CN-121982565-A - Ship behavior discrimination method and device based on satellite-borne SAR image

CN121982565ACN 121982565 ACN121982565 ACN 121982565ACN-121982565-A

Abstract

The invention discloses a ship behavior discriminating method and device based on a satellite-borne SAR image, and belongs to the technical field of signal processing. The method comprises the steps of slicing a satellite-borne SAR image to be judged, obtaining a scene classification result and a ship target detection result of the slice, deleting the ship target detection result corresponding to the slice with the scene classification result being inland, preliminarily determining that the behavior of the ship target in the slice is sailing when the scene classification result is a marine slice, preliminarily determining that the behavior of the ship target in the slice is sailing or berthing when the scene classification result is a sea-land handover slice, and determining that the non-parallel vector distance and the safety threshold value between any two ship targets are determined when the non-parallel vector distance and the safety threshold value between any two ship targets are smaller than the safety threshold value when the scene classification result is the marine slice, otherwise, finally determining that the behavior of the two ship targets is sailing. The invention can judge the behavior of the ship target.

Inventors

ZENG HONGCHENG
PAN YONGCHEN
YANG WEI
WANG YAMIN
CHEN JIE

Assignees

北京航空航天大学

Dates

Publication Date: 20260505
Application Date: 20260120

Claims (10)

1. The ship behavior discriminating method based on the satellite-borne SAR image is characterized by comprising the following steps of: acquiring a satellite-borne SAR image to be discriminated, and slicing the satellite-borne SAR image to be discriminated to obtain a plurality of slices to be discriminated; Inputting the slices to be distinguished into a pre-trained scene classification network and a ship detection network aiming at each slice to be distinguished to obtain respectively output scene classification results and ship target detection results, wherein the scene classification results comprise one of inland, sea Liu Jiaojie and offshore; Deleting the ship target detection result corresponding to the inland slicing of the scene classification result; aiming at a scene classification result which is a marine slice, if a ship target exists in a ship target detection result of the slice, primarily determining that the behavior of the ship target in the slice is sailing; aiming at a slice with a scene classification result of sea-land handover, if a ship target exists in a ship target detection result of the slice, primarily determining that the behavior of the ship target in the slice is sailing or berthing; And determining the non-parallel vector distance between any two ship targets and the safety threshold according to the ship targets preliminarily determined to be sailing in the satellite-borne SAR image to be distinguished, if the non-parallel vector distance is smaller than the safety threshold, determining the two ship targets as temporary replenishment, otherwise, finally determining the behavior of the two ship targets as sailing.
2. The method of claim 1, wherein for a slice in which the scene classification result is a sea-land handover, if a ship target exists in the ship target detection result of the slice, the behavior of the ship target in the slice is primarily determined to be sailing or berthing by: mapping the existing ship targets into the satellite-borne SAR image to be distinguished, and aiming at the ship targets with sea-land handover as the scene classification result in the satellite-borne SAR image to be distinguished, executing the following operations: taking the central position of the rotating boundary frame of the ship target as the center, taking the designated multiple of the ship target length as the side length, and generating a new target slice; Classifying the new target slice by using the scene classification network to obtain a scene classification result of the current slice; If the scene classification result of the current slice is sea Liu Jiaojie, preliminarily determining that the behavior of the ship target in the current slice is berthing; And if the scene classification result of the current slice is offshore, preliminarily determining that the behavior of the ship target in the current slice is navigation.
3. The method of claim 2, further comprising, prior to classifying the new target slice with the scene classification network: and (3) shielding other ship targets detected in the new target slice, generating a polygonal mask based on the rotating frame, and filling other ship target areas with zero values to obtain a current slice to be classified.
4. The method of claim 1, wherein the non-parallel vector distance between any two ship targets is calculated by: Respectively calculating transverse vectors aiming at each of the current two ship targets; constructing a center vector of each of the current two ship targets based on the center point coordinates of each of the current two ship targets; And determining the dot product of the center vector and the transverse vector of one of the ship targets as the non-parallel vector distance between the two current ship targets.
5. The method of claim 4, wherein the safety threshold is determined by determining a maximum of the vessel lengths of the current two vessel targets as the safety threshold of the current two vessel targets based on a suction effect.
6. The method of any of claims 1-5, further comprising, after said determining that the two vessel targets are temporary supplies: Determining the captain of each of the two ship targets; And determining the ship targets with larger ship length as the supplied ships, and determining the ship targets with smaller ship length as the supplied ships.
7. The utility model provides a naval vessel action discriminating apparatus based on spaceborne SAR image which characterized in that, the device includes: The first acquisition unit is used for acquiring a satellite-borne SAR image to be distinguished and slicing the satellite-borne SAR image to be distinguished to obtain a plurality of slices to be distinguished; The second acquisition unit inputs the slices to be discriminated into a pre-trained scene classification network and a ship detection network for each slice to be discriminated so as to acquire a scene classification result and a ship target detection result which are respectively output, wherein the scene classification result comprises one of inland, sea Liu Jiaojie and sea; The first processing unit is used for deleting the ship target detection result corresponding to the inland slicing of the scene classification result; for the section with the scene classification result being sea-land handover, if the ship target detection result of the section has the ship target, the behavior of the ship target in the section is primarily determined to be sailing or berthing; the second processing unit is used for determining the size of the non-parallel vector distance and the safety threshold between any two ship targets according to the ship targets preliminarily determined to be sailing in the satellite-borne SAR image to be distinguished, judging that the two ship targets are temporarily replenished if the non-parallel vector distance is smaller than the safety threshold, and finally judging that the behaviors of the two ship targets are sailing if the non-parallel vector distance is smaller than the safety threshold.
8. A computer device, characterized in that it comprises a memory for storing a computer program and a processor for executing the computer program stored on the memory for carrying out the steps of the method according to any of the preceding claims 1-6.
9. A computer-readable storage medium, characterized in that the storage medium has stored therein a computer program which, when executed by a processor, implements the steps of the method of any of claims 1-6.
10. A computer program product comprising a computer program which, when executed by a processor, implements the steps of the method of any of claims 1-6.

Description

Ship behavior discrimination method and device based on satellite-borne SAR image Technical Field The invention relates to the technical field of signal processing, in particular to a ship behavior judging method and device based on a satellite-borne SAR image. Background The ship is used as a key way of marine transportation, and the grasping of ship information is an important means for guaranteeing the safety of the ocean. Traditional ship monitoring mainly relies on a ship automatic identification system (Automatic Identification System, AIS), and basic information such as position, heading, speed and the like can be provided through a radio signal actively transmitted by a ship, so that the method plays an important role in civil shipping management. However, the AIS data can only reflect basic dynamic parameters of individual ships, cannot directly describe interaction behaviors among ships, and may cause signal loss, information principle errors and even risk of data tampering due to equipment failure, manual closing or environmental interference, so that high-reliability behavior research and judgment are difficult to support. SAR is mostly used for acquiring sea surface image information in ocean scenes due to the all-day and all-weather working capacity of the SAR, and ship target detection is carried out. In recent years, as resolution is continuously increased from 10m to 3m and 1m, SAR, and detailed information of the ship which can be excavated in the image is remarkably increased. However, automatic judgment of ship behaviors is not yet achieved. Disclosure of Invention The invention provides a ship behavior discriminating method and device based on a satellite-borne SAR image. The technical proposal is as follows: In one aspect, a ship behavior discrimination method based on a satellite-borne SAR image is provided, and the method comprises the following steps: acquiring a satellite-borne SAR image to be discriminated, and slicing the satellite-borne SAR image to be discriminated to obtain a plurality of slices to be discriminated; Inputting the slices to be distinguished into a pre-trained scene classification network and a ship detection network aiming at each slice to be distinguished to obtain respectively output scene classification results and ship target detection results, wherein the scene classification results comprise one of inland, sea Liu Jiaojie and offshore; Deleting the ship target detection result corresponding to the inland slicing of the scene classification result; aiming at a scene classification result which is a marine slice, if a ship target exists in a ship target detection result of the slice, primarily determining that the behavior of the ship target in the slice is sailing; aiming at a slice with a scene classification result of sea-land handover, if a ship target exists in a ship target detection result of the slice, primarily determining that the behavior of the ship target in the slice is sailing or berthing; And determining the non-parallel vector distance between any two ship targets and the safety threshold according to the ship targets preliminarily determined to be sailing in the satellite-borne SAR image to be distinguished, if the non-parallel vector distance is smaller than the safety threshold, determining the two ship targets as temporary replenishment, otherwise, finally determining the behavior of the two ship targets as sailing. On the other hand, a ship behavior discriminating device based on a satellite-borne SAR image is provided, and the device comprises: The first acquisition unit is used for acquiring a satellite-borne SAR image to be distinguished and slicing the satellite-borne SAR image to be distinguished to obtain a plurality of slices to be distinguished; The second acquisition unit inputs the slices to be discriminated into a pre-trained scene classification network and a ship detection network for each slice to be discriminated so as to acquire a scene classification result and a ship target detection result which are respectively output, wherein the scene classification result comprises one of inland, sea Liu Jiaojie and sea; The first processing unit is used for deleting the ship target detection result corresponding to the inland slicing of the scene classification result; for the section with the scene classification result being sea-land handover, if the ship target detection result of the section has the ship target, the behavior of the ship target in the section is primarily determined to be sailing or berthing; the second processing unit is used for determining the size of the non-parallel vector distance and the safety threshold between any two ship targets according to the ship targets preliminarily determined to be sailing in the satellite-borne SAR image to be distinguished, judging that the two ship targets are temporarily replenished if the non-parallel vector distance is smaller than the safety threshold, and finally jud