Oil film detection method and system based on hierarchical self-organizing network and scale self-adaption

Abstract

The invention provides an oil film detection method and system based on a hierarchical self-organizing network and scale adaptation, and relates to the technical field of oil film detection. According to the technical key points, the method comprises the following steps: preprocessing an obtained original radar image; performing oil film region identification on the preprocessed radar image by using a trained oil film identification model based on the improved growth layered neural gas network to obtain an oil film candidate region; and segmenting the oil film candidate region by using an improved multi-scale threshold segmentation algorithm to obtain a final oil film detection result. According to the method, a local feature enhancement mechanism, dynamic level adjustment, adaptive error attenuation, sparse constraint optimization and a real-time adaptive learning mechanism are introduced, so that the adaptivity, robustness and calculation efficiency of the detection model are remarkably improved. The method can be integrated in a shipborne radar system, is high in calculation efficiency, can meet the real-time processing requirements of shipborne equipment, and is suitable for practical application scenes such as marine oil spill monitoring and emergency response.

Description

Technical Field

[0001] The present invention relates to the technical field of oil film detection, and particularly relates to an oil film detection method and system based on a hierarchical self-organizing network and scale adaptability. Background Art

[0002] At present, remote sensing technology has been widely used in the field of oil spill monitoring, mainly using optical remote sensing and microwave radar as monitoring means. Optical remote sensing can capture information in multiple bands, and its spectral data is very rich, which helps to accurately identify oil films. However, the monitoring effect of optical remote sensing is easily affected by lighting conditions and meteorological conditions, and it is difficult to meet the requirement of real-time monitoring of oil film diffusion. Different from optical remote sensing, microwave radar emits electromagnetic waves to the sea surface and receives the scattered signals reflected back, and then generates a sea surface echo image. In these images, compared with the sea water area not covered by the oil film, the backscattering signal of the sea surface area with an oil spill is relatively weak, so dark spots will be formed on the sea surface echo image of the microwave radar. Since microwave radar has the ability of all-weather, real-time and efficient monitoring, and can overcome the influence of bad sea conditions to a certain extent, it has become one of the most effective oil spill monitoring means at present.

[0003] Traditional methods rely on threshold segmentation and are difficult to deal with problems such as complex backgrounds, uneven lighting, and variable oil film shapes. In the prior art, although the algorithms based on supervised learning can classify data, they need to label the data during the learning process; in addition, although the traditional threshold segmentation method can perform local optimization segmentation, it is insufficient in dealing with multi-scale features. Summary of the Invention

[0004] In view of the above problems, the present invention proposes an oil film detection method and system based on a hierarchical self-organizing network and scale adaptability, which improves the oil film detection accuracy by integrating a dynamic self-organizing network and a multi-scale threshold technology.

[0005] According to one aspect of the present invention, an oil film detection method based on a hierarchical self-organizing network and scale adaptability is proposed, and the method includes:

[0006] Preprocess the acquired original radar image;

[0007] Use the trained oil film recognition model based on the improved growing hierarchical neural gas network to identify the oil film area of the preprocessed radar image and obtain the oil film candidate area;

[0008] Use the improved multi-scale threshold segmentation algorithm to segment the oil film candidate area and obtain the final oil film detection result.

[0009] Further, the preprocessing includes: performing horizontal operator convolution on the original radar image, detecting vertical direction noise using the Otsu algorithm, noise smoothing, speckle removal, and contrast enhancement.

[0010] Further, the training process of the oil film recognition model based on the improved growing hierarchical neural gas network includes:

[0011] Initializing neurons, where the reference vector of each neuron is a sample point randomly selected from the training set S; initializing the error variable to zero and the age of the edge to zero;

[0012] In the growing stage, each iteration randomly selects an input sample point, i.e., a certain neuron, calculates the Euclidean distance between this neuron and all other neurons, and finds the neuron closest to this neuron and the second closest neuron , increasing the ages of all connections starting from neuron q; updating the error variable of the closest neuron , the reference vectors of neuron and its neighbors;

[0013] If there is no connection between neuron q and neuron s, create a new connection, otherwise reset the age of this connection to zero; delete connections whose ages exceed the threshold, remove neurons with no connections; and insert new neurons;

[0014] If the number of neurons in a certain graph is less than or equal to 2, prune it, otherwise generate its corresponding subgraph; recursively generate subgraphs for each neuron and use the receptive fields of multiple neurons as the training set for training until the preset maximum hierarchical depth is reached;

[0015] Calculate the average quantization error of the backup graph and the current graph respectively, and compare the average quantization errors. If the stop condition is met, stop growing and enter the convergence stage;

[0016] In the convergence stage, the error variables of all neurons are decayed through an adaptive decay mechanism;

[0017] Output the clustering results of each layer, extract the oil film candidate regions, and the training is completed.

[0018] Further, update the reference vector according to the following formula :

[0019]

[0020] In the formula, w i (t) represents the reference vector of neuron i at time t; η(t) represents the learning rate that decreases with time; α represents the global feature weight, and its value range is [0, 1]; β represents the local feature weight, and its value range is [0, 1]; A sample point randomly selected from the training set S of the local features.

[0021] Furthermore, update the reference vector according to the following formula :[[]]

[0022]

[0023] In the formula, min represents minimization; W represents the set of reference vectors of all neurons; S is the training set; represents the best matching unit of the input sample point x(t) in the set of reference vectors of all neurons; represents the weight parameter of the sparse constraint; represents the neuron reference vector of the L1 norm; N represents the total number of neurons.

[0024] Furthermore, the formula for attenuating the error variables of all neurons through the adaptive decay mechanism is:

[0025]

[0026] In the formula, represents the error variable of neuron i at time t; γ(t) represents the dynamic decay factor,

[0027]

[0028] Among them, γ0 is the initial decay constant, and its value range is [0, 1]; Δ Data Complexity represents the change rate of the input data complexity, which is obtained by calculating the difference in information entropy of the input data at adjacent time steps; represents the decay adjustment parameter, which is used to control the sensitivity of the dynamic decay factor changing with the data complexity.

[0029] Furthermore, the segmentation of the oil film candidate region using the improved multi-scale threshold segmentation algorithm includes:

[0030] For the oil film candidate region, calculate the local threshold at different neighborhood scales:

[0031]

[0032] In the formula, represents the local threshold of the pixel point at the kth scale; represents the local mean; represents the local standard deviation; is the sensitivity coefficient; represents the pixel point Local feature saliency

[0033]

[0034] Among them represents the gradient vector of pixel point ; max represents taking the maximum value

[0035] Take the average of the local thresholds at each scale to generate the final threshold:

[0036] In the formula represents the final threshold after fusion; M represents the total number of scales; w k represents the spatial weight of pixel point at the k-th scale

[0037] According to the final threshold Segment the oil film candidate region, extract the oil film boundary, and obtain the final oil film detection result

[0038] Furthermore, the sensitivity coefficient in the improved multi-scale threshold segmentation algorithm has the following calculation formula:

[0039]

[0040] In the formula, k0 represents the initial sensitivity coefficient represents the sensitivity adjustment parameter, which is used to control the degree of change of sensitivity with local complexity represents the entropy value of the neighborhood gray value of pixel point ; MaxComplexit represents the maximum value of the entropy values of all neighborhood gray values

[0041] Furthermore, the spatial weight w of pixel point k at the k-th scale in the improved multi-scale threshold segmentation algorithm has the following calculation formula:

[0042]

[0043] In the formula represents the central position of the k-th scale, representing the core position of this scale in the image space represents the distance between pixel point and the center of the k-th scale; σ k represents the standard deviation related to the scale

[0044] According to another aspect of the present invention, an oil film detection system based on a hierarchical self-organizing network and scale adaptation is proposed. The system includes:

[0045] An image preprocessing module configured to preprocess the acquired original radar image;

[0046] An oil film region recognition module configured to use a trained oil film recognition model based on an improved growing hierarchical neural gas network to recognize the oil film region in the preprocessed radar image and obtain oil film candidate regions;

[0047] An oil film segmentation module configured to use an improved multi-scale threshold segmentation algorithm to segment the oil film candidate regions and obtain the final oil film detection result.

[0048] The beneficial technical effects of the present invention are:

[0049] Determining the oil film region is an important link in oil spill identification. Supplementary feature information during the experiment helps improve the identification accuracy. The present invention proposes an oil film detection method and system based on a hierarchical self-organizing network and scale adaptation. It performs unsupervised clustering of the oil film region through an improved growing hierarchical neural gas network and uses an improved multi-scale adaptive threshold segmentation algorithm to accurately extract the oil film. The present invention makes innovative improvements to the traditional growing hierarchical neural gas network model and multi-scale adaptive threshold segmentation algorithm, combines their unique advantages for efficient oil film identification; by introducing a local feature enhancement mechanism, dynamic hierarchical adjustment, adaptive error decay, sparse constraint optimization, and real-time adaptive learning mechanism, the self-adaptability, robustness, and computational efficiency of the model are significantly improved; in terms of multi-scale adaptive threshold segmentation processing, by introducing local feature saliency analysis, multi-scale threshold fusion based on spatial weights, adaptive sensitivity adjustment, and threshold optimization based on saliency analysis, the ability to accurately extract the oil film boundary is further improved, and the adaptability to complex backgrounds and changing environments is enhanced. Through these improvements and optimizations, the parameters of the model can be dynamically optimized according to the threshold processing results, thus significantly improving the oil film detection accuracy in complex scenarios.

[0050] The present invention can be integrated into a shipborne radar system, has high computational efficiency, can meet the real-time processing requirements of shipborne equipment, and is applicable to practical application scenarios such as offshore oil spill monitoring and emergency response; the present invention improves the accuracy of the shipborne radar system in detecting marine oil spill oil films under complex sea conditions, provides efficient technical support for marine environmental protection and oil spill accident emergency handling, and has significant social and economic benefits. Brief Description of the Drawings

[0051] By reading the following detailed description with reference to the accompanying drawings, the above and other objects, features, and advantages of the exemplary embodiments of the present invention will become readily understandable. In the drawings, several embodiments of the present invention are shown by way of example and not limitation, wherein:

[0052] Figure 1 is a flowchart of an oil film detection method based on a hierarchical self-organizing network and scale adaptability according to an embodiment of the present invention;

[0053] Figure 2 is another schematic flowchart of an oil film detection method based on a hierarchical self-organizing network and scale adaptability according to an embodiment of the present invention;

[0054] Figure 3 is an example diagram of a radar image in an embodiment of the present invention;

[0055] Figure 4 is an example diagram of a preprocessed radar image in an embodiment of the present invention;

[0056] Figure 5 is an example diagram of an oil film candidate region image obtained by using an oil film recognition model based on an improved growing hierarchical neural gas network in an embodiment of the present invention;

[0057] Figure 6 is an example diagram of an image of a final oil film detection result obtained by using an improved multi-scale threshold segmentation algorithm in an embodiment of the present invention;

[0058] Figure 7 is an example diagram of an oil film detection result image after removing spots in an embodiment of the present invention;

[0059] Figure 8 is a schematic structural diagram of an oil film detection system based on a hierarchical self-organizing network and scale adaptability according to an embodiment of the present invention. Detailed implementation manners

[0060] The principles and spirit of the present invention will be described below with reference to several exemplary embodiments. It should be understood that these embodiments are provided only to enable those skilled in the art to better understand and then implement the present invention, and do not limit the scope of the present invention in any way. On the contrary, these embodiments are provided to make the present disclosure more thorough and complete, and to fully convey the scope of the present disclosure to those skilled in the art.

[0061] Those skilled in the art know that the embodiments of the present invention can be implemented as a system, device, equipment, method, or computer program product. Therefore, the present disclosure can be specifically implemented in the following forms: completely hardware, completely software (including firmware, resident software, microcode, etc.), or a combination of hardware and software. In this article, it should be understood that any number of elements in the drawings is for illustration rather than limitation, and any naming is only for distinction and does not have any limiting meaning.

[0062] The present invention proposes an oil film detection method and system based on a hierarchical self-organizing network and scale adaptability, providing an intelligent processing method for oil spill identification for shipborne microwave radar oil spill monitoring tasks.

[0063] An embodiment of the present invention proposes an oil film detection method based on a hierarchical self-organizing network and scale adaptability, as Figures 1 - 2 shown, the method includes:

[0064] S1. Preprocess the acquired original radar image;

[0065] S2. Use the trained oil film recognition model based on the improved growing hierarchical neural gas network to identify the oil film area in the preprocessed radar image and obtain the oil film candidate area;

[0066] S3. Use the improved multi-scale threshold segmentation algorithm to segment the oil film candidate area and obtain the final oil film detection result.

[0067] The method starts with S1. In S1, the acquired original radar image is preprocessed.

[0068] According to the embodiment of the present invention, the acquired original radar image is as Figure 3 shown. The following preprocessing is performed on the original radar image: perform horizontal operator convolution on the image, use the Otsu algorithm for vertical direction noise detection, and then perform operations such as noise smoothing, speckle removal, and contrast enhancement to convert the original radar image into high-quality output data, providing a clear and low-noise basis for subsequent oil film detection. It should be noted that the preprocessing process is the existing preprocessing process for radar images, so it will not be elaborated in this embodiment. The preprocessed image is as Figure 4 shown.

[0069] Then in S2, use the trained oil film recognition model based on the improved growing hierarchical neural gas network to identify the oil film area in the preprocessed radar image and obtain the oil film candidate area.

[0070] According to the embodiments of the present invention, the training process of the original Growing Hierarchical Neural Gas Network (GHNG) is as follows.

[0071] 1) Initialization phase: Starting from the root graph, initialize two neurons. The reference vector of each neuron is randomly drawn from the training set S, and the positions of the initial neurons are determined by the input data distribution; initialize the error variable to zero, and initialize the age of the edges to zero.

[0072] When inputting a grayscale image, the grayscale value of each pixel point is used as an input sample, and the model completes the task by learning the distribution and characteristics of these grayscale values; all pixel points in an image can form a training set or sample set, and each pixel point is used as an independent sample.

[0073] 2) In the growth phase, the GHNG model dynamically adds neurons through an error-driven learning mechanism; the specific steps are as follows: Calculate the Euclidean distance between the input sample point (i.e., a certain neuron) and all other neurons, find the neuron nearest to it and the second-nearest neuron

[0074]

[0075]

[0076] Update the error variable of the nearest neuron , and add to the error variable e q of q:

[0077]

[0078] Update the reference vectors of neuron and its neighbors:

[0079]

[0080] where is the learning rate, defined as: ; respectively control the learning rates of q and its neighbors.

[0081] 3) If there is no connection between q and s, create a new connection; otherwise, reset the age of this connection to zero. Delete the connections whose age exceeds the threshold, and remove the neurons without connections.

[0082] 4) Insert new neurons (every λ steps): Select the neuron r with the largest error and its neighbor z, and insert a new neuron k with the prototype . Connect k to r and z, and decay e r and e z (i.e., the error variables of neuron r and its neighbor neuron z).

[0083] 5) Hierarchical training and recursive generation of subgraphs: The GHNG model generates a hierarchical structure through recursive training. If the number of neurons H in a graph ≤ 2, prune it; otherwise, generate a corresponding subgraph for each neuron i, and the receptive field of each neuron is defined as: where, is the training set of the current graph; for each neuron , recursively generate a subgraph and use as the training set for training until the preset maximum hierarchical depth is reached.

[0084] 6) Growth control mechanism: The GHNG model controls the growth of the graph through the parameter ; the specific steps are as follows: Calculate the mean quantization error (MQE) of the backup graph and the current graph:

[0085] where, and are the neuron sets of the backup graph and the current graph respectively; compare the mean quantization errors. If the following conditions are met, stop growing and enter the convergence stage:

[0086] 7) Error decay and convergence stage: In the convergence stage, the error variables of all neurons are decayed by the constant :

[0087]

[0088] where, is the decay constant.

[0089] 8) Transform invariance of the hierarchical structure: The hierarchical structure of the GHNG model is invariant to uniform scaling and rigid transformation of the input data. The transformed neuron prototypes and error variables are:

[0090]

[0091] where, is the scaling factor; is an orthogonal matrix; is the translation vector.

[0092] 9) Output the clustering results of each layer and extract the oil film candidate regions.

[0093] In this embodiment, the original growing hierarchical neural gas network is improved as follows.

[0094] 1) Optimize the update process of the reference vectors to enable the model to better adapt to the oil film detection requirements under complex sea conditions and improve the detection accuracy and robustness. Two different improvement strategies are adopted.

[0095] 11) Introduce a local feature enhancement mechanism. When updating the reference vectors, combine local feature information to enhance the robustness of the model to complex backgrounds and noises. The updated formula for the reference vectors is:

[0096]

[0097] where w i (t) represents the reference vector of neuron i at time t; η(t) represents the learning rate that decreases with time, , is the initial learning rate, λ is the decay coefficient, which can be adjusted according to the actual situation; α represents the global feature weight, with a value range of [0, 1], used to balance the role of global features in the reference vector update; β represents the local feature weight, with a value range of [0, 1], used to control the influence degree of local features on the reference vector update; represents the local features of the input image x(t). The local features include edge features, texture features, etc., which can be extracted through edge detection operators (such as the Sobel operator) or texture analysis methods (such as the gray-level co-occurrence matrix) to capture local detail information such as edges and textures of the input samples.

[0098] 12) Introduce a sparse constraint in the reference vector update stage to reduce redundant neurons and lower the computational complexity. The updated formula for the reference vectors is:

[0099]

[0100] where min represents minimization; W represents the set of reference vectors of all neurons; S is the training set; represents the best matching unit of the input sample x(t) in the set of reference vectors of all neurons, where represents the best matching unit of the input sample x(t), that is, the neuron q with the closest Euclidean distance to the sample x(t); represents the weight parameter of the sparse constraint. The larger its value, the higher the requirement for the sparsity of the neuron reference vectors. The appropriate value can be determined through experimental methods such as cross-validation; represents the neuron reference vector The L1 norm is used to measure the sparsity of vectors, which encourages the reference vectors of some neurons to approach zero, thereby reducing redundant neurons; N represents the total number of neurons.

[0101] On the one hand, the local feature enhancement mechanism or sparse constraint optimization ensures that the model can accurately identify the oil film, and on the other hand, it improves the operating efficiency of the model; during the training and optimization process of the model, the local feature enhancement mechanism can make the reference vectors better adapt to the local feature distribution of the input data, while the sparse constraint optimization helps to screen out the neurons and features that are more important for oil film detection, making the optimization direction of the model more focused on the features related to the oil film, thereby improving the optimization efficiency and effect.

[0102] 2) In the convergence stage, the error variable is decayed through a constant and improved to: the error variable is decayed through an adaptive mechanism to enhance the convergence speed and stability of the model. The formula for updating the error variable is:

[0103]

[0104] In the formula, represents the error variable of neuron i at time t; γ(t) represents the dynamic decay factor,

[0105]

[0106] where γ0 is the initial decay constant, and its value range is [0, 1], which can be set according to experimental experience; ΔData Complexity represents the change rate of the complexity of the input data, which is obtained by calculating the difference in information entropy of the input data at adjacent time steps to reflect the dynamic change of data complexity. Specifically, the probability distribution of the current image pixel values is calculated at each iteration step to obtain the information entropy, and ΔData Complexity is the difference in information entropy between the current step and the previous step; represents the decay adjustment parameter, which is used to control the sensitivity of the dynamic decay factor changing with the data complexity, and this parameter can be adjusted through experiments to achieve the best convergence effect. Figure 5 Shows the oil film candidate region image obtained by using the improved growing hierarchical neural gas network.

[0107] Then execute S3, and use the improved multi-scale threshold segmentation algorithm to segment the oil film candidate region to obtain the final oil film detection result.

[0108] According to the embodiments of the present invention, the multi-scale threshold segmentation algorithm is used to further refine the oil film boundary, eliminate noise and improve the detection accuracy. The specific steps are as follows:

[0109] a. Multi-scale threshold calculation: For the oil film candidate region image output by the GHNG model, calculate the local threshold T at different neighborhood scales:

[0110] Where, and are the mean and standard deviation of the scale neighborhood, is the sensitivity coefficient. Different sensitivity parameters have different effects. Set the sensitivity parameter according to needs. Calculate the thresholds at different scales through this formula combined with local statistical characteristics.

[0111] b. Threshold fusion: Take the average of the thresholds at each scale to generate the final threshold:

[0112] c. Binary segmentation: According to Binaryize the image to accurately extract the oil film boundary and eliminate background noise.

[0113] In this embodiment, the original multi-scale threshold segmentation algorithm is improved as follows.

[0114] 1) Introduce local feature saliency analysis and dynamically adjust the threshold to adapt to the local feature intensity. Specifically as follows:

[0115]

[0116] In the formula, represents the local threshold of the pixel point at the k-th scale; represents the local mean; represents the local standard deviation; is the sensitivity coefficient; represents the local feature saliency of the pixel point ,

[0117]

[0118] Where, represents the gradient vector of the pixel point . By calculating the ratio of the gradient vector norm to the maximum norm, measure the significance degree of the local feature of this pixel. The larger the value, the more significant the local feature; max represents taking the maximum value.

[0119] 2) The sensitivity coefficient k is adaptively and dynamically adjusted according to the local feature complexity. The formula is:

[0120]

[0121] In the formula, k0 represents the initial sensitivity coefficient; represents the sensitivity adjustment parameter, which is used to control the degree of change of the sensitivity with the local complexity; Represents the entropy value of the neighborhood gray values of a pixel point. The local complexity is measured by calculating the entropy value of the neighborhood gray levels. The larger the entropy value, the higher the local complexity; Max Complexity represents the maximum value of the entropy values of all neighborhood gray levels and is used for normalizing the local complexity. The entropy value of the neighborhood gray values of is used to measure the local complexity by calculating the entropy value of the neighborhood gray levels. The larger the entropy value, the higher the local complexity; Max Complexity represents the maximum value of the entropy values of all neighborhood gray levels and is used for normalizing the local complexity.

[0122] 3) Introduce spatial weights in multi-scale threshold fusion, considering the spatial distribution of pixels. The formula is:

[0123]

[0124] In the formula, Represents the final threshold after fusion, which is obtained by weighted summation of the thresholds at M scales. The threshold at each scale contributes to the final threshold according to the degree of spatial association of the pixel point with this scale (i.e., the spatial weight ); M represents the total number of scales; w k Represents the spatial weight of the pixel point at the k-th scale and is defined as:

[0125]

[0126] In the formula, Represents the central position of the k-th scale, representing the core position of this scale in the image space; represents the distance between the pixel point and the center of the k-th scale; σ k Represents the standard deviation related to the scale, controlling the rate at which the weight changes with distance. The smaller σ k , the faster the weight decays with increasing distance. The exponential term in the numerator represents the degree of spatial association of the pixel point with the threshold of the k-th scale. The closer the distance, the larger the exponential value, indicating that the influence weight of this scale on the pixel point is greater; in the denominator part, the exponential terms of all M scales are summed to normalize the exponential value of the numerator, making range from 0 to 1, ensuring that the sum of the weights of each scale is 1, thus reasonably distributing the contribution of each scale in the fusion.

[0127] Introducing spatial weights in multi-scale threshold fusion makes the fusion process pay more attention to the scale thresholds that are spatially close to the current pixel. Through the assignment of spatial weights, the accuracy of the threshold after fusion in depicting the oil film boundary is improved. Figure 6 Shows the oil film detection results obtained using the improved multi-scale threshold segmentation algorithm.

[0128] Further, optionally, small spots and spots in the ship wake area can be removed by using the image meta - area threshold to obtain the final oil film. This process is also the prior art. An example of the final oil film detection result image is shown as Figure 7 shown.

[0129] Another embodiment of the present invention proposes an oil film detection system based on a hierarchical self - organizing network and scale adaptability, as shown in Figure 8 shown. The system includes:

[0130] An image pre - processing module 810, which is configured to pre - process the acquired original radar image;

[0131] An oil film area recognition module 820, which is configured to use the trained oil film recognition model based on the improved growing hierarchical neural gas network to recognize the oil film area in the pre - processed radar image and obtain the oil film candidate area;

[0132] An oil film segmentation module 830, which is configured to use the improved multi - scale threshold segmentation algorithm to segment the oil film candidate area and obtain the final oil film detection result.

[0133] It should be noted that the functions of the oil film detection system based on a hierarchical self - organizing network and scale adaptability described in this embodiment can be illustrated by the aforementioned oil film detection method based on a hierarchical self - organizing network and scale adaptability. For the parts not detailed in the system embodiment, refer to the above method embodiment.

[0134] It should be noted that although several units, modules or sub - modules are mentioned in the above detailed description, this division is merely exemplary and not mandatory. In fact, according to the embodiments of the present invention, the features and functions of two or more of the above - described modules can be embodied in one module. Conversely, the features and functions of one module described above can be further divided and embodied by multiple modules.

[0135] In addition, although the operations of the method of the present invention are described in a specific order in the drawings, this does not require or imply that these operations must be performed in that specific order, or that all the shown operations must be performed to achieve the desired result. Additionally or alternatively, some steps may be omitted, multiple steps may be combined into one step for execution, and / or one step may be decomposed into multiple steps for execution.

[0136] Although the spirit and principles of the present invention have been described with reference to several specific embodiments, it should be understood that the present invention is not limited to the disclosed specific embodiments, and the division of each aspect does not mean that the features in these aspects cannot be combined for benefit. This division is only for the convenience of expression. The present invention aims to cover various modifications and equivalent arrangements included within the spirit and scope of the appended claims.

Claims

1. An oil film detection method based on a hierarchical self-organizing network and scale adaptability, characterized in that Including: Preprocessing the acquired original radar image; Using the trained oil film recognition model based on the improved growing hierarchical neural gas network to identify the oil film area in the preprocessed radar image and obtain the oil film candidate area; Using the improved multi-scale threshold segmentation algorithm to segment the oil film candidate area and obtain the final oil film detection result.

2. The oil film detection method based on a hierarchical self-organizing network and scale adaptation according to claim 1, wherein The preprocessing includes: performing horizontal operator convolution on the original radar image, detecting vertical direction noise using the Otsu algorithm, noise smoothing, speckle removal, and contrast enhancement.

3. The oil film detection method based on a hierarchical self-organizing network and scale adaptability according to claim 1, characterized in that, The training process of the oil film recognition model based on the improved growing hierarchical neural gas network includes: Initialize the neurons, and the reference vector of each neuron is a sample point randomly selected from the training set S; initialize the error variable to zero and the age of the edge to zero; In the growth phase, each iteration randomly selects an input sample point, i.e., a certain neuron, calculates the Euclidean distance between this neuron and all other neurons, and finds the neuron that is closest to this neuron and the second-closest neuron , increases the ages of all connections starting from neuron q; updates the error variable of the closest neuron , the reference vectors of the neuron and its neighbors; If there is no connection between neuron q and neuron s, create a new connection; otherwise, reset the connection age to zero; delete the connections whose age exceeds the threshold, remove the unconnected neurons; and insert new neurons; If the number of neurons in a certain graph is less than or equal to 2, prune it; otherwise, generate its corresponding subgraph; recursively generate subgraphs for each neuron and use the receptive fields of multiple neurons as the training set for training until the preset maximum hierarchical depth is reached; Calculate the average quantization error of the backup graph and the current graph respectively, and compare the average quantization errors. If the stop condition is met, stop growing and enter the convergence stage; In the convergence stage, the error variables of all neurons are attenuated through the adaptive attenuation mechanism; Output the clustering results of each layer, extract the oil film candidate area, and the training is completed.

4. The oil film detection method based on a hierarchical self-organizing network and scale adaptability according to claim 3, characterized in that, Update the reference vector according to the following formula :[[]]END]] ； In the formula, represents the reference vector of neuron i at time t; η(t) represents the learning rate that decreases over time; α represents the global feature weight, with a value range of [0, 1]; β represents the local feature weight, with a value range of [0, 1]; represents the sample point randomly selected from the training set S and its local feature.

5. The oil film detection method based on a hierarchical self-organizing network and scale adaption according to claim 3, characterized in that Update the reference vector according to the following formula : ； where, min represents minimization; W represents the set of reference vectors of all neurons; S is the training set; represents the input sample point in the set of reference vectors of all neurons of the best matching unit; represents the weight parameter of the sparse constraint; represents the neuron reference vector of the L1 norm; N represents the total number of neurons.

6. The oil film detection method based on a hierarchical self-organizing network and scale adaptation according to claim 4 or 5, characterized in that, The formula for attenuating the error variables of all neurons through the adaptive attenuation mechanism is: ； wherein, represents the error variable of neuron i at time t; γ(t) represents the dynamic decay factor, ； Among them, is the initial decay constant, and its value range is [0, 1]; Δ Data Complexity represents the change rate of the input data complexity, which is obtained by calculating the difference in information entropy of the input data at adjacent time steps; represents the decay adjustment parameter, which is used to control the sensitivity of the dynamic decay factor to the change in data complexity.

7. The oil film detection method based on a hierarchical self-organizing network and scale adaptation according to claim 1 or 2, characterized in that, The using of the improved multi-scale threshold segmentation algorithm to segment the oil film candidate area includes: For the oil film candidate area, calculate the local threshold at different neighborhood scales: ； In the formula, represents the local threshold of the pixel point at the k-th scale ; represents the local mean; represents the local standard deviation; is the sensitivity coefficient; represents the local feature significance of the pixel point ; ； Among them, represents the gradient vector of the pixel point ; max represents taking the maximum value; Average the local thresholds at each scale to generate the final threshold: ; In the formula, represents the final threshold after fusion; M represents the total number of scales; w k represents the spatial weight of the pixel point at the k-th scale; According to the final threshold Segment the oil film candidate region, extract the oil film boundary, and obtain the final oil film detection result.

8. The oil film detection method based on a hierarchical self-organizing network and scale adaption according to claim 7, wherein The sensitivity coefficient in the improved multi-scale threshold segmentation algorithm has the following calculation formula: ； wherein, k0 represents an initial sensitivity coefficient; represents a sensitivity adjustment parameter for controlling the degree of change of sensitivity with local complexity; represents a pixel entropy value of the neighborhood gray value; Max Complexit represents the maximum value of the entropy values of all neighborhood gray values.

9. The oil film detection method based on a hierarchical self-organizing network and scale adaptability according to claim 8, characterized in that In the improved multi-scale threshold segmentation algorithm, the spatial weight w of the pixel point at the k-th scale is k The calculation formula is as follows: ； In the formula, represents the central position of the k-th scale, representing the core position of the scale in the image space; represents the pixel point and the distance from the center of the k-th scale; σ k represents the standard deviation related to the scale.

10. An oil film detection system based on a hierarchical self-organizing network and scale adaptation, characterized in that, Including: An image preprocessing module configured to preprocess the acquired original radar image; An oil film area recognition module configured to use the trained oil film recognition model based on the improved growing hierarchical neural gas network to identify the oil film area in the preprocessed radar image and obtain the oil film candidate area; An oil film segmentation module configured to use the improved multi-scale threshold segmentation algorithm to segment the oil film candidate area and obtain the final oil film detection result.

Images