CN-121661042-B - Oil film detection method and system based on region sensing and structure guiding

CN121661042BCN 121661042 BCN121661042 BCN 121661042BCN-121661042-B

Abstract

The invention provides an oil film detection method and system based on region sensing and structure guiding, relates to the technical field of oil film detection, and aims to solve the problem of accurate segmentation of oil film regions in low-contrast and high-noise radar images. The method comprises the steps of preprocessing acquired original radar image data, detecting feature points of the preprocessed image by using an improved KAZE feature detector, converting detected discrete feature points into a continuous probability density field, extracting a high-density area from the continuous probability density field to serve as an interested area to generate an interested image, acquiring an optimal segmentation threshold by using a neural gas network-guided sea-city building particle swarm optimization algorithm, segmenting the interested image by using the optimal segmentation threshold, and carrying out post-processing on segmented images based on the optimal segmentation threshold to acquire a final oil film detection result. The invention is especially suitable for low-contrast and high-noise scenes, and can provide an effective technical tool for real-time monitoring of marine oil spill.

Inventors

GUO ZEKUN
XU JIN
JIA BAOZHU
Qian Sihan
HUANG YUANYUAN

Assignees

广东海洋大学深圳研究院
广东海洋大学

Dates

Publication Date: 20260508
Application Date: 20260204

Claims (8)

1. The oil film detection method based on region sensing and structure guiding is characterized by comprising the following steps of: s1, preprocessing collected original radar image data; S2, performing feature point detection on the preprocessed image by using an improved KAZE feature detector, wherein the improved KAZE feature detector improves a diffusion mode into a local consistency-based region diffusion mode, and the local consistency-based region diffusion mode comprises a conduction function for controlling diffusion intensity Is introduced into a local gray level consistency factor Then a new conduction function is constructed The method comprises the following steps: ; in the formula, Representing a Gaussian smoothed image At the pixel position The gradient vector at which the gradient is located, Representing a scale control factor; Representing a gradient threshold; , The enhancement factor is represented by a coefficient of enhancement, Represents the adjustment coefficient of the device, Representing local gray standard deviation; S3, converting the detected discrete feature points into a continuous probability density field, extracting a high-density region from the continuous probability density field as an interested region, and further generating an interested image, wherein the specific steps include calculating a density value at any pixel position on an image plane according to a feature point set obtained by detection by using a kernel density estimation method, further obtaining the continuous probability density field, normalizing the continuous probability density field to obtain a normalized density field, determining an optimal segmentation threshold value of the interested region and a background region by adopting an Otsu maximum inter-class variance method based on the normalized density field, generating a binary mask based on the optimal segmentation threshold value, and multiplying the binary mask with the preprocessed image pixel by pixel to generate the interested image; S4, acquiring an optimal segmentation threshold value by utilizing a mirage particle swarm optimization algorithm guided by a neural gas network, and segmenting the interest image by utilizing the optimal segmentation threshold value to acquire a segmented image based on the optimal segmentation threshold value; based on the characteristic anchor point set, adopting a particle swarm optimization algorithm fused with a mirage guidance mechanism to search an optimal segmentation threshold value by minimizing a fusion fitness function; s5, carrying out post-processing on the segmented image based on the optimal segmentation threshold value to obtain a final oil film detection result.
2. The method for detecting the oil film based on the region sensing and the structure guiding as claimed in claim 1, wherein the preprocessing in the step S1 comprises the following steps of coordinate conversion, co-channel interference detection and suppression, speckle noise filtering, gray level correction and contrast enhancement in sequence.
3. The oil film detection method based on region sensing and structure guidance according to claim 1, wherein the mirage guidance mechanism in S4 is implemented by the following speed update equation: ; in the formula, A velocity vector representing particle i at the t+1st iteration; a velocity vector representing particle i at the t-th iteration; Representing inertial weights for controlling the motion inertia of the particles i; , , all represent learning factors for adjusting individual cognitive coefficients, social cognitive coefficients and mirage guidance coefficients, respectively; , , all represent random numbers uniformly distributed in the interval [0,1 ]; a mirage point assigned to particle i; representing individual historical optimal positions of particle i; and G (t) represents the global optimal position of the population.
4. A method for detecting an oil film based on area sensing and structure guiding as defined in claim 3, wherein said inertial weight With iteration number The formula of the self-adaptive adjustment is as follows: ; in the formula, 、 Respectively the maximum value and the minimum value of the inertia weight; Representing a maximum number of iterations; learning factor , , With iteration number The formulas of the self-adaptive adjustment are respectively as follows: ; ; ; in the formula, Representing the rate of increase of the individual cognitive coefficients; an initial value representing an individual cognitive coefficient; representing the rate of increase of social cognition coefficients; an initial value representing a social cognition coefficient; a reference value representing mirage guidance coefficients; And The training final value and the reference standard value respectively represent the neural gas network learning rate; indicating the rate of decrease.
5. The method for detecting an oil film based on area sensing and structure guidance according to claim 4, wherein the spot of the mirage is The generation process of (1) comprises: calculating policy selection probabilities The formula is: ; in the formula, And The strategy selection probabilities at the initial stage and the final stage of iteration are respectively represented; Generating random decision variables xi-Uniform (0, 1); representing random variables subject to uniform distribution; if the random decision variable xi is less than or equal to Generating the mirage point according to elite disturbance guiding strategy or NGN guiding strategy Otherwise, generating the mirage point according to diversity exploration guiding strategy Wherein, generating the mirage point according to elite disturbance guiding strategy The formula of (2) is: ; in the formula, Representing randomly selected elite particle positions; representing a random disturbance that is indicative of the random disturbance, Wherein Disturbance amplitude representing attenuation: , representing the initial value of the disturbance amplitude, And Representing the upper and lower bounds of the search space respectively, A random vector representing each dimension subject to a standard normal distribution, for providing a random direction of disturbance, Representing element-by-element multiplication; generating mirage points according to NGN guidance strategy The formula of (2) is: ; in the formula, Representing feature anchors randomly selected from the feature anchor set; Generating mirage points according to diversity exploration guidance strategy The method comprises the following steps of randomly and uniformly sampling a point in a global search space, wherein the formula is as follows: 。
6. the oil film detection method based on region sensing and structure guiding as claimed in claim 3, wherein said fusion fitness function is : ; In the formula, Representing a candidate segmentation threshold vector to be optimized, Representing a segmentation quality term for measuring the uniformity of the gray scale of pixels within each segment region defined by a segmentation threshold vector, , For the class k pixel scale, For the gray variance of the kth class of pixels, Representing a preset threshold number; And The weight coefficients are used for balancing the consistency of the segmentation quality and the characteristics; A reward term representing neural gas network characteristic consistency, defined as: ; Wherein, the Representing adaptive matching sensitivity parameters; Representing candidate segmentation threshold vectors The d-th segmentation threshold in (2) Distance to the nearest neural gas network extracted feature anchor point w.
7. The oil film detection method based on region sensing and structure guiding according to claim 1 is characterized by comprising the specific steps of performing binarization processing on a segmented image based on an optimal segmentation threshold to extract oil film candidate regions, removing background interference by using an ROI mask to keep targets in the oil film candidate regions, extracting edges of the ROI mask by a Sobel operator, expanding an edge range by adopting morphological expansion operation, deleting the expanded edges from the oil film candidate regions, and obtaining a final oil film detection result.
8. An oil film detection system based on region sensing and structure guiding, which is characterized in that the system is used for realizing the oil film detection method based on region sensing and structure guiding according to any one of claims 1-7, and comprises the following steps: A data preprocessing module configured to preprocess the acquired raw radar image data; a feature point detection module configured to perform feature point detection on the preprocessed image using a modified KAZE feature detector that modifies a diffusion mode to a local consistency-based region diffusion mode; The interest region extraction module is configured to convert the detected discrete feature points into a continuous probability density field, extract a high-density region from the continuous probability density field as an interest region and further generate an interest image; the optimal segmentation module is configured to acquire an optimal segmentation threshold value by utilizing a mirage particle swarm optimization algorithm guided by the neural gas network, segment the interest image by utilizing the optimal segmentation threshold value and acquire a segmented image based on the optimal segmentation threshold value; and the oil film detection module is configured to post-process the segmented image based on the optimal segmentation threshold value to obtain a final oil film detection result.

Description

Oil film detection method and system based on region sensing and structure guiding Technical Field The invention relates to the technical field of oil film detection, in particular to an oil film detection method and system based on region sensing and structure guiding. Background The marine oil spill accident forms a serious threat to the ecological environment and the economic development, and it is important to rapidly and accurately detect the oil film on the sea surface. The ship-borne radar is an important means for carrying out large-scale all-weather sea surface monitoring, and the collected radar image can reflect sea surface oil film information. However, radar images generally have problems of low contrast, high noise, blurred edges of the target, and the like, so that a stable and accurate oil film region is difficult to obtain by the traditional image segmentation and target detection method. When processing such complex images, the methods based on threshold segmentation, edge detection or traditional feature extraction in the prior art are often sensitive to noise, and the segmentation results are prone to region discontinuity, boundary misalignment or over/under segmentation phenomena. Although some improved population intelligent optimization algorithms (such as particle swarm optimization) are introduced to find a better segmentation threshold, when low-contrast, multi-modal gray scale distribution is processed, local optimization is still easily trapped, and the prior information of the structure of the image is not fully utilized, so that the detection capability of a weak characteristic oil film region is limited. Therefore, there is a need for an oil film detection method that can effectively suppress noise, adaptively enhance target features, and perform intelligent guidance using image essential structures, so as to improve the accuracy, robustness, and automation level of oil film detection under complex sea conditions. Disclosure of Invention In view of the above problems, the invention provides an oil film detection method and system based on region sensing and structure guiding, which aim to solve the problem of accurate segmentation of an oil film region in a low-contrast and high-noise radar image. According to an aspect of the present invention, an oil film detection method based on region sensing and structure guiding is provided, the method comprising: s1, preprocessing collected original radar image data; S2, detecting characteristic points of the preprocessed image by using an improved KAZE characteristic detector, wherein the improved KAZE characteristic detector improves a diffusion mode into a local consistency-based region diffusion mode; s3, converting the detected discrete feature points into a continuous probability density field, extracting a high-density region from the continuous probability density field as an interested region, and further generating an interested image; S4, acquiring an optimal segmentation threshold value by utilizing a mirage particle swarm optimization algorithm guided by a neural gas network, and segmenting the interest image by utilizing the optimal segmentation threshold value to acquire a segmented image based on the optimal segmentation threshold value; s5, carrying out post-processing on the segmented image based on the optimal segmentation threshold value to obtain a final oil film detection result. Further, the preprocessing in S1 comprises coordinate conversion, co-channel interference detection and suppression, speckle noise filtering, gray correction and contrast enhancement. Further, the local uniformity-based region diffusion mode in S2 includes a conduction function for controlling diffusion intensityIs introduced into a local gray level consistency factorThen a new conduction function is constructedThe method comprises the following steps: ; in the formula, Representing a Gaussian smoothed imageAt the pixel positionThe gradient vector at which the gradient is located,Representing a scale control factor; Representing a gradient threshold; , The enhancement factor is represented by a coefficient of enhancement, Represents the adjustment coefficient of the device,Representing the local gray standard deviation. Further, the specific step of S3 comprises the steps of calculating density values at any pixel position on an image plane according to a detected feature point set by using a kernel density estimation method to obtain a continuous probability density field, normalizing the continuous probability density field to obtain a normalized density field, determining an optimal segmentation threshold value of an interested region and a background region by adopting an Otsu maximum inter-class variance method based on the normalized density field, generating a binary mask based on the optimal segmentation threshold value, and multiplying the binary mask with preprocessed images pixel by pixel to generate the interested image. Furth