CN107527331B - Remote sensing identification method of polar iceberg based on double-bubble method - Google Patents

Abstract

The invention discloses a polar iceberg remote sensing identification method based on a double bubbling method. The image noise is removed by preprocessing Sentinel-1A satellite images; a detector is used in MATLAB to detect the edge area of the iceberg; the image is assigned a value to remove the image The streak noise in the background sea area; the maximum value of each pixel group in each row of the image is obtained by the double-bubble method, that is, the boundary pixel of the iceberg row; the pixels between the boundary pixels of the iceberg row are filled; The boundary of the iceberg is extracted, and finally the iceberg boundary in pixels can be obtained, and then the iceberg can be identified. Based on this, information such as iceberg area and linear size can be calculated, which improves the accuracy of iceberg identification.

Description

Remote sensing identification method of polar iceberg based on double-bubble method

technical field

The invention relates to a polar iceberg remote sensing identification method based on a double bubbling method, and belongs to the technical field of remote sensing geoscience application.

technical background

Icebergs play an important role in revealing changes in Antarctic climate and marine environmental conditions in recent years. Accurate positioning of the iceberg boundary is the primary condition for obtaining information such as iceberg area and linear size. Obtaining the iceberg boundary is of great significance for the study of iceberg ablation, generation rate and ocean current movement.

The ship-based observation method is limited by factors such as time and distance, and has great limitations, and can only provide information such as the area and linear size of a small number of icebergs. The ship-based iceberg identification method has the advantage of identifying multiple icebergs at the same time, but because it cannot precisely locate the boundary of the iceberg, the obtained information such as the area and linear size of the iceberg is not reliable.

Remote sensing images can quickly and accurately reflect sea surface information. The spatial resolution of Sentinel-1A SAR images is 40m, which is sufficient to identify most icebergs and obtain information such as iceberg area and linear size. Icebergs mostly exist in the near-polar seas. Except for a small amount of icebergs and sea ice, there are few impurities on the sea surface. Therefore, the background sea area is mainly black or dark gray on the SAR image. The iceberg has a clear contrast with the background sea area, and it is mainly white or light gray on the SAR image, and most of the icebergs are large and have obvious target characteristics on the SAR image. Therefore, the possible regions of icebergs can be pre-extracted by visual interpretation, the targets can be preliminarily extracted by the backscattering difference between the iceberg and the background sea area, and then the iceberg boundary can be determined by the pixel distribution law in the edge area of the iceberg, and the area and linear size of the iceberg can be further obtained. and other information, and finally achieve the purpose of identifying the iceberg. This provides a completely new method for identifying icebergs, which has important scientific implications for analyzing the role of iceberg melting in ocean and global climate change.

SUMMARY OF THE INVENTION

The technical problem to be solved by the present invention is to provide a polar iceberg remote sensing identification method based on the double-bubbling method, which can be fast and efficient, aiming at the difficulty of obtaining reliable information such as iceberg area and linear size for the iceberg that cannot be accurately identified. The iceberg boundary in pixels is obtained from the image, and then the iceberg area and linear size are obtained to complete the precise identification of the iceberg.

In order to solve the above-mentioned technical problems, the technical scheme proposed by the present invention is: a polar iceberg remote sensing identification method based on the double-bubbling method, comprising the following steps:

The first step is to download the Sentinel-1A SAR image in a suitable time, the suitable time is November or December;

The second step is to crop the original remote sensing image, leaving only the suspected iceberg part, and determine the iceberg with less than 10 pixels through manual identification and MATLAB data extraction, and use the assignment method to remove it;

The third step is to determine the edge area of the iceberg, use the edge detector of 3*3 window in MATLAB to scan the image, and calculate the standard deviation σ and the mean μ of the 9 pixel values in the window, and compare the standard deviation σ and the mean μ The quotient σ/μ is assigned to the center pixel of the window, and the image of the edge area of the iceberg is obtained;

The fourth step is to determine the boundary pixels of the iceberg row, which includes the following steps:

a1. Remove the noise existing in the background sea area and eliminate the interference of floating sea ice in the background sea area;

a2. Each line of the image in the edge area of the iceberg contains several pixel groups consisting of consecutive non-zero pixels. Scan each pixel group repeatedly. During the scanning process, compare the values of two adjacent pixel values in sequence. In the reverse order, the positions are exchanged. In the process of continuous comparison, all the data are finally made into order and the pixel with the largest pixel value in each pixel group is respectively reserved, and the other pixels in the pixel group are removed using the assignment method. The pixel with the largest pixel value is the iceberg row boundary pixel;

The fifth step, iceberg boundary separation;

b1. In the image obtained after processing in step a2, when there are 3 or more pixels in the same row with a value greater than 0, there may be noise points in the row, and the noise points may be removed by manual identification or with the help of MATLAB;

b2. Iceberg filling, divided into several situations:

Ⅰ. If there are two or three non-zero value pixels in a row, assign the value of the first and last two non-zero value pixels and the pixels between them as 255;

Ⅱ. There are 2N non-zero value pixels in a row and N is a natural number greater than 1, then the 2i+1 and 2i+2 non-zero value pixels and the pixels between them are assigned as 255, i∈ [0,1,...,N-1];

Ⅲ. There are 2N+1 non-zero value pixels in a row and N is a natural number greater than 1, then manually identify the iceberg area, and assign the pixel in the iceberg area to 255;

b3. Use the boundary extraction algorithm to process the filled iceberg to obtain the iceberg boundary in pixels.

The present invention is based on the polar iceberg remote sensing identification method based on the double bubbling method, and also has the following features:

1. In the second step, preprocessing such as mean filtering is performed on the cropped image to eliminate noise.

2. In step a1, for the pixels whose pixel value is less than 50 in the edge area image, assign the pixel value to 0, and the pixel value greater than or equal to 50 remains unchanged, so as to realize the elimination of floating sea ice in the background sea area. .

3. The method of removing pixels or noise is to assign the value of the pixel that needs to be removed as 0.

Identifying icebergs and obtaining their associated area and linear dimensions is of great significance for studying the interrelationships between Antarctic ice sheets, sea ice, oceans and climate change. The invention realizes the remote sensing identification of polar icebergs, according to the difference of backscattered signals between the icebergs and the background sea area and the pixel distribution law of the edge area, using the exchange ordering of the "double bubble method" and the filling of the ice surface to quickly and accurately Identifying icebergs, and then obtaining information such as iceberg area and linear size, has unparalleled advantages compared to traditional identification methods such as ship bases.

The specific beneficial effects of the present invention are as follows:

First, the Sentinel-1A SAR image used in the present invention can be obtained free of charge, which is convenient and quick.

Second, the iceberg area and linear size and other information obtained by the invention after identifying the iceberg can be further applied to the study of iceberg melting, generation rate and ocean current movement, so as to analyze the interaction between Antarctic ice sheet, sea ice and ocean and climate change. Relationships provide reliable data support.

Thirdly, the present invention utilizes the difference of backscattering intensity of different objects on the SAR image, combined with the pixel distribution law in the edge area, and uses MATLAB programming to complete the work of iceberg identification and boundary extraction, thereby obtaining information such as iceberg area and linear size. The method is simple and the processing process is concise.

Fourth, the present invention has simple operation steps, combined with pixel distribution rules and simple software operations, does not require a large number of calculations and in-depth field investigations, and has high precision, which is suitable for iceberg identification in special environments in polar regions.

Description of drawings

The present invention will be further described below in conjunction with the accompanying drawings.

Figure 1 is a flowchart of the polar iceberg identification method.

Figure 2 shows the original image preprocessing.

Figure 3 shows the edge region of the iceberg obtained by the edge detector.

Figure 4 shows the image noise removal in the edge area.

Figure 5 shows the iceberg line boundary obtained by the "double-bubbling method".

Figure 6 shows the interior filling of the iceberg boundary.

Figure 7 shows the extraction result of the iceberg boundary.

Detailed ways

The image data of this example is the Sentinel-1A SAR image, which can be downloaded for free from the ASF website (https://vertex.daac.asf.alaska.edu/). The geographic coordinates of the image center are S 77°, W 38°, and the time is 2016. November 29.

Figure 1 is a flowchart of this example. The specific implementation steps of the iceberg identification method include the following: Step 1: Download Sentinel-1A SAR images in November and December (summer in the southern hemisphere). The image location is generally the sea area on the edge of the Antarctic continent.

The second step is to crop the original remote sensing image, leaving only the suspected iceberg part. Use ENVI software to read the cropped remote sensing images, and perform preprocessing such as mean filtering to eliminate noise. And the iceberg with the number of pixels less than 10 was determined by manual identification and MATLAB data extraction, and it was removed by the assignment method. The results are shown in Figure 2.

The third step is to determine the edge area of the iceberg. Scan the image using a detector with a 3*3 window in MATLAB, and calculate the standard deviation (σ) and mean (μ) of the 9 pixel values in the window, and assign the quotient of the standard deviation and the mean (σ/μ) Give the center pixel of the window, and thus obtain an image of the edge area of the iceberg. Through the analysis of the iceberg edge area, it can be seen that in the newly generated image, the two pixels in the center of the iceberg edge area have the largest values, among which the larger value belongs to the background sea area pixel, and the smaller value belongs to the iceberg pixel, which belongs to the background. Ocean cells always have larger values than iceberg cells. The iceberg edge area is shown in Figure 3.

The fourth step is to determine the boundary pixels of the iceberg row, and the iceberg row boundary pixels refer to the boundary pixels on the left and right sides of the iceberg in the image of the edge region of the iceberg. Including the following aspects:

a1. Remove the noise existing in the pixels of the background sea area by assigning them, mainly to eliminate the interference of floating sea ice in the background sea area. In this example, for the pixels whose pixel value is less than 50 in the edge area image, the pixel value is assigned to 0, and the pixel value greater than or equal to 50 remains unchanged, so as to eliminate the floating sea ice in the background sea area. The processing results are shown in Figure 4.

a2. According to the distribution law of the pixels in the edge area of the iceberg, use the "double bubble method" to determine the boundary pixels of the iceberg row. The core idea of the "double bubble method" is to regard the row pixels in the edge area of the iceberg as several continuous pixel groups, scan each pixel group repeatedly, and compare two adjacent pixels in sequence during the scanning process. If the order is reversed, the positions are exchanged. In the process of continuous comparison, all the data are finally ordered and the maximum value of the pixel group is determined respectively. The pixel with the largest pixel value is the iceberg line boundary. Other cells in the cell group are removed using the assignment method, that is, the cell value that needs to be removed is assigned 0. The processing results are shown in Figure 5.

The fifth step, the iceberg boundary separation. Including the following aspects:

b1. There is some noise in the processing result of the "double bubbling method". When there are 3 or more pixels in the same row with a value greater than 0, there may be noise points in the row. Manual identification or MATLAB is used to remove the noise points.

b2. Iceberg filling, divided into several situations:

Ⅰ. If there are two or three non-zero value pixels in a row, assign the value of the first and last two non-zero value pixels and the pixels between them as 255;

Ⅱ. There are 2N non-zero value pixels in a row and N is a natural number greater than 1, then the 2i+1 and 2i+2 non-zero value pixels and the pixels between them are assigned as 255, i∈ [0,1,...,N-1];

Ⅲ. There are 2N+1 non-zero value pixels in a row and N is a natural number greater than 1, then manually identify the iceberg area, and assign the pixel in the iceberg area to 255.

The iceberg filling results are shown in Figure 6.

b3. Use the boundary extraction algorithm to process the filled iceberg to obtain the iceberg boundary in pixels. From this, information such as iceberg area and linear size can be obtained. The iceberg boundary identified by this method in pixel units is shown in Figure 7. Through calculation, the number of iceberg pixels is 138021, the area is 220.8336km ² , the maximum vertical length is 24.5124km, and the maximum horizontal length is 11.1427km .

In addition to the above-described embodiments, the present invention may also have other embodiments. All technical solutions formed by equivalent replacement or equivalent transformation fall within the protection scope of the present invention.

Claims (4)

1. A polar iceberg remote sensing identification method based on double bubbling method, comprising the following steps: The first step is to download the Sentinel-1A SAR image in a suitable time, the suitable time is November or December; The second step is to crop the original remote sensing image, leaving only the suspected iceberg part, and determine the iceberg with less than 10 pixels through manual identification and MATLAB data extraction, and use the assignment method to remove it; The third step is to determine the edge area of the iceberg, use the edge detector of 3*3 window in MATLAB to scan the image, and calculate the standard deviation σ and the mean μ of the 9 pixel values in the window, and compare the standard deviation σ and the mean μ The quotient σ/μ is assigned to the center pixel of the window, and the image of the edge area of the iceberg is obtained; The fourth step is to determine the boundary pixels of the iceberg row, which includes the following steps: a1. Remove the noise existing in the background sea area and eliminate the interference of floating sea ice in the background sea area; a2. Each line of the image in the edge area of the iceberg contains several pixel groups consisting of consecutive non-zero pixels. Scan each pixel group repeatedly. During the scanning process, compare the values of two adjacent pixel values in sequence. In the reverse order, the positions are exchanged. In the process of continuous comparison, all the data are finally made into order and the pixel with the largest pixel value in each pixel group is respectively reserved, and the other pixels in the pixel group are removed using the assignment method. The pixel with the largest pixel value is the iceberg row boundary pixel; The fifth step, the iceberg boundary separation, the specific process is as follows: b1. In the image obtained after processing in step a2, when there are at least 3 pixels whose pixel value is greater than 0 in the same row, manually identify or remove noise points with the help of MATLAB; b2. Iceberg filling, divided into several situations: I. If there are two or three non-zero value pixels in a row, assign the value of the first and last two non-zero value pixels and the pixels between them as 255; II. There are 2N non-zero value pixels in a row and N is a natural number greater than 1, then the 2i+1 and 2i+2 non-zero value pixels and the pixels between them are assigned as 255, i∈ [0,1,...,N-1]; Ⅲ. There are 2N+1 non-zero value pixels in a row and N is a natural number greater than 1, then manually identify the iceberg area, and assign the pixel in the iceberg area to 255; b3. Use the boundary extraction algorithm to process the filled iceberg to obtain the iceberg boundary in pixels. 2 . The polar iceberg remote sensing identification method based on the double bubbling method according to claim 1 , wherein in the second step, mean filtering preprocessing is performed on the cropped image to eliminate noise. 3 . 3. The polar iceberg remote sensing identification method based on double bubbling method according to claim 1, is characterized in that: in step a1, for the pixel whose pixel value is less than 50 in the edge area image, assign its pixel value as 0, the pixel value greater than or equal to 50 remains unchanged, to achieve the elimination of floating sea ice in the background sea area. 4 . The polar iceberg remote sensing identification method based on the double-bubbling method according to claim 1 , wherein the method for removing pixels or noises is to assign the value of the pixel that needs to be removed as 0. 5 .

Images