CN111860534A - SAR image oil spill detection method based on image significance analysis - Google Patents
SAR image oil spill detection method based on image significance analysis Download PDFInfo
- Publication number
- CN111860534A CN111860534A CN202010534064.5A CN202010534064A CN111860534A CN 111860534 A CN111860534 A CN 111860534A CN 202010534064 A CN202010534064 A CN 202010534064A CN 111860534 A CN111860534 A CN 111860534A
- Authority
- CN
- China
- Prior art keywords
- image
- pixel
- value
- saliency
- oil
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 239000003305 oil spill Substances 0.000 title claims abstract description 54
- 238000001514 detection method Methods 0.000 title claims abstract description 51
- 238000004458 analytical method Methods 0.000 title abstract description 5
- 238000000034 method Methods 0.000 claims abstract description 61
- 230000003044 adaptive effect Effects 0.000 claims abstract description 11
- 238000001914 filtration Methods 0.000 claims description 17
- 238000006243 chemical reaction Methods 0.000 claims description 4
- 239000000126 substance Substances 0.000 claims description 4
- 238000000605 extraction Methods 0.000 abstract description 14
- 230000003993 interaction Effects 0.000 abstract description 2
- 239000003921 oil Substances 0.000 description 69
- 230000000052 comparative effect Effects 0.000 description 14
- 230000011218 segmentation Effects 0.000 description 13
- 230000006870 function Effects 0.000 description 7
- 230000010287 polarization Effects 0.000 description 6
- 238000012545 processing Methods 0.000 description 6
- 239000000284 extract Substances 0.000 description 5
- 238000013528 artificial neural network Methods 0.000 description 4
- 230000002146 bilateral effect Effects 0.000 description 4
- 238000013461 design Methods 0.000 description 3
- 238000003709 image segmentation Methods 0.000 description 3
- 238000012544 monitoring process Methods 0.000 description 3
- 238000011160 research Methods 0.000 description 3
- 238000012360 testing method Methods 0.000 description 3
- 239000010779 crude oil Substances 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 230000018109 developmental process Effects 0.000 description 2
- 230000004927 fusion Effects 0.000 description 2
- 239000011159 matrix material Substances 0.000 description 2
- 238000010606 normalization Methods 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 238000012549 training Methods 0.000 description 2
- 101100423891 Caenorhabditis elegans qars-1 gene Proteins 0.000 description 1
- 230000009471 action Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 239000003638 chemical reducing agent Substances 0.000 description 1
- 238000013527 convolutional neural network Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000003708 edge detection Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 230000000877 morphologic effect Effects 0.000 description 1
- 239000003129 oil well Substances 0.000 description 1
- 230000002093 peripheral effect Effects 0.000 description 1
- 239000003208 petroleum Substances 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 238000004321 preservation Methods 0.000 description 1
- 238000003672 processing method Methods 0.000 description 1
- 229910052702 rhenium Inorganic materials 0.000 description 1
- WUAPFZMCVAUBPE-UHFFFAOYSA-N rhenium atom Chemical compound [Re] WUAPFZMCVAUBPE-UHFFFAOYSA-N 0.000 description 1
- 238000012216 screening Methods 0.000 description 1
- 239000013535 sea water Substances 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/46—Descriptors for shape, contour or point-related descriptors, e.g. scale invariant feature transform [SIFT] or bags of words [BoW]; Salient regional features
- G06V10/462—Salient features, e.g. scale invariant feature transforms [SIFT]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/20—Image enhancement or restoration using local operators
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/90—Dynamic range modification of images or parts thereof
- G06T5/94—Dynamic range modification of images or parts thereof based on local image properties, e.g. for local contrast enhancement
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0004—Industrial image inspection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/136—Segmentation; Edge detection involving thresholding
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/194—Segmentation; Edge detection involving foreground-background segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/90—Determination of colour characteristics
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Multimedia (AREA)
- Quality & Reliability (AREA)
- Image Analysis (AREA)
Abstract
An oil spilling region detection method without manual interaction. The method is used for detecting the SAR image oil spilling region based on image significance analysis and a self-adaptive iteration threshold method. In the scheme, an image significance detection method is introduced into SAR oil spill detection, and then the accurate extraction of the oil spill area is realized by using the size relation between the adaptive iteration threshold and the significance.
Description
Technical Field
The invention relates to the technical field of image processing and remote sensing images, in particular to an SAR image oil spilling detection method based on image significance analysis.
Background
In today's society, petroleum remains a very important resource. With the increasing scarcity of land resources and the rapid increase in human demand for energy, the marine oil industry and the marine oil transportation industry are developing vigorously. The offshore oil spill is loss of oil in different degrees in the process of offshore exploitation or transportation, and mainly comprises oil well crude oil leakage caused in the process of offshore oil exploration and development, leakage caused by loading and unloading of offshore oil pipelines or oil tankers, crude oil leakage caused by accidents such as collision, overturning, grounding and the like of ships, even oil spill caused by natural disasters and the like. These accidents all pollute the marine ecological environment to varying degrees and also cause a great deal of economic loss.
In order to reduce the occurrence of oil spill accidents, the monitoring and detection of offshore oil spill needs to be enhanced. Synthetic Aperture Radar (SAR) has the advantages of all-time, all-weather, large-range, high precision and the like, and is the most effective means for monitoring oil spill at present. In the detection of oil spill at sea, the oil spill area is an important parameter for measuring the oil spill accident. Therefore, in the SAR marine oil spill image, how to accurately extract the boundary of the oil spill area is a key problem for calculating the oil spill area. In the SAR marine oil spill image, the oil spill area usually has dark color, and is obviously different from the peripheral non-oil spill area. Based on this, the oil spill area can be extracted by an image processing method.
Aiming at the oil spill detection of the SAR image, a great deal of significant research is carried out by many scholars at home and abroad. The Liuwei and the like combine the FCM and the DRLSE model to be applied to SAR oil spill extraction, and the effectiveness of the method is verified; in addition, the complete oil spill SAR map is established by analyzing the complete polarization SAR oil spill image according to the statistical characteristics, textural characteristics and polarization characteristics of oil spill and by feature extraction and selection, and the oil spill is detected and extracted by introducing a multi-core learning method of prior labels.
Wehne and rhenium and the like adopt a single threshold segmentation method, a maximum entropy segmentation method and an unsupervised classification method to carry out target detection on SAR oil spilling images, roughly divide the images into a foreground region and a background region, manually select partial oil spilling regions and non-oil spilling regions as regions of interest, respectively count texture features commonly used by the SAR images on the regions of interest, and carry out classification based on a BP neural network by combining different target detection results and original images to obtain a good effect.
The yaoqi and the like perform an ocean oil spill extraction experiment on the sea surface near the zhujiang opening stretcher island by utilizing the SAR image, and analyze different applicability of the artificial neural network method and the Markov chain method in the aspect of oil spill monitoring. The Seattle and the like research an SAR image oil spill detection method based on a deep learning method and realize an oil spill detection algorithm based on the combination of a gray level co-occurrence matrix and a convolutional neural network.
Guo Yue et al fuse gray level co-occurrence matrix and Tamura characteristics, directly extract the characteristics of the SAR original image, and then use the classification method of the deep belief network to classify and identify the 3 types of samples of oil films, oil-like films and seawater, so as to obtain better identification accuracy.
Zhenghong Lei and the like introduce polarization characteristics and single scattering relative difference as oil spill detection characteristic parameters, and an oil spill detection algorithm based on the polarization characteristics and an artificial neural network is developed.
CN201210024538.7 discloses a sea surface oil spill image segmentation method of polarized SAR data fusion. The method comprises the steps of firstly constructing an active contour energy functional based on a maximum posterior probability criterion of a segmentation region, expressing the distribution of the segmentation region into a Gibbs prior probability model, then embedding the active contour model into a high-dimensional level set function, obtaining a development equation by using a Euler-Lagrange formula, wherein the model comprises a boundary length term weighted by CFAR edge detection and a fusion data statistical distance term.
CN201110277737.4 discloses a method and device for detecting sea surface oil spill based on SAR images. The method comprises the following steps: converting the SAR image into a binary image by performing threshold segmentation on the gray level of a pixel point in the SAR image on the sea surface; for a neighborhood with a preset size of each pixel point in the binary image, determining whether the number and/or proportion of the pixel points with the pixel value of 1 in the neighborhood is larger than a preset value or not; identifying whether the quantity and/or proportion of the pixel points with the pixel value of 1 is larger than the neighborhood of the preset value or not; and (4) taking the boundary of the image formed by all the identified neighborhoods as an initial zero level set for detecting level set segmentation so as to detect sea surface oil spill.
CN201310382104.9 discloses an oil spill detection method for a complex SAR image scene, which comprises the following steps: firstly, reading in a detection image; secondly, carrying out image segmentation on the detected image, extracting dark spots, and processing to obtain a bright sea dark spot image; thirdly, in the bright sea dark spot image, setting the part except the dark sea area to be 0 or 1 to obtain a dark sea image, carrying out image segmentation on the dark sea image, extracting the dark spot, and processing to obtain a dark sea dark spot image; fourthly, adding the dark spots in the dark sea dark spot image into the bright sea dark spot image to obtain a full dark spot image, and removing false dark spots to obtain a partial dark spot image; and fifthly, setting a reference gray level according to the partial dark spot image, performing omission retrieval on the de-noised detection image to obtain omitted dark spots, and adding the omitted dark spots into the partial dark spot image to obtain a final oil spilling dark spot image. The method is suitable for extracting the oil spilling dark spots in the complex scene.
CN201610715334.6 discloses a sea surface oil spill detection method based on C-band polarized SAR images, which comprises the following steps: step 1: preprocessing a radar image; step 2: constructing a high-dimensional polarization feature set; and step 3: constructing a linear Laplace mapping image dimensionality reducer and carrying out k-mean classification; and 4, step 4: the sea surface wind field data assists oil spill detection; and 5: and (6) evaluating the precision.
CN201810373026.9 discloses a precise oil spill detection method based on CFAR, which first performs coarse detection on a to-be-detected region by using global CFAR, and extracts a suspected oil film region to obtain an oil film target binary reference map. And then screening the oil film target binary reference image by using methods such as morphological filtering and the like, and eliminating interference of the clutter. And finally, carrying out fine detection on the filtered image by adopting a self-adaptive window CFAR algorithm to finally obtain an oil film region.
CN201811463066.9 discloses a level set SAR oil spill extraction method based on bilateral filtering. According to the method, a bilateral filter is used for filtering an oil spill SAR image; a DRLSE model energy function based on bilateral filtering is constructed
For DRLSE model energy function based on bilateral filtering
Energy minimization is performed; by using
The energy minimization equation F extracts SAR oil spill information.
In view of the disadvantages that the level set method needs to be manually initialized, the neural network method needs to provide manually calibrated samples in advance, training is needed to generate a recognition model, and the like, how to design a higher-precision oil spilling region extraction method on the basis of the saliency image is an important research direction.
Disclosure of Invention
Aiming at the problems in the prior art, the invention provides an oil spilling region detection method without manual interaction. The method is used for detecting the SAR image oil spilling region based on image significance analysis and a self-adaptive iteration threshold method. In the scheme, an image significance detection method is introduced into SAR oil spill detection, and then the accurate extraction of the oil spill area is realized by using the size relation between the adaptive iteration threshold and the significance.
The invention provides a method for detecting an oil spilling area of an SAR image, which comprises the following steps:
step 1) obtaining a saliency image:
carrying out image significance detection to generate a normalized significance image;
step 2) calculating an adaptive iteration threshold T:
calculating a self-adaptive iteration threshold T by using a self-adaptive iteration threshold algorithm;
step 3), judging an oil spilling area:
after the SAR oil spilling image is subjected to significance detection, in the significance image, the pixel value of each pixel represents the significance degree of the pixel; and (3) solving a threshold value T of the whole saliency image by using a self-adaptive iterative threshold value method, wherein if the value of a certain pixel in the saliency image is greater than T, the saliency image belongs to an oil spilling region, and otherwise, the saliency image belongs to a non-oil spilling region.
Further, the step 1) of acquiring the saliency image is as follows:
1.1) converting an original image from an RGB space to a Lab space;
1.2) carrying out Gaussian filtering on the Lab space image;
1.3) taking the average values LM, AM and BM of the images of three channels L, a and b after conversion respectively; calculating Euclidean distances of the mean value images of the three channels and the images after Gaussian filtering respectively and summing;
and 1.4) normalizing the saliency image by using the maximum value and the minimum value in the saliency image to generate a normalized saliency image.
Further, the step 1) of acquiring the saliency image is as follows:
1.1) firstly converting the RGB color space of the original image I into XYZ space by means of formula (1), and then converting the XYZ space into Lab space by means of formula (2) to obtain a Lab space image I corresponding to the original imageLab；
Wherein the content of the first and second substances,
1.2) image I for Lab spaceLabThree components I ofL、IaAnd IbFiltering all with a 3 x 3 gaussian convolution kernel to obtain a filtered image IGLab；
Wherein the 3 × 3 Gaussian convolution kernel is
1.3) respectively obtaining ILabThree components I of the middle Lab spaceL、IaAnd IbThen using equation (3) to find the three means and IGLabThe Euclidean distance of the three components in the image to obtain a saliency image SM;
SM(x,y)＝(LIGLab(x,y)-LM)2+(aIGLab(x,y)-AM)2+(bIGLab(x,y)-BM)2(3)
wherein (x, y) is pixel coordinate, LIGLab(x,y)、aIGLab(x, y) and b IGLab(x, y) are each IGLabThe value of the three components at coordinates (x, y).
1.4) maximum Max in saliency map SM using equation (4)SMAnd minimum MinSMAnd carrying out normalization processing on the SM to obtain a normalized saliency image NSM.
NSM(x,y)＝(SM(x,y)-MinSM)/(MaxSM-MinSM) (4)。
Further, the step 2) of the adaptive iterative threshold method comprises the following steps:
2.1) calculating the maximum gray value and the minimum gray value of the image, respectively recording as Zmax and Zmin, and making the initial threshold value
T0＝(Zmax+Zmin)/2 (5)
2.2) dividing the image into foreground and background according to the threshold Tk, and respectively calculating the average gray value Z of the foreground and the backgroundoAnd Zb；
2.3) find the new threshold:
Tk+1＝(Zo+Zb)/2 (6)
2.4) if Tk＝Tk+1The iteration is stopped and the final threshold is obtained, otherwise go to step 2.2).
Further, the step 3) of judging the oil spilling region is as follows:
and (4) judging the relation between the significance value of each pixel in the normalized significance image NSM and T by using the formula (7) to determine whether the pixel is an oil spill pixel.
Wherein R isoil(x, y) represents whether the pixel at the coordinate (x, y) is the pixel of the oil spilling region, if 1, the pixel is regarded as the pixel of the oil spilling region, and if 0, the pixel is regarded as the pixel of the non-oil spilling region.
The method combines an image significance detection method and a self-adaptive iteration threshold value method and introduces the method into the extraction of the oil spilling region of the SAR image.
Firstly, processing an SAR image by using an image saliency detection method to generate a saliency image, so that an oil spilling region in the image is more obvious visually; then, an optimal segmentation threshold value is calculated by using a self-adaptive iteration threshold value method; and finally, extracting an accurate oil spilling region according to the relation between the significance value of each pixel and the threshold value on the basis of the significance image.
Experimental results show that the scheme designed by the invention can effectively extract the oil spilling region in the SAR image and has higher recall rate and accuracy.
The method designed by the invention has the following advantages:
1) the image significance detection method can improve the contrast between the oil spilling region and the non-oil spilling region, so that the oil spilling region and the non-oil spilling region are more visually distinguished, and a foundation is laid for the next extraction of the oil spilling region;
2) the threshold most suitable for a certain image can be efficiently obtained by adopting a self-adaptive iteration threshold method, so that an oil spilling area and a non-oil spilling area are better distinguished;
3) the method can provide more efficient detection efficiency for large-batch detection of the offshore oil spilling areas and provide reference for manual interpretation.
Drawings
The above and/or additional aspects and advantages of the present invention will become apparent and readily appreciated from the following description of the embodiments, taken in conjunction with the accompanying drawings of which:
fig. 1 is an SAR oil spill image example 1.
FIG. 2 shows the results of the significance test in example 1.
Fig. 3(a) shows the detection result of the oil spilling region in example 1.
Fig. 3(b) is the oil spilling region detection result of comparative example 1.
Fig. 3(c) is the oil spilling region detection result of comparative example 2.
FIG. 3(d) is the result of the manual interpretation of example 1.
Fig. 4 is an SAR oil spill image example 2.
Fig. 5 is the significance test result of example 2.
Fig. 6(a) shows the detection result of the oil spilling region in example 2.
Fig. 6(b) is the oil spilling region detection result of comparative example 3.
Fig. 6(c) is the oil spilling region detection result of comparative example 4.
FIG. 6(d) is the result of manual interpretation of example 2.
Detailed Description
The experimental data used herein are derived from the NOWPAP (Northwest Pacific Action plan) database.
The geographical range covered by NOWPAP is marine environment and coastal area from east longitude about 121 degrees to east longitude 143 degrees, north latitude about 33 degrees to north latitude 52 degrees.
The SAR image is a C-band SAR image from two satellites of ERS-1(European Remote Sensing Satellite-1) and ESR-2(European Remote Sensing Satellite-2) by adopting a VV polarization mode, as shown in FIG. 1 and FIG. 4.
Embodiment 1 is a method for detecting an oil spill area in an SAR image, where the SAR image is shown in fig. 1 and includes:
step 1) obtaining a saliency image:
1.1) converting an original image from an RGB space to a Lab space;
firstly, converting the RGB color space of the original image I into an XYZ space by means of formula (1), and then converting the XYZ space into a Lab space by means of formula (2) to obtain a Lab space image I corresponding to the original image Lab；
Wherein the content of the first and second substances,
1.2) carrying out Gaussian filtering on the Lab space image;
image I for Lab spaceLabThree components I ofL、IaAnd IbFiltering all with a 3 x 3 gaussian convolution kernel to obtain a filtered image IGLab；
Wherein the 3 × 3 Gaussian convolution kernel is
1.3) taking the average values LM, AM and BM of the images of three channels L, a and b after conversion respectively; respectively calculating Euclidean distances of the mean value images of the three channels and the images after Gaussian filtering;
respectively obtain ILabThree components I of the middle Lab spaceL、IaAnd IbThen using equation (3) to find the three means and IGLabThe Euclidean distance of the three components in the image to obtain a saliency image SM;
SM(x,y)＝(LIGLab(x,y)-LM)2+(aIGLab(x,y)-AM)2+(bIGLab(x,y)-BM)2(3)
wherein (x, y) is pixel coordinate, LIGLab(x,y)、aIGLab(x, y) and bIGLab(x, y) are each IGLabThe value of the three components of (a) at coordinates (x, y);
1.4) normalizing the saliency image by using the maximum value and the minimum value in the saliency image to generate a normalized saliency image;
maximum value Max in saliency map SM using equation (4)SMAnd minimum MinSMNormalizing the SM to obtain a normalized saliency image NSM;
NSM(x,y)＝(SM(x,y)-MinSM)/(MaxSM-MinSM) (4)
carrying out image significance detection to generate a normalized significance image;
step 2) calculating an adaptive iteration threshold T:
calculating a self-adaptive iteration threshold T by using a self-adaptive iteration threshold algorithm;
2.1) determining the maximum and minimum gray-scale values of the image, respectively denoted as ZmaxAnd ZminLet an initial threshold value
T0＝(Zmax+Zmin)/2 (5)
Wherein Z ismax＝240，Zmin＝22
2.2) dividing the image into foreground and background according to the threshold Tk, and respectively calculating the average gray value Z of the foreground and the backgroundoAnd Zb；
2.3) find the new threshold:
Tk+1＝(Zo+Zb)/2 (6)
2.4) if Tk＝Tk+1Then go to restStopping iteration to obtain a final threshold, otherwise, turning to the step 2.2);
the adaptive iteration threshold T of the normalized saliency image NSM is 135.884;
step 3), judging an oil spilling area:
and (4) judging the relation between the significance value of each pixel in the normalized significance image NSM and T by using the formula (7) to determine whether the pixel is an oil spill pixel.
Wherein R isoil(x, y) represents whether the pixel at the coordinate (x, y) is the pixel of the oil spilling region, if 1, the pixel is regarded as the pixel of the oil spilling region, and if 0, the pixel is regarded as the pixel of the non-oil spilling region.
Comparative example 1
Different from example 1, the level set method was used for oil spill area detection.
The level set method employs the Demo1 program in the level set code version v0 of the lisoming design. Wherein the Gaussian fuzzy variance sigma is 1.5, the smooth dirac function parameter epsilon is 1.5, the time step length timestep is 5, the internal energy penalty parameter mu is 0.2/timestep, the weighting length coefficient lambda is 5, the weighting area coefficient alf is 10, and the iteration number is 600.
Comparative example 2
Different from the embodiment 1, the OTSU dynamic threshold method is used for detecting the oil spilling area.
The OTSU dynamic threshold method is as follows:
1) for an image I, setting T as a segmentation threshold of a foreground and a background, wherein the ratio of foreground points to image is omega0Average gray of μ0(ii) a The number of background points in the image is omega1Average gray of μ1
2) When T is traversed from the minimum gray value to the maximum gray value, the variance value is taken to be omega0×ω1×(μ0-μ1)×(μ0-μ1) Maximum time T0Is the optimal segmentation threshold.
The specific implementation method is to directly adopt MATLAB softMethod for solving segmentation threshold value T of OTSU (on-the-go) method by graythresh () function in piece0。
Embodiment 2 is a method for detecting an oil spill area in an SAR image, where the SAR image is shown in fig. 4, and includes:
step 1) obtaining a saliency image:
1.1) converting an original image from an RGB space to a Lab space;
firstly, converting the RGB color space of the original image I into an XYZ space by means of formula (1), and then converting the XYZ space into a Lab space by means of formula (2) to obtain a Lab space image I corresponding to the original imageLab；
Wherein the content of the first and second substances,
1.2) carrying out Gaussian filtering on the Lab space image;
image I for Lab spaceLabThree components I ofL、IaAnd IbFiltering all with a 3 x 3 gaussian convolution kernel to obtain a filtered image IGLab；
Wherein the 3 × 3 Gaussian convolution kernel is 
1.3) taking the average values LM, AM and BM of the images of three channels L, a and b after conversion respectively; respectively calculating Euclidean distances of the mean value images of the three channels and the images after Gaussian filtering;
respectively obtain ILabThree components I of the middle Lab spaceL、IaAnd IbThen using equation (3) to find the three means and IGLabThe euclidean distance of the three components of (a),further obtaining a significant image SM;
SM(x,y)＝(LIGLab(x,y)-LM)2+(aIGLab(x,y)-AM)2+(bIGLab(x,y)-BM)2(3)
wherein (x, y) is pixel coordinate, LIGLab(x,y)、aIGLab(x, y) and bIGLab(x, y) are each IGLabThe value of the three components of (a) at coordinates (x, y);
1.4) normalizing the saliency image by using the maximum value and the minimum value in the saliency image to generate a normalized saliency image;
maximum value Max in saliency map SM using equation (4)SMAnd minimum MinSMAnd carrying out normalization processing on the SM to obtain a normalized saliency image NSM.
NSM(x,y)＝(SM(x,y)-MinSM)/(MaxSM-MinSM) (4)
Carrying out image significance detection to generate a normalized significance image;
step 2) calculating an adaptive iteration threshold T:
calculating a self-adaptive iteration threshold T by using a self-adaptive iteration threshold algorithm;
2.1) determining the maximum and minimum gray-scale values of the image, respectively denoted as ZmaxAnd ZminMaking an initial threshold value;
T0＝(Zmax+Zmin)/2 (5)
in this embodiment, Zmax is 255, Zmin is 0;
2.2) dividing the image into foreground and background according to the threshold Tk, and respectively calculating the average gray value Z of the foreground and the background oAnd Zb；
2.3) find the new threshold:
Tk+1＝(Zo+Zb)/2 (6)
2.4) if Tk＝Tk+1Stopping iteration to obtain a final threshold, otherwise, turning to the step 2.2);
the adaptive iteration threshold T of the normalized saliency image NSM is 119.0381;
step 3), judging an oil spilling area:
3.1) solving the self-adaptive iteration threshold T of the normalized saliency image NSM by using the method in the step 2);
3.2) judging the relation between the significance value of each pixel in the normalized significance image NSM and T by using the formula (7) to determine whether the pixel is an oil spilling pixel;
wherein R isoil(x, y) represents whether the pixel at the coordinate (x, y) is the pixel of the oil spilling region, if 1, the pixel is regarded as the pixel of the oil spilling region, and if 0, the pixel is regarded as the pixel of the non-oil spilling region.
Comparative example 3
Different from example 2, the level set method was used for oil spill area detection.
The level set method employs the Demo1 program in the level set code version v0 of the lisoming design. Wherein the Gaussian fuzzy variance sigma is 1.5, the smooth dirac function parameter epsilon is 1.5, the time step length timestep is 5, the internal energy penalty parameter mu is 0.2/timestep, the weighting length coefficient lambda is 5, the weighting area coefficient alf is 10, and the iteration number is 600.
Comparative example 4
Different from the embodiment 2, the OTSU dynamic threshold method is used for detecting the oil spilling area.
The OTSU dynamic threshold method is as follows:
1) for an image I, setting T as a segmentation threshold of a foreground and a background, wherein the ratio of foreground points to image is omega0Average gray of μ0(ii) a The number of background points in the image is omega1Average gray of μ1
2) When T is traversed from the minimum gray value to the maximum gray value, the variance value is taken to be omega0×ω1×(μ0-μ1)×(μ0-μ1) Maximum time T0Is the optimal segmentation threshold.
The specific implementation method is to directly adopt graythresh () function in MATLAB software to obtain the segmentation threshold value T of the OTSU method0。
On the basis of the manually interpreted data (Ground Truth), namely manually marking the oil spilling area, the oil spilling area extraction performance of the three methods is compared. The indexes of comparison comprise recall rate and accuracy rate, and the calculation mode is as follows:
the recall rate and accuracy of the oil spilling region detection results of examples 1-2 and comparative examples 1-4 are shown in table 1.
TABLE 1 recall and accuracy of oil spill area test results of examples 1-2 and comparative examples 1-4
As shown in the table 1 below, the following examples,
the recall rate for example 1 was 81.04%, the accuracy was 85.86%;
the recall rate of comparative example 1 was 58.37%, the accuracy was 96.28%;
the recall rate of comparative example 2 was 57.12% and the accuracy was 98.03%.
The recall rate for example 2 was 81.79%, the accuracy was 94.48%;
Comparative example 3 had a recall of 81.32% and an accuracy of 90.97%;
the recall rate of comparative example 4 was 79.43% and the accuracy was 94.31%.
When the results obtained were analyzed, it was found that:
(1) oil spill area extraction results: the level set method performs poorly in edge preservation as shown by red circles in fig. 3(b) and fig. 6(b), while the OTSU method distinguishes oil spilling regions from non-oil spilling regions according to a threshold value, does not consider correlation between pixels inside the regions, and also performs poorly in partial regions as shown by green circles in fig. 3(c) and fig. 6 (c).
(2) Oil spilling region extraction accuracy: for the SAR image shown in fig. 1, although both the level set method and the OTSU method have high accuracy, their recall rate is too low; for the SAR image shown in fig. 4, the three methods all have relatively high recall rate and accuracy, but the adaptive iterative threshold method performs optimally.
(3) Degree of human involvement: the level set method needs to manually set initial profiles in advance, and the initial profiles are different, so that the extraction results of oil spilling regions are different due to different iteration times; the scheme designed by the method does not need manual participation, and does not need a calibrated sample training set.
Therefore, the recall rate and the accuracy rate of the detection result of the oil spilling region can be simultaneously higher than 80%, and the extraction precision is higher.
Claims (5)
1. A method for detecting an oil spilling region of an SAR image is characterized by comprising the following steps:
step 1) obtaining a saliency image:
carrying out image significance detection to generate a normalized significance image;
step 2) calculating an adaptive iteration threshold T:
calculating a self-adaptive iteration threshold T by using a self-adaptive iteration threshold algorithm;
step 3), judging an oil spilling area:
after the SAR oil spilling image is subjected to significance detection, in the significance image, the pixel value of each pixel represents the significance degree of the pixel; and (3) solving a threshold value T of the whole saliency image by using a self-adaptive iterative threshold value method, wherein if the value of a certain pixel in the saliency image is greater than T, the saliency image belongs to an oil spilling region, and otherwise, the saliency image belongs to a non-oil spilling region.
2. The detection method according to claim 1, characterized in that:
step 1) the step of obtaining a saliency image is as follows:
1.1) converting an original image from an RGB space to a Lab space;
1.2) carrying out Gaussian filtering on the Lab space image;
1.3) taking the average values LM, AM and BM of the images of three channels L, a and b after conversion respectively; calculating Euclidean distances of the mean value images of the three channels and the images after Gaussian filtering respectively and summing;
and 1.4) normalizing the saliency image by using the maximum value and the minimum value in the saliency image to generate a normalized saliency image.
3. The detection method according to claims 1 and 2, characterized in that:
step 1) the step of obtaining a saliency image is as follows:
1.1) firstly converting the RGB color space of the original image I into XYZ space by means of formula (1), and then converting the XYZ space into Lab space by means of formula (2) to obtain a Lab space image I corresponding to the original imageLab；
Wherein the content of the first and second substances,
1.2) image I for Lab spaceLabThree components I ofL、IaAnd IbFiltering all with a 3 x 3 gaussian convolution kernel to obtain a filtered image IGLab；
Wherein the 3 × 3 Gaussian convolution kernel is
1.3) respectively obtaining ILabThree components I of the middle Lab spaceL、IaAnd IbAnd then the three mean values LM, AM and BM are obtained by using the formula (3)And IGLabThe Euclidean distance of the three components in the image to obtain a saliency image SM;
SM(x,y)＝(LIGLab(x,y)-LM)2+(aIGLab(x,y)-AM)2+(bIGLab(x,y)-BM)2(3)
wherein (x, y) is pixel coordinate, LIGLab(x,y)、aIGLab(x, y) and bIGLab(x, y) are each IGLabThe value of the three components of (a) at coordinates (x, y);
1.4) maximum Max in saliency map SM using equation (4)SMAnd minimum MinSMNormalizing the SM to obtain a normalized saliency image NSM;
NSM(x,y)＝(SM(x,y)-MinSM)/(MaxSM-MinSM) (4)。
4. the detection method according to claims 1 and 2, characterized in that:
step 2) the steps of the adaptive iterative threshold method are as follows:
2.1) calculating the maximum gray value and the minimum gray value of the image, respectively recording as Zmax and Zmin, and making the initial threshold value
T0＝(Zmax+Zmin)/2 (5)
2.2) dividing the image into foreground and background according to the threshold Tk, and respectively calculating the average gray value Z of the foreground and the backgroundoAnd Zb；
2.3) find the new threshold:
Tk+1＝(Zo+Zb)/2 (6)
2.4) if Tk＝Tk+1The iteration is stopped and the final threshold is obtained, otherwise go to step 2.2).
5. The detection method according to claims 1 and 2, characterized in that:
step 3) judging the oil spilling region as follows:
and (4) judging the relation between the significance value of each pixel in the normalized significance image NSM and T by using the formula (7) to determine whether the pixel is an oil spill pixel.
Wherein R isoil(x, y) represents whether the pixel at the coordinate (x, y) is the pixel of the oil spilling region, if 1, the pixel is regarded as the pixel of the oil spilling region, and if 0, the pixel is regarded as the pixel of the non-oil spilling region.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010534064.5A CN111860534A (en) | 2020-06-12 | 2020-06-12 | SAR image oil spill detection method based on image significance analysis |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010534064.5A CN111860534A (en) | 2020-06-12 | 2020-06-12 | SAR image oil spill detection method based on image significance analysis |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN111860534A true CN111860534A (en) | 2020-10-30 |
Family
ID=72987624
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010534064.5A Pending CN111860534A (en) | 2020-06-12 | 2020-06-12 | SAR image oil spill detection method based on image significance analysis |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111860534A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112560929A (en) * | 2020-12-09 | 2021-03-26 | 中国石油大学(华东) | Oil spilling area determining method and device and storage medium |
| CN113191328A (en) * | 2021-05-26 | 2021-07-30 | 辽宁工程技术大学 | LSMA-IBAI comprehensive index-based impervious surface extraction method |
-
2020
- 2020-06-12 CN CN202010534064.5A patent/CN111860534A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112560929A (en) * | 2020-12-09 | 2021-03-26 | 中国石油大学(华东) | Oil spilling area determining method and device and storage medium |
| CN113191328A (en) * | 2021-05-26 | 2021-07-30 | 辽宁工程技术大学 | LSMA-IBAI comprehensive index-based impervious surface extraction method |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110310264B (en) | DCNN-based large-scale target detection method and device | |
| Williams | Fast unsupervised seafloor characterization in sonar imagery using lacunarity | |
| CN107330875B (en) | Water body surrounding environment change detection method based on forward and reverse heterogeneity of remote sensing image | |
| CN109427055B (en) | Remote sensing image sea surface ship detection method based on visual attention mechanism and information entropy | |
| CN111008664B (en) | Hyperspectral sea ice detection method based on space-spectrum combined characteristics | |
| CN111339827A (en) | SAR image change detection method based on multi-region convolutional neural network | |
| Stankov et al. | Building detection in very high spatial resolution multispectral images using the hit-or-miss transform | |
| CN111860534A (en) | SAR image oil spill detection method based on image significance analysis | |
| CN114764801A (en) | Weak and small ship target fusion detection method and device based on multi-vision significant features | |
| Zhang et al. | Saliency-driven oil tank detection based on multidimensional feature vector clustering for SAR images | |
| Yang et al. | Evaluation and mitigation of rain effect on wave direction and period estimation from X-band marine radar images | |
| CN111310771B (en) | Road image extraction method, device and equipment of remote sensing image and storage medium | |
| Song et al. | A residual convolution neural network for sea ice classification with Sentinel-1 SAR imagery | |
| CN115452759A (en) | River and lake health index evaluation method and system based on satellite remote sensing data | |
| Li et al. | Automatic infrared ship target segmentation based on structure tensor and maximum histogram entropy | |
| Sun et al. | Probabilistic neural network based seabed sediment recognition method for side-scan sonar imagery | |
| Scarsi et al. | An automated flood detection framework for very high spatial resolution imagery | |
| CN113705441A (en) | High-spatial-temporal-resolution surface water body extraction method cooperating with multispectral and SAR images | |
| CN115546199B (en) | Method for acquiring offshore oil spill area by analyzing SAR image through self-attention | |
| Han et al. | A robust LCSE-ResNet for marine man-made target classification based on optical remote sensing imagery | |
| CN116797941A (en) | Marine oil spill risk source rapid intelligent identification and classification method for high-resolution remote sensing image | |
| Wang et al. | An automatic thresholding method for water body detection from SAR image | |
| CN113177473B (en) | Automatic water body extraction method and device for remote sensing image | |
| CN114862883A (en) | Target edge extraction method, image segmentation method and system | |
| Lewiński et al. | Testing texture of VHR panchromatic data as a feature of land cover classification |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |