CN102999897A - Method and device for sea surface oil spillage detection based on SAR (synthetic aperture radar) image - Google Patents

Abstract

Disclosed are a method and a device for sea surface oil spillage detection based on an SAR (synthetic aperture radar) image. The method includes: converting the SAR image into a binary image by performing threshold segmentation for gray of pixels in the sea surface SAR image; for a neighborhood in predetermined size of each pixel in the binary image, determining whether number and/or ratio of the pixels with the pixel value of 1 in the neighborhood is larger than a preset value or not; identifying the neighborhoods that whether the number and/or the ratio of the pixels with the pixel value of 1 is larger than the preset value or not; and using an image boundary formed by all the identified neighborhoods as an initial zero level set used for detecting level set segmentation so as to detect sea surface oil spillage.

Description

Method and apparatus based on SAR image detection offshore spilled oil

Technical field

The application relates to the technical fields such as image processing, remote sensing image, marine environmental management, particularly, relates to the method and apparatus based on SAR (Synthetic Aperture Radar, synthetic-aperture radar) image detection offshore spilled oil.

Background technology

In the last few years, the marine oil overflow event was the trend of magnifying year by year, detected and monitor that the diffusion in offshore spilled oil zone and the technology that moves have caused that people pay close attention to greatly, had very high researching value.

At present, utilizing the SAR image to carry out that oil spilling surveys is comparatively mainly also to be means preferably in the world.The trickle wave that SAR imagery exploitation sea produces shows as light tone and the oil spilling zone shows as dark-coloured principle and identifies contaminated area at radar image.The fundamental sum committed step of SAR image detection oil spilling is the blackspot zone of finding in the image, is the place in oil spilling zone most probably namely, and then is determined further and analyzes.The mode of drawing by hand the blackspot zone wastes time and energy and is confined to artificial professional knowledge, and the Computer Automatic Recognition technology then can improve a lot of efficient.But the speckle noise in the SAR image forming course brings very large challenge can for the Computer Automatic Recognition technology.

The edge detection method of traditional optical imagery (for example, Canny operator, Zero Crossing operator) is owing to the impact of very noisy is not suitable for the SAR image segmentation.Therefore, the algorithm of much integrating boundary information and area information is suggested, a kind of active contour method that just is based on level set wherein.Level Set Method is embedded into curve in the curved surface of high one dimension, can express easily the geometric parameter of curve, and has good change in topology structure and numerical discretization form easily, therefore is widely used in image segmentation.Yet, utilize the Level Set Method of non-convex functional model to depend on especially the selection of initial position, thus so that net result not necessarily global minimum or the area-of-interest that will cut apart.Overcome this point and often need the human assistance input.Yet for high-resolution SAR image, the workload of human assistance input is appreciable.

Therefore, how automatically selecting starting condition is the emphasis that this class model is realized auto Segmentation to obtain results needed.Some algorithms utilize pretreated method (for example image recovery method) to remove noise, thereby realize that automatic level set cuts apart.Yet this algorithm is effectively denoising and destroyed some intrinsic informations of original image on the one hand, has also increased on the one hand in addition the execution time of algorithm, has reduced efficient.

The application aim to provide a kind of method and apparatus can overcome in the above shortcoming one of at least, find quickly and accurately initial level collection position, reach the purpose of automatic identification blackspot.

Summary of the invention

The application's purpose provides a kind of method and apparatus based on SAR image detection offshore spilled oil, and it can identify the initial zero level collection of cutting apart for level set from the SAR image, with the zone of the oil spilling in the automatic detection SAR image.

According to the application's one side, a kind of method based on SAR image detection offshore spilled oil is disclosed, comprising: by the pixel gray scale in the SAR image on sea is carried out Threshold segmentation, be bianry image with described SAR image transitions; For the neighborhood of the pre-sizing of each pixel in the described bianry image, determine that pixel value in the described neighborhood is whether the quantity of 1 pixel and/or ratio are greater than predetermined value; The identification pixel value is whether the quantity of 1 pixel and/or ratio be greater than the neighborhood of described predetermined value; The border of the image that all neighborhoods that identify are consisted of is used for initial zero level collection that level set cuts apart to detect offshore spilled oil as detecting.

According to the application on the other hand, disclose a kind of device based on SAR image detection offshore spilled oil, having comprised: the Threshold segmentation module by the pixel gray scale in the SAR image on sea is carried out Threshold segmentation, is bianry image with described SAR image transitions; Determination module for the neighborhood of the pre-sizing of each pixel in the described bianry image, determines that pixel value in the described neighborhood is whether the quantity of 1 pixel and/or ratio are greater than predetermined value; Identification module, identification pixel value are whether the quantity of 1 pixel and/or ratio be greater than the neighborhood of described predetermined value; Boundary Extraction module, the Boundary Extraction of the image that all neighborhoods that identify are consisted of out are used for initial zero level collection that level set cuts apart to detect offshore spilled oil as detecting.

Description of drawings

Fig. 1 is according to the process flow diagram of illustrative embodiments of the application based on the method for SAR image detection offshore spilled oil;

Fig. 2 is the SAR image with the oil spilling zone among embodiment of the application;

Fig. 3 is according to SAR image the bianry image through Threshold segmentation after obtain of illustrative embodiments of the application to Fig. 2;

Fig. 4 shows blackspot zone most probably for according to illustrative embodiments of the application the bianry image of Fig. 3 further being processed the image that obtains;

Fig. 5 is cut apart the result who obtains for carrying out level set with the boundary position that identifies among Fig. 4 as initial zero level collection according to illustrative embodiments of the application; And

Fig. 6 is according to the block diagram of illustrative embodiments of the application based on the device of SAR image detection offshore spilled oil.

Embodiment

Be described in conjunction with the illustrative embodiments of embodiment to the application with reference to the accompanying drawings.

Fig. 1 shows according to the method 100 of embodiment of the application based on SAR image detection offshore spilled oil.As shown in the figure, at step S101, the pixel gray scale in the SAR image on sea is carried out Threshold segmentation, take with the SAR image transitions as bianry image, find out possible blackspot zone.

Threshold segmentation is a kind of image Segmentation Technology based on the zone, and its ultimate principle is by setting gray threshold the image slices vegetarian refreshments to be classified.Fig. 2 exemplarily shows the SAR image with the oil spilling zone.From visually, the gray-scale value that can depend on simply image distinguishes blackspot zone and background: the zone that gray-scale value is lower is identified as the blackspot zone usually, and the zone that gray-scale value is higher is identified as background.For example, in Fig. 2, the darker band-like portions of center section color can be identified as the blackspot zone usually, and other parts then are identified as background.Particularly, can set gray threshold, and gray-scale value in the SAR image is set as 1 greater than the zone of threshold value or the pixel value of pixel, gray-scale value is not more than the zone of threshold value or the pixel value of pixel is set as 0.SAR image shown in Figure 2 for example is shown in Figure 3 through the image that threshold method generates.

In one embodiment, gray threshold can followingly arrange:

lambda＝0.382*d_max+0.618*d_min，

Wherein, lambda is the gray threshold that is used to form bianry image, and d_max and d_min are respectively maximal value and the minimum value of pixel grey scale in the SAR image.Like this, pixel value then is set as 1 greater than the pixel value in the zone (or pixel) of lambda in the SAR image, and the pixel value that pixel value is not more than the zone (or pixel) of lambda then is set as 0.

In another embodiment, can utilize Bayes's threshold method to determine gray threshold based on the probability distribution in blackspot zone.In this case, gray threshold can be the maximum T value of the mixing probability density p (T) that makes image=p (o) p (x|o)+p (b) p (x|b).Wherein, x is image pixel, p (o) and p (b) are respectively the probability of blackspot zone o (zone of x＜T in the image) and background area b (zone of x＞T in the image), and p (x|o) and p (x|b) are respectively the known in advance regional o of blackspot and the prior probability of background area b.P (o) and p (b) for example can utilize the statistics with histogram of SAR image to obtain.

Can see that from bianry image shown in Figure 3 the threshold application method can't once be finished and cut apart, the zone also is marked as 1 because the interference of noise can make a lot of non-blackspots, causes excessive cutting apart.In order to filter out blackspot zone most probably, at step S102, for the neighborhood of the pre-sizing of each pixel in the bianry image, be whether the quantity of 1 pixel and/or this neighborhood of ratio-dependent are blackspot zone most probably according to pixel value wherein.Particularly, can judge that pixel value in each neighborhood is whether the quantity of 1 pixel and/or ratio are greater than predetermined value.

At step S103, with the border in the blackspot zone most probably identified as the initial zero level collection that is used for level set movements.For example, be whether the quantity of 1 pixel and/or ratio are identified as most probably blackspot zone greater than the neighborhood of predetermined value with pixel value, and will be wherein all the pixel value of pixels be made as 1; The pixel value that with pixel value is the whole pixels in the quantity of 1 pixel and/or the neighborhood that ratio is not more than predetermined value is made as-1.Like this, can further obtain image shown in Figure 4 according to bianry image shown in Figure 3, wherein, middle white portion represents that the blackspot most probably that identifies is regional.

In one embodiment, can select 15 * 15 square block to whole image point by point scanning, if most pixel values are 1 in this square block, then the pixel value of the pixel in the whole square area is labeled as 1, be considered as blackspot zone most probably.Here, the quantity of exhausted most pixel value is for to deduct the value that obtains after its diameter length value less than selected foursquare area value, and wherein the diameter length value is the maximal value of any two points distance in the zone, and for example regional diameter length is 21 among the embodiment.

After this, at step S104, can utilize this initial zero level collection to carry out the profile that level set movements detects finally to obtain blackspot.For example, can utilize Level Set Method to find the solution image segmentation problem by following energy functional:

$E(\mathrm{\φ},C_1,C_2)=\underset{\mathrm{\Ω}}{\∫}\left|H\left(\mathrm{\φ}\right)\right|\mathrm{dx}+\underset{\mathrm{\Ω}}{\∫}{\mathrm{\λ}}_1\left(\mathrm{\φ}\right){(I\left(x\right)-C_1)}^2+{\mathrm{\λ}}_2(1-H\left(\mathrm{\φ}\right)){(I\left(x\right)-C_2)}^2\mathrm{dx}$

Wherein, φ＞0 o'clock, H (φ)=1; And φ≤0 o'clock, H (φ)=0.In this energy functional formula, by the long-pending formula of lap as can be known first be length of curve; Next one then is the k-average sorting criterion of domain of dependence information.λ ₁, λ ₂Be weighting parameters, if it is less to choose their value, it is shorter then to cause cutting apart the border that obtains the zone, thereby avoids noise, in an embodiment, can select λ ₁=λ ₂=0.0006.

In order to find the solution the minimum value of this energy functional, can utilize Euler-Lagrangian method and gradient descent method, the following development-oriented partial differential equation of controlled level set movements:

$\left\{\begin{array}{c}\frac{\∂\mathrm{\φ}}{\∂t}=|\▿\mathrm{\φ}|[\mathrm{div}\left(\frac{\▿\mathrm{\φ}}{|\▿\mathrm{\φ}|}\right)-{\mathrm{\λ}}_1{(I\left(x\right)-C_1)}^2+{\mathrm{\λ}}_2{(I\left(x\right)-C_2)}^2]\\ C_1=\frac{{\∫}_{\mathrm{\Ω}}H\left(\mathrm{\φ}\right)I\left(x\right)\mathrm{dx}}{{\∫}_{\mathrm{\Ω}}H\left(\mathrm{\φ}\right)\mathrm{dx}},C_2=\frac{{\∫}_{\mathrm{\Ω}}(1-H\left(\mathrm{\φ}\right))I\left(x\right)\mathrm{dx}}{{\∫}_{\mathrm{\Ω}}(1-H\left(\mathrm{\φ}\right))\mathrm{dx}},\\ \frac{\∂\mathrm{\φ}}{\∂\stackrel{\→}{n}}=0,\mathrm{on}\∂\mathrm{\Ω},\\ \mathrm{\φ}(x,0)={\mathrm{\φ}}_0\left(x\right),\mathrm{in\Ω}\end{array}\right.$

Wherein, φ ₀(x) be the initial level collection, that is, and the border in the blackspot zone most probably of in step S103, determining.By with above-mentioned partial differential equation discretize, can utilize computer solving, until level set do not developing, perhaps reach till certain iterations (for example can be set to 500 times), obtain the blackspot profile with detection.Because the technology of utilizing initial zero level collection to carry out level set movements is existing image Segmentation Technology, therefore do not repeat them here.

Fig. 5 exemplarily shows and carries out level set with the boundary position that identifies among Fig. 4 as initial zero level collection and cut apart the result who obtains.

After this, the blackspot profile that obtains is further analyzed, for example analyzes girth and the Area Ratio of cut zone, the factors such as square estimation can find the oil spilling zone.

According to an embodiment, when not recognizing at step S102 in the situation in blackspot zone most probably, for example, pixel value is in the quantity of 1 pixel and/or the situation that ratio is not more than predetermined value in each neighborhood, can select initial zero level collection by bianry image being carried out artificial judgment.

The application also provides a kind of device based on SAR image detection offshore spilled oil.Fig. 6 has used the device 600 based on SAR image detection offshore spilled oil according to an illustrative embodiments of the application, and it comprises Threshold segmentation module 601, determination module 602, identification module 603 and Boundary Extraction module 604.Threshold segmentation module 601 is bianry image by the pixel gray scale in the SAR image on sea is carried out Threshold segmentation with the SAR image transitions.Determination module 602 is for the neighborhood of the pre-sizing of each pixel in the bianry image, determines that pixel value in each neighborhood is whether the quantity of 1 pixel and/or ratio are greater than predetermined value.Identification module 603 identification pixel values are whether the quantity of 1 pixel and/or ratio be greater than the neighborhood of this predetermined value.The Boundary Extraction of the image that all neighborhoods that Boundary Extraction module 604 will identify consist of is out as detect being used for initial zero level collection that level set cuts apart to detect offshore spilled oil.

According to an embodiment, Threshold segmentation module 601 can be made as 0 greater than the gray scale of the pixel of predetermined threshold with gray scale in the SAR image, and the gray scale that gray scale is not more than the pixel of predetermined threshold is made as 1.For example, this predetermined threshold can be 0.382*d_max+0.618*d_min, and wherein, d_max and d_min are respectively maximal value and the minimum value of the pixel gray scale in the SAR image.

According to another embodiment, Threshold segmentation module 601 can be set as predetermined threshold the maximum T value of the mixing probability density p (T) that makes the SAR image=p (o) p (x|o)+p (b) p (x|b), wherein, x is the pixel value of the pixel in the SAR image, p (o) and p (b) are respectively the probability in the zone of x＜T and x＞T in the SAR image, and p (x|o) and p (x|b) are respectively in advance the prior probabilities in the zone of x＜T and x＞T in the known SAR image.P (o) and p (b) for example can utilize the statistics with histogram of SAR image to obtain.

According to an embodiment, to be the quantity of 1 pixel and/or ratio with the pixel value that identifies be made as 1 greater than the pixel value of the whole pixels in the neighborhood of predetermined value to identification module 603, and be that the pixel value of the whole pixels in the quantity of 1 pixel and/or the neighborhood that ratio is not more than predetermined value is made as-1 with the pixel value that identifies.Above-mentioned neighborhood can be square, circle or the polygon with pre-sizing.

According to an embodiment, device 600 can also comprise the indicating module (not shown), go out quantity that pixel value is 1 pixel and/or ratio whether in the situation greater than the neighborhood of predetermined value unidentified, indicating module can be indicated based on the bianry image artificial cognition and is used for initial zero level collection that level set cuts apart to detect offshore spilled oil.

Abovely be described in conjunction with the method and apparatus of exemplary embodiment to the application with reference to accompanying drawing.Should be appreciated that scope of the present invention is not limited to specifically described embodiment, various modifications and the distortion of above-mentioned embodiment are also contained in protection scope of the present invention.

Claims (14)

1. based on the method for SAR image detection offshore spilled oil, comprising:

By the pixel gray scale in the SAR image on sea is carried out Threshold segmentation, be bianry image with described SAR image transitions;

For the neighborhood of the pre-sizing of each pixel in the described bianry image, determine that pixel value in the described neighborhood is whether the quantity of 1 pixel and/or ratio are greater than predetermined value;

The identification pixel value is whether the quantity of 1 pixel and/or ratio be greater than the neighborhood of described predetermined value;

The border of the image that all neighborhoods that identify are consisted of is used for initial zero level collection that level set cuts apart to detect offshore spilled oil as detecting.

2. the method for claim 1, wherein described Threshold segmentation comprises:

Gray scale in the described SAR image is made as 0 greater than the gray scale of the pixel of predetermined threshold; And

The gray scale that gray scale in the described SAR image is not more than the pixel of described predetermined threshold is made as 1.

3. method as claimed in claim 2, wherein, described predetermined threshold is:

0.382*d_max+0.618*d_min，

Wherein, d_max and d_min are respectively maximal value and the minimum value of the pixel gray scale in the described SAR image.

4. method as claimed in claim 2, wherein, described predetermined threshold is the mixing probability density p (T) that makes described SAR image=maximum T value of p (o) p (x|o)+p (b) p (x|b),

Wherein, x is the pixel value of the pixel in the described SAR image, p (o) and p (b) are respectively the probability in the zone of x＜T and x＞T in the described SAR image, and p (x|o) and p (x|b) are respectively the prior probabilities in the zone of x＜T and x＞T in the known in advance described SAR image.

5. the method for claim 1, wherein, being the quantity of 1 pixel and/or ratio with the pixel value that identifies is made as 1 greater than the pixel value of the whole pixels in the neighborhood of described predetermined value, and is that the pixel value of the whole pixels in the quantity of 1 pixel and/or the neighborhood that ratio is not more than described predetermined value is made as-1 with the pixel value that identifies.

6. the method for claim 1, wherein described neighborhood is square, circle or the polygon with pre-sizing.

7. based on the method for SAR image detection offshore spilled oil, comprising:

By the pixel gray scale in the SAR image on sea is carried out Threshold segmentation, be bianry image with described SAR image transitions;

For the neighborhood of the pre-sizing of each pixel in the described bianry image, determine that pixel value in the described neighborhood is whether the quantity of 1 pixel and/or ratio are greater than predetermined value;

Go out quantity that pixel value is 1 pixel and/or ratio whether in the situation greater than the neighborhood of described predetermined value unidentified, be used for initial zero level collection that level set cuts apart to detect offshore spilled oil based on described bianry image artificial cognition.

8. based on the device of SAR image detection offshore spilled oil, comprising:

The Threshold segmentation module by the pixel gray scale in the SAR image on sea is carried out Threshold segmentation, is bianry image with described SAR image transitions;

Determination module for the neighborhood of the pre-sizing of each pixel in the described bianry image, determines that pixel value in the described neighborhood is whether the quantity of 1 pixel and/or ratio are greater than predetermined value;

Identification module, identification pixel value are whether the quantity of 1 pixel and/or ratio be greater than the neighborhood of described predetermined value;

Boundary Extraction module, the Boundary Extraction of the image that all neighborhoods that identify are consisted of out are used for initial zero level collection that level set cuts apart to detect offshore spilled oil as detecting.

9. device as claimed in claim 8, wherein, described Threshold segmentation module is made as 0 with gray scale in the described SAR image greater than the gray scale of the pixel of predetermined threshold, and the gray scale that gray scale in the described SAR image is not more than the pixel of described predetermined threshold is made as 1.

10. device as claimed in claim 9, wherein, described predetermined threshold is:

0.382*d_max+0.618*d_min，

Wherein, d_max and d_min are respectively maximal value and the minimum value of the pixel gray scale in the described SAR image.

11. device as claimed in claim 9, wherein, described predetermined threshold is the maximum T value of the mixing probability density p (T) that makes described SAR image=p (o) p (x|o)+p (b) p (x|b),

Wherein, x is the pixel value of the pixel in the described SAR image, p (o) and p (b) are respectively the probability in the zone of x＜T and x＞T in the described SAR image, and p (x|o) and p (x|b) are respectively the prior probabilities in the zone of x＜T and x＞T in the known in advance described SAR image.

12. device as claimed in claim 8, wherein, to be the quantity of 1 pixel and/or ratio with the pixel value that identifies be made as 1 greater than the pixel value of the whole pixels in the neighborhood of described predetermined value to described identification module, and be that the pixel value of the whole pixels in the quantity of 1 pixel and/or the neighborhood that ratio is not more than described predetermined value is made as-1 with the pixel value that identifies.

13. device as claimed in claim 8, wherein, described neighborhood is square, circle or the polygon with pre-sizing.

14. device as claimed in claim 8 also comprises:

Indicating module goes out quantity that pixel value is 1 pixel and/or ratio whether in the situation greater than the neighborhood of described predetermined value unidentified, and indication is used for initial zero level collection that level set cuts apart to detect offshore spilled oil based on described bianry image artificial cognition.

Images