船舶航行过程中，受到风、浪、流等自然因素，以及船舶自身运动特性等因素的影响，导致船舶偏离预定航迹，影响航行的安全性和稳定性，为了有效控制船舶航迹，保证船舶的运行安全，提出一种基于光学图像分类的船舶航迹自抗扰检测技术。使用卷积神经网络提取光学图像中的船舶特征，并把船舶导航目标的光学图像划分成大、小导航目标切片，并基于大导航目标特征构建第一层SVM分类器训练漏检大、小导航目标数据集，形成第二、三层SVM分类器。利用该分类器对挖掘到特征参数进行逐层的剔除筛选，检测漏检大、小导航目标，最终对船舶航迹中的光学图像数据进行识别和分类；在此基础上，利用自抗扰技术结合高斯核映射，捕捉船舶航迹的复杂变化，实现船舶航迹自抗扰检测。实验结果表明，应用该方法能有效区分目标光学图像，并可以在干扰下较稳定的完成船舶航迹检测，从而确保船舶航行的安全和稳定。
During the navigation process of ships, natural factors such as wind, waves, and currents, as well as the movement characteristics of the ship itself, can cause the ship to deviate from the predetermined trajectory, affecting the safety and stability of navigation. In order to effectively control the ship's trajectory and ensure the safe operation of the ship, a ship's trajectory self disturbance detection technology based on optical image classification is proposed. Using convolutional neural networks to extract ship features from optical images, and dividing the optical images of ship navigation targets into large and small navigation target slices, the first layer SVM classifier is constructed based on the large navigation target features to train missed large and small navigation target datasets, forming the second and third layers of SVM classifiers. Using this classifier to remove and filter the mined feature parameters layer by layer, detecting missed large and small navigation targets, and ultimately identifying and classifying optical image data in ship trajectories; On this basis, using self disturbance rejection technology combined with Gaussian kernel mapping, complex changes in ship trajectories are captured to achieve self disturbance rejection detection of ship trajectories. The experimental results show that the method can effectively distinguish the target optical image, and can complete the ship track detection under interference, so as to ensure the safety and stability of ship navigation.
2024,46(15): 164-168 收稿日期：2024-01-22
DOI：10.3404/j.issn.1672-7649.2024.15.029
分类号：TN957.51
基金项目：重庆市教委基金项目（22SKGH493）
作者简介：孙宝刚（1979 – ），男，本科，副教授，研究方向为数据挖掘与分析、深度学习
参考文献：
[1] 何春城, 曹伟, 管志强. 一致性检测和航向编码的船舶航迹聚类[J]. 计算机仿真, 2023, 40(5): 318-322.
HE Chuncheng, CAO Wei, GUAN Zhiqiang. Consistency detection and vessels trajectory clustering based on course coding[J]. Computer Simulation, 2023, 40(5): 318-322.
[2] 张逸宸, 水鹏朗, 廖沫. 目标回波特征辅助的海面多目标跟踪方法[J]. 西安电子科技大学学报(自然科学版), 2023, 50(5): 1-10.
ZHANG Yichen, SHUI Penglang, LIAO Mo. Sea-surface multi-target tracking method aided by target returns features[J]. Journal of Xidian University (Natural Science), 2023, 50(5): 1-10.
[3] 薄钧天, 于洪波, 张翔宇, 等. 基于峰值自寻的多目标检测前跟踪算法[J]. 系统工程与电子技术, 2023, 45(10): 3083-3090.
BO Jun tian, YU Hongbo, ZHANG Xiangyu, et al. Multi-target track before detect algorithm based on peak self-seeking[J]. Systems Engineering and Electronics, 2023, 45(10): 3083-3090.
[4] 古毅杰, 张闯, 房美含. 基于输入延迟神经网络的船舶GPS/INS组合导航[J]. 船舶工程, 2022, 44(7): 96-102.
GU Yijie, ZHANG Chuang, FANG Meihan. Ship GPS/INS integrated navigation based on input delay neural network[J]. Ship Engineering, 2022, 44(7): 96-102.
[5] 贾翊文, 荆文龙, 杨骥, 等. 基于深度学习的SAR影像海洋涡旋检测算法对比分析[J]. 海洋科学进展, 2024, 42(1): 137-148.
JIA Yiwen, JING Wenlong, YANG Ji, et al. Comparative analysis of ocean eddy detection algorithms based on deep learning in SAR images[J]. Advances in Marine Science, 2024, 42(1): 137-148.
[6] 张天文, 张晓玲. 一种大场景SAR图像中舰船船舶检测虚警抑制方法[J]. 现代雷达, 2022, 44(2): 1-8.
ZHANG Tianwen, ZHANG Xiaoling. A false alarm suppression method for ship detection in large-scene SAR image[J]. Modern Radar, 2022, 44(2): 1-8.
[7] 薛安克, 毛克成, 张乐. 多分类器联合虚警可控的海上小目标检测方法[J]. 电子与信息学报, 2023, 45(7): 2528-2536.
XUE Anke, MAO Kecheng, ZHANG Le. Multi-feature marine small target detection based on multi-class classifier[J]. Journal of Electronics & Information Technology, 2023, 45(7): 2528-2536.
[8] 施赛楠, 姜丽, 李东宸, 等. 基于双重虚警控制XGBoost的海面小目标检测[J]. 雷达科学与技术, 2023, 21(3): 314-323+328.
SHI Sainan, JIANG Li, LI Dongchen, et al. Sea-surface small target detection based on XGBoost with dual false alarm control[J]. Radar Science and Technology, 2023, 21(3): 314-323+328.
