uses the CEO algorithm to perform a large-scale search across the entire solution space through a multi-center gravity mechanism and a Lévy flight strategy. The general subgroup uses the RCOA algorithm to perform a fine-grained search around elite individuals through a spiral following mechanism and a Gaussian perturbation strategy. A bidirectional interaction mechanism is triggered at fixed intervals of iterations, enabling information sharing and enhanced population diversity between the two subgroups. The fitness function is constructed as a weighted function that integrates regional statistical features and edge structural features, and a low-threshold dynamic boosting factor and multiple constraint penalty terms are introduced to constrain the search space, avoiding convergence to a mathematically optimal but poorly segmented solution. This improves the efficiency and accuracy of oil spill monitoring.

[0017] Furthermore, this study introduces a low-threshold dynamic boosting factor. .when hour, The weights of inter-class variance and edge consistency are dynamically increased by up to 20%, forcing the algorithm to prioritize higher thresholds that can identify target areas, thereby further improving the efficiency and accuracy of oil film detection. Attached Figure Description

[0018] Figure 1 The image is the original image before data preprocessing in this embodiment of the invention; Figure 2 This is an image showing the data preprocessing result of an embodiment of the present invention; Figure 3 This is the initial classification result image of an embodiment of the present invention; Figure 4 This is a post-processed image of the oil spill detection results according to an embodiment of the present invention; Figure 5 This is an image showing the oil spill detection results according to an embodiment of the present invention. Detailed Implementation

[0019] 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.

[0020] 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.

[0021] This embodiment of the oil film detection method based on CEO (Cosmic Evolution Optimization) and RCOA (Reindeer Cyclone Optimization Algorithm) includes: Data preprocessing; The population is divided into elite and ordinary subpopulations based on fitness values. The elite subgroup uses the CEO algorithm to perform a large-scale search in the entire solution space through a multi-center gravity mechanism and a Lévy flight strategy. The ordinary subgroup uses the RCOA algorithm to perform a fine search around the elite individuals through a spiral following mechanism and a Gaussian perturbation strategy. A bidirectional interaction mechanism is triggered at fixed intervals to achieve information sharing and enhanced population diversity between the two subgroups; Perform post-processing operations to obtain oil spill detection results.

[0022] Specifically, the data preprocessing employs a Contrast-Limited Adaptive Histogram Equalization (CLAHE) algorithm to enhance edge details of the oil spill area and suppress excessive amplification of background noise. The original image is shown below. Figure 1 As shown, the preprocessing results are as follows: Figure 2 As shown.

[0023] This invention employs a hierarchical parallel algorithm that integrates Cosmic Evolution Optimization (CEO) and Reindeer Cyclone Optimization Algorithm (RCOA) to achieve dual-threshold segmentation. The population is divided into an elite subgroup and a general subgroup based on fitness values. The elite subgroup uses the CEO algorithm to conduct a large-scale search across the entire solution space through a multi-center gravity mechanism and a Lévy flight strategy. The general subgroup uses the RCOA algorithm to conduct a fine-grained search around elite individuals through a spiral following mechanism and a Gaussian perturbation strategy. A bidirectional interaction mechanism is triggered at fixed intervals of iterations to achieve information sharing and enhanced population diversity between the two subgroups. The fitness function is constructed as a weighted function that integrates regional statistical features and edge structural features, and a low-threshold dynamic boosting factor and multiple constraint penalty terms are introduced to constrain the search space, avoiding convergence to a mathematically optimal but poorly segmented solution.

[0024] The constraint penalty items include dark area percentage penalty, below-average absolute penalty, and area balance penalty.

[0025] Specifically, this study used CLAHE to enhance the image, and the contrast limit threshold was set to 0.01.

[0026] This invention uses a threshold Divide the image into ( The classification results are as follows: Figure 3 As shown, a minimum distance constraint is also set. avoid Too close. To improve the threshold shift problem caused by large areas of low threshold in the image, this study introduces a low threshold dynamic boosting factor. .when hour, The weights of inter-class variance and edge consistency are dynamically increased by up to 20%, forcing the algorithm to prioritize higher thresholds that can identify target regions. Fitness function This serves as a standard for evaluating the quality of thresholds.

[0027] The fitness value is determined according to the fitness function. calculate, in, As the basic weight for the entropy component, this study recommends a value of 0.3; , is the normalized entropy score; As a low-threshold dynamic boosting factor, when Effective at time ; For the basic weights of the inter-class variance components, this study recommends a value of 0.6; Normalized inter-class variance score; For the basic weight of the edge consistency component, this study recommends a value of 0.3; This is the normalized edge consistency score; Penalty for the percentage of dark areas; An absolute penalty is imposed for values ​​below the specified value. Punishment for regional balance.

[0028] Specifically, this invention quantifies the segmentation results from three dimensions: region uniformity, inter-class discriminability, and boundary clarity. The normalized entropy score... Used to measure the uniformity of gray-level distribution within each segmented region. in, For the first Pixel percentage of each region; For the first Shannon entropy of each region; Indicates grayscale value The percentage of pixels in the entire image; The normalized entropy score The smaller the value, the more uniform the area.

[0029] Specifically, the normalized inter-class variance score Used to measure the degree of grayscale difference between different regions after segmentation. in, The inter-class variance for a three-category classification; The global grayscale variance; For the first The average gray level of each region; The global grayscale mean; The normalized inter-class variance score The larger the value, the higher the regional differentiation.

[0030] Specifically, the normalized edge consistency score Used to measure the degree of overlap between the segmentation boundary and the actual edge of the image. in, This is the ratio of the boundary average gradient to the global average gradient. This represents the average value of the global gradient magnitude; Indicates only in the boundary area The average gradient magnitude over the gradient; The normalized edge consistency score A higher value indicates a more accurate segmentation.

[0031] Specifically, to avoid problems such as excessively large dark areas, excessively low thresholds, and severely unbalanced region proportions during segmentation, constraint penalty terms are set. Optimize the fitness function. The dark area percentage penalty... Absolute penalty for values ​​below Regional balance penalty , in, For the first The percentage of pixels in each region.

[0032] Specifically, the initialization sets a population size. The population is divided into an elite subpopulation (top 30%) and a normal subpopulation (bottom 70%) based on fitness, with a maximum number of iterations. The elite subgroup of the CEO performs a wide-ranging search using a multi-center gravity mechanism and a Lévy flight strategy, avoiding getting trapped in local optima. The ordinary subgroup of the RCOA performs a fine-grained search around the elite individuals using a spiral following mechanism and a Gaussian perturbation strategy, quickly converging to the vicinity of the optimal solution.

[0033] The CEO simulation model explores the gravitational interactions and random motions of celestial bodies in the universe, providing two update modes for each elite individual, resulting in new candidate thresholds after the update. , in, For the first An elite individual; This is the current globally optimal solution; Another elite individual selected randomly; This represents the current iteration number; This represents the maximum number of iterations. Represents Lévy's flight step size vector; , is a random number; , which is the upper bound of the threshold; , which is the lower bound of the threshold; when At that time, individuals are updated under the dual attraction of the global optimal solution and random elite individuals, realizing multi-center collaborative search; when At that time, the body performs a Lévy flight random walk, jumping out of local optima by long-distance jumps, and the search range gradually shrinks as the number of iterations increases.

[0034] Specifically, the RCOA simulates the spiral movement pattern of a caribou herd around its leader, guiding ordinary individuals to converge towards elite individuals. in, Shrinkage factor; The spiral angle decreases in range with iteration; , is a standard normally distributed random vector; when At that time, individuals move in a spiral motion centered on random elite individuals to achieve a precise local search; when At this time, individuals perform random walks guided by the global optimal solution, enhancing the diversity of the population.

[0035] Specifically, the two-way interaction mechanism, triggered every 10 iterations, aims to achieve information sharing between the two subgroups and prevent stagnation in the elite subgroup and blind searching in the ordinary subgroup. This mechanism consists of two parts: an elite promotion mechanism and an ordinary subgroup perturbation mechanism. For the elite promotion mechanism, the best individual in the ordinary subgroup is compared with the worst individual in the elite subgroup. If the best individual in the ordinary subgroup has a higher fitness, the worst individual in the elite subgroup is replaced, and the elite subgroup is re-sorted by fitness. For the ordinary subgroup perturbation mechanism, for each ordinary individual, a Gaussian perturbation with a standard deviation of 5 is applied around a random elite individual with a 50% probability, reinitializing some individuals and enhancing population diversity.

[0036] Specifically, post-processing operations, retain The results of the interval are then subjected to small-area noise filtering. Post-treatment operations, including filling holes and shielding areas affected by ship interference, are performed. The results of the post-treatment oil spill detection are as follows: Figure 4 As shown, the final oil spill detection results are as follows: Figure 5 As shown.

[0037] 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.

[0038] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for detecting oil film based on CEO and RCOA, characterized in that, include: Data preprocessing: The contrast-limited adaptive histogram equalization algorithm is used to preprocess the original image to enhance the edge details of the oil spill area and suppress excessive amplification of background noise, thus forming the corresponding preprocessed image. The population was divided into an elite subpopulation and a normal subpopulation based on fitness values. The elite subgroup uses the CEO algorithm to perform a large-scale search in the entire solution space through a multi-center gravity mechanism and a Lévy flight strategy. The ordinary subgroup uses the RCOA algorithm to perform a fine search around the elite individuals through a spiral following mechanism and a Gaussian perturbation strategy. A bidirectional interaction mechanism is triggered at fixed intervals to achieve information sharing and enhanced population diversity between the two subgroups; Perform post-processing operations to obtain oil spill detection results; Wherein, the fitness value is determined according to the fitness function. calculate, in, The basic weights for entropy components; , is the normalized entropy score; As a low-threshold dynamic boosting factor, when Effective at time ; The basic weights for the inter-class variance components; Normalized inter-class variance score; The basic weights for edge consistency components; This is the normalized edge consistency score; Penalty for the percentage of dark areas; Absolute penalty for low threshold; Punishment for regional balance; The CEO simulation model explores the gravitational interactions and random motions of celestial bodies in the universe, providing two update modes for each elite individual, resulting in new candidate thresholds after the update. , in, For the first An elite individual; This is the current globally optimal solution; Another elite individual selected randomly; This represents the current iteration number; This represents the maximum number of iterations. Represents Lévy's flight step size vector; , is a random number; , which is the upper bound of the threshold; , which is the lower bound of the threshold; when At that time, individuals are updated under the dual attraction of the global optimal solution and random elite individuals, realizing multi-center collaborative search; when At that time, the individual performs a Lévy flight random walk, jumping out of the local optimum by long-distance jumps, and the search range gradually shrinks as the number of iterations increases; The RCOA simulation model depicts the spiral movement pattern of a reindeer herd around its leader, guiding ordinary individuals to converge towards elite individuals. in, Shrinkage factor; The spiral angle decreases in range with iteration; , is a standard normally distributed random vector; when At that time, individuals move in a spiral motion centered on random elite individuals to achieve a precise local search; when At this time, individuals perform random walks guided by the global optimal solution, enhancing the diversity of the population.

2. The oil film detection method based on CEO and RCOA according to claim 1, characterized in that, The normalized entropy score Used to measure the uniformity of gray-level distribution within each segmented region. in, For the first Pixel percentage of each region; For the first Shannon entropy of each region; Indicates grayscale value The percentage of pixels in the entire image; The normalized entropy score The smaller the value, the more uniform the area.

3. The oil film detection method based on CEO and RCOA according to claim 1, characterized in that, The normalized inter-class variance score Used to measure the degree of grayscale difference between different regions after segmentation. in, The inter-class variance for a three-category classification; The global grayscale variance; For the first The average gray level of each region; The global grayscale mean; The normalized inter-class variance score The larger the value, the higher the regional differentiation.

4. The oil film detection method based on CEO and RCOA according to claim 1, characterized in that, The normalized edge consistency score Used to measure the degree of overlap between the segmentation boundary and the actual edge of the image. in, This is the ratio of the boundary average gradient to the global average gradient. This represents the average value of the global gradient magnitude; Indicates only in the boundary area The average gradient magnitude over the gradient; The normalized edge consistency score A higher value indicates a more accurate segmentation.

5. The oil film detection method based on CEO and RCOA according to claim 1, characterized in that, The penalty for the proportion of dark areas Low threshold absolute penalty Regional balance penalty Used to optimize the fitness function. in, For the first The percentage of pixels in each region.

6. The oil film detection method based on CEO and RCOA according to claim 1, characterized in that, reserve The results of the interval are processed and post-processing operations such as small-area noise filtering, hole filling, and shielding of ship interference areas are performed to obtain the oil spill detection results.

Images