CN-121999379-A - Sea ice edge determination method based on remote sensing satellite

CN121999379ACN 121999379 ACN121999379 ACN 121999379ACN-121999379-A

Abstract

The application discloses a sea ice edge determining method based on a remote sensing satellite. The method comprises the steps that a first image corresponding to a remote sensing image and an actual sea ice edge image can be obtained to serve as training samples, then a horizontal gradient image and a vertical gradient image corresponding to the first image are determined, then a first sea ice edge prediction image corresponding to the first image can be generated based on a first generation network through the first image, the horizontal gradient image and the vertical gradient image, and then a first judging result corresponding to the actual sea ice edge image and the first sea ice edge prediction image can be respectively determined through a first judging network. Then, training the first generation network and the first discrimination network by a training mode of circularly generating the countermeasure network. And determining a second sea ice edge prediction image corresponding to the remote sensing image to be identified through the first generation network after training is completed, so that the prediction of the sea ice edge is realized, and the accuracy of sea ice edge identification can be improved.

Inventors

QU TIANLONG
LIU ANG
XU QIKAI

Assignees

银河航天科技(安徽)有限公司

Dates

Publication Date: 20260508
Application Date: 20260108

Claims (10)

1. The sea ice edge determining method based on the remote sensing satellite is characterized by comprising the following steps of: Acquiring a remote sensing image acquired by a satellite aiming at the ocean surface, determining a first image corresponding to the remote sensing image, and acquiring an actual sea ice edge image corresponding to the first image; Determining a horizontal gradient map corresponding to the first image and a vertical gradient map corresponding to the first image; Generating a first sea ice edge prediction image through a first generation network according to the first image, the horizontal gradient map and the vertical gradient map; Inputting the actual sea ice edge image and the first image into a first discrimination network to obtain a first discrimination result corresponding to the actual sea ice edge image, and inputting the first sea ice edge prediction image and the first image into the first discrimination network to obtain a first discrimination result corresponding to the first sea ice edge prediction image, wherein the first discrimination result is used for representing the authenticity of the image in the input own network judged by the first discrimination network; training the first generation network and the first discrimination network according to the first discrimination result to obtain a trained first generation network, and Under the condition that a remote sensing image to be identified of the sea ice edge to be identified is received, a second sea ice edge prediction image corresponding to the remote sensing image to be identified is generated through the first generation network after training is completed, and the sea ice edge is determined according to the second sea ice edge prediction image.
2. The method of claim 1, wherein the operation of determining a horizontal gradient map corresponding to the first image comprises: Performing convolution processing on the first image through a horizontal kernel, and determining a horizontal gradient map corresponding to the first image, wherein the horizontal gradient map is used for displaying brightness variation in the first image in the horizontal direction; an operation of determining a vertical gradient map corresponding to the first image, comprising: And carrying out convolution processing on the first image through a vertical kernel, and determining a vertical gradient map corresponding to the first image, wherein the horizontal gradient map is used for displaying brightness change in the first image in the vertical direction, and the vertical kernel is a transpose matrix of the horizontal kernel.
3. The method of claim 2, wherein the horizontal kernel is in the form of: In the horizontal core ~ Is a negative number of the number, ~ Is a positive number.
4. The method of claim 1, wherein prior to generating a first sea-ice edge prediction image from the first image, the horizontal gradient map, and the vertical gradient map by generating a network, further comprising: inputting the first image into an edge detection network, and acquiring a second image corresponding to the first image; An operation of generating a first sea ice edge prediction image through a first generation network according to the first image, the horizontal gradient map and the vertical gradient map, comprising: Inputting the first image, the second image, the horizontal gradient map and the vertical gradient map into the first generation network to generate the first sea ice edge prediction image.
5. The method of claim 4, wherein generating a second sea ice edge prediction image corresponding to the remote sensing image to be identified via the trained first generation network comprises: Determining an input image corresponding to the remote sensing image to be identified, and determining a horizontal gradient map and a vertical gradient map corresponding to the input image; Inputting the input image into the edge detection network, and acquiring an edge detection image corresponding to the input image; And inputting the input image, the edge detection image, a horizontal gradient map and a vertical gradient map corresponding to the input image into a first generation network after training is completed, and generating the second sea ice edge prediction image.
6. The method of claim 1, wherein training the first generation network and the first discrimination network according to the first discrimination result, before obtaining a trained generation network, further comprises: Inputting the actual sea ice edge image into a second generation network to obtain a third image corresponding to the actual sea ice edge image, wherein the third image is the same as the image domain of the first image; inputting the third image into a second discrimination network to obtain a second discrimination result corresponding to the third image, and inputting the first image into the second discrimination network to obtain a second discrimination result corresponding to the first image; training the first generating network and the first judging network according to the judging result to obtain a trained generating network, wherein the operation comprises the following steps: And training the first generation network, the first discrimination network, the second generation network and the second discrimination network according to the first discrimination result, the second discrimination result, the first image, the actual sea ice edge image, the first sea ice edge prediction image and the third image.
7. The method of claim 6, wherein training the first generation network, the first discrimination network, the second generation network, the second discrimination network, based on the first discrimination result, the second discrimination result, the first image, the actual sea-ice edge image, the first sea-ice edge prediction image, and the third image, comprises: Training the first generation network, the first discrimination network, the second generation network, the second discrimination network by the following loss functions: , , , Wherein X represents a first image, X represents a first sample set comprising a number of first images, Y represents an actual sea ice edge image, Y represents a second sample set comprising a number of actual sea ice edge images, Representing a first generation network, Representing a first discrimination network, Representing a second generation network, A second discrimination network is represented and is shown, Representing a first image X of a first sample set X is expected, Representing the actual sea ice edge image Y expectations with respect to the second sample set Y.
8. A storage medium comprising a stored program, wherein the method of any one of claims 1 to 7 is performed by a processor when the program is run.
9. A device for determining sea ice edges based on remote sensing satellites, comprising: The acquisition module is used for acquiring a remote sensing image acquired by a satellite aiming at the ocean surface, determining a first image corresponding to the remote sensing image and acquiring an actual sea ice edge image corresponding to the first image; A gradient map determining module, configured to determine a horizontal gradient map corresponding to the first image and a vertical gradient map corresponding to the first image; The edge image generation module is used for generating a first sea ice edge prediction image through a first generation network according to the first image, the horizontal gradient map and the vertical gradient map; The judging module is used for inputting the actual sea ice edge image and the first image into a first judging network to obtain a first judging result corresponding to the actual sea ice edge image, inputting the first sea ice edge predicted image and the first image into the first judging network to obtain a first judging result corresponding to the first sea ice edge predicted image, wherein the first judging result is used for representing the authenticity of the image which is judged by the first judging network and is input into the network; The training module is used for training the first generation network and the first discrimination network according to the first discrimination result to obtain a first generation network after training; The sea ice edge recognition module is used for generating a second sea ice edge prediction image corresponding to the sea ice edge to be recognized through the first generation network after training is completed under the condition that the sea ice edge to be recognized is received, and determining the sea ice edge according to the second sea ice edge prediction image.
10. A device for determining sea ice edges based on remote sensing satellites, comprising: processor, and A memory, coupled to the processor, for providing instructions to the processor to process the following processing steps: Acquiring a remote sensing image acquired by a satellite aiming at the ocean surface, determining a first image corresponding to the remote sensing image, and acquiring an actual sea ice edge image corresponding to the first image; Determining a horizontal gradient map corresponding to the first image and a vertical gradient map corresponding to the first image; Generating a first sea ice edge prediction image through a first generation network according to the first image, the horizontal gradient map and the vertical gradient map; Inputting the actual sea ice edge image and the first image into a first discrimination network to obtain a first discrimination result corresponding to the actual sea ice edge image, and inputting the first sea ice edge prediction image and the first image into the first discrimination network to obtain a first discrimination result corresponding to the first sea ice edge prediction image, wherein the first discrimination result is used for representing the authenticity of the image in the network input by the first discrimination network; training the first generation network and the first discrimination network according to the first discrimination result to obtain a first generation network after training is completed; Under the condition that a remote sensing image to be identified of the sea ice edge to be identified is received, a second sea ice edge prediction image corresponding to the remote sensing image to be identified is generated through the first generation network after training is completed, and the sea ice edge is determined according to the second sea ice edge prediction image.

Description

Sea ice edge determination method based on remote sensing satellite Technical Field The application relates to the technical field of satellites, in particular to a sea ice edge determining method and device based on remote sensing satellites. Background In the fields of offshore engineering, shipping safety and marine ecological protection, sea ice state monitoring has important practical significance. The sea ice in the coastal area in winter not only directly affects the port navigation, offshore operation and coastal infrastructure safety, but also changes the sea gas exchange process, and has profound effects on regional climate and marine ecosystem. Traditional monitoring means such as manual observation, shipborne radar or buoy station and the like are limited by factors such as narrow space coverage, low data updating frequency, difficult operation in severe weather and the like, and the large-range, high-precision, continuous and dynamic business monitoring requirements are difficult to realize. With the rapid development of satellite remote sensing technology, the satellite remote sensing technology has the advantages of macroscopic, objective and high-frequency observation, and provides a reliable technical approach for real-time inversion of sea ice range, type, thickness and motion parameters. In the prior art, the edge detection can be performed on remote sensing images acquired by satellites aiming at the ocean surface through traditional edge detection operators (such as Sobel and Canny), so that the edge of ocean surface sea ice can be determined, and further the areas where sea water and sea ice are respectively located can be determined. For the existing sea ice monitoring technology, the following disadvantages exist when the edge of the sea ice is extracted: (1) When the satellite passes through the border, the sun is at an oblique angle. The same sea ice has very bright sunny side (high gray scale) and relatively dark shady side (medium gray scale). While the adjacent sea water may also appear brighter in certain areas because the waves reflect sunlight. Thereby possibly erroneously classifying sea ice (gray scale value, etc.) on the back-to-back surface as sea water. Sea water (highlight region) that is glistened may also be erroneously judged as sea ice. Thus the extracted "sea ice edge line" becomes jagged, discontinuous, exaggerated toward the male surface edge and lost toward the female surface edge. (2) The sea ice surface is not uniform. It may be covered with white snow (very high grey values) or may have a "pool" of melted ice water (very low grey values, like sea water). Traditional edge detection operators (e.g., sobel, canny) will find places in the image where grey scale is abrupt. Now, it finds not only a high gradient at the sea water-sea ice interface, but also the same or even higher gradient at the snow-melt pool interface, even at the ice cracks. The algorithm may erroneously extract the edges of these internal textures on the ice surface as well, misidentifying the outer edge lines of sea ice. This can create a large number of "false edges" that make the true sea ice contour illegible. (3) At the sea ice edge there is typically a transition zone within which crushed ice, floating ice and sea water mix together. One pixel (pixel) of a satellite sensor may contain both ice and water. Therefore, the accuracy of edge detection of the remote sensing image by the traditional edge detection operator is low. Aiming at the technical problem that the accuracy of edge detection on a remote sensing image by a traditional edge detection operator in the prior art is low, no effective solution is proposed at present. Disclosure of Invention The embodiment of the disclosure provides a sea ice edge determining method, device and storage medium based on a remote sensing satellite, which at least solve the technical problem that in the prior art, the accuracy of edge detection on a remote sensing image by a traditional edge detection operator is low. According to one aspect of the embodiment of the disclosure, a method for determining sea ice edges based on remote sensing satellites is provided, and the method comprises the steps of obtaining remote sensing images acquired by satellites aiming at sea surfaces, determining first images corresponding to the remote sensing images, obtaining actual sea ice edge images corresponding to the first images, determining horizontal gradient diagrams corresponding to the first images and vertical gradient diagrams corresponding to the first images, generating first sea ice edge prediction images through a first generation network according to the first images, the horizontal gradient diagrams and the vertical gradient diagrams, inputting the actual sea ice edge images and the first images into the first discrimination network, obtaining first discrimination results corresponding to the actual sea ice edge images, inputting the first sea ice edge