US 12,008,801 B1
Tracking and identification method, device, electronic device, and storage medium for multiple vessel targets
Wen Liu, Wuhan (CN); Jingxiang Qu, Wuhan (CN); Yu Guo, Wuhan (CN); Mengwei Bao, Wuhan (CN); Chenjie Zhao, Wuhan (CN); and Jingxian Liu, Wuhan (CN)
Assigned to WUHAN UNIVERSITY OF TECHNOLOGY, Wuhan (CN)
Filed by WUHAN UNIVERSITY OF TECHNOLOGY, Wuhan (CN)
Filed on Oct. 27, 2023, as Appl. No. 18/495,776.
Claims priority of application No. 202310387654.3 (CN), filed on Apr. 12, 2023.
Int. Cl. G06V 10/80 (2022.01); B63B 79/00 (2020.01); G06T 7/246 (2017.01); G06T 7/73 (2017.01); G06V 10/74 (2022.01); G06V 10/82 (2022.01); G06V 20/40 (2022.01); G06V 20/52 (2022.01)
CPC G06V 10/806 (2022.01) [B63B 79/00 (2020.01); G06T 7/248 (2017.01); G06T 7/74 (2017.01); G06V 10/761 (2022.01); G06V 10/82 (2022.01); G06V 20/41 (2022.01); G06V 20/52 (2022.01); G06T 2207/10016 (2013.01); G06T 2207/20081 (2013.01); G06T 2207/20084 (2013.01); G06T 2207/30232 (2013.01); G06T 2207/30241 (2013.01); G06T 2207/30244 (2013.01); G06V 2201/07 (2022.01)] 11 Claims
OG exemplary drawing
 
1. A tracking and identification method for multiple vessel targets, comprising:
obtaining video surveillance data and initial AIS data;
filtering the initial AIS data to obtain effective AIS data;
based on the effective AIS data, determining a current position of a vessel, based on the video surveillance data, obtaining parameters of a camera, based on the parameters of the camera, combined with a pinhole imaging model, projecting the current position of the vessel to an image coordinate system corresponding to the video surveillance data; extracting motion characteristics of the effective AIS data in the video surveillance data from projected data, and obtaining a visual motion trajectory of the vessel based on the motion characteristics of the effective AIS data in the video surveillance data;
inputting the video surveillance data into a target detection network to obtain target detection boxes corresponding to multiple vessels; the target detection network is obtained based on a training of Yolov5-s network;
determining an occluded area based on the target detection boxes corresponding to the multiple vessels at a previous time, after deleting a part of the target detection boxes corresponding to the multiple vessels at the current time that falls into the occluded area, determining a prediction detection box of the occluded area based on the visual tracking trajectory of the vessel at the previous time, and determining real-time appearance features based on appearance features of the prediction detection box before occlusion;
inputting the prediction detection box, the real-time appearance features, and the target detection boxes under a non-occlusion state at the current time into a preset DeepSORT algorithm model, and carrying out an anti-occlusion tracking of the vessel based on the preset DeepSORT algorithm model to obtain a visual tracking trajectory of the vessel at the current time;
based on the visual motion trajectory and the visual tracking trajectory of the vessel at the current time, fusing the effective AIS data corresponding to the multiple vessels into the video surveillance data to determine identities of the multiple vessels;
wherein, the parameters of the camera include position of the camera, orientation of the camera, horizontal field angle of the camera, vertical field angle of the camera, height of the camera from water surface, and resolution of the video surveillance data taken by the camera.