利用高分辨率地球同步轨道卫星对海洋区域重复观测的特点，本文提出一种舰船运动目标检测算法，通过快速区域提取算法，提取云掩膜与水体的NDWI（Normalized Difference Water Index），基于梯度算法对海洋背景进行建模，随后使用深度学习的方法，将多尺度特征融合目标检测技术与基于上下文分析的模型相结合，得到时序影像中舰船目标位置信息。本研究同时利用真实目标和仿真目标相结合的技术，验证了算法的可行性。在50 m分辨率的高分四序列影像实验中表明，本研究提出的算法可以有效提取多尺度运动舰船目标的位置和状态信息，有效为舰船目标检测和跟踪领域提供技术支撑。该算法在工程实践领域中具有较高的应用价值。
Taking advantage of the characteristics of repeated observation of ocean areas by high-resolution geosynchronous orbit satellites, this paper proposes a ship target detection algorithm, which extracts the NDWI (Normalized Difference Water Index) and cloud mask through a fast area extraction algorithm, and models the ocean background based on the gradient algorithm. Then, the multi-scale feature fusion target detection technology is combined with the target detection model based on context analysis to obtain the target position information of the ship. In this study, the feasibility of the algorithm is verified by using the combination of real target and simulated target. In the 50 m resolution high-resolution four-sequence image experiment, it is shown that the algorithm proposed in this study can effectively extract the position and status information of ship targets of various sizes, and effectively provide technical support for the field of ship target detection and tracking. This algorithm has strong engineering application value.
2025,47(19): 190-196 收稿日期：2025-1-17
DOI：10.3404/j.issn.1672-7649.2025.19.031
分类号：TP751
作者简介：郑义成(1976-),男,硕士,高级工程师,研究方向为目标特性
参考文献：
[1] YU Y, YANG J. Visual saliency using binary spectrum of walsh - hadamard transf -orm and its applications to ship detection in multispectral imagery[J]. Neural Processing Letters, 2017, 45(3): 759-776.
[2] 路威, 余旭初, 马永刚, 等. 高光谱遥感影像海上舰船目标检测算法的研究[J]. 海洋测绘, 2005, 25(4): 8-12.
[3] 李明, 赵俊霞, 胡芬. 国家航空航天遥感影像获取现状及发展[J]. 测绘通报, 2015(10): 12-15.
[4] WANG Y, ZHOU X, ZUO Y, et al. A comparative study of moving target detection algorithms[J]. IOP Conference Series: Materials Science and Engineering, 2020, 790(1): 012061.
[5] CHEN Z, WANG Y, LIU X, et al. Physical model-driven deep learning for single image dehazing[J]. IEEE Transactions on Neural Networks and Learning Systems, 2023, 34(5): 2314-2326.
[6] 张振. 高分辨率可见光遥感图像港口及港内目标识别方法研究[D]. 合肥: 中国科学技术大学, 2009.
[7] LETU H, YANG K, NAKAJIMA TY, et al. High-resolution retrieval of cloud microphysical properties and surface solar radiation using Himawari-8/AHI next-generation geostationary satellite[J]. Remote Sensing of Environment, 2020, 239: 1-16.
[8] XI Q, ZHANG Z, PENG L. Small object detection algorithm based on improved dense network and quadratic regression[J]. Computer Engineering, 2021, 47(4): 241-247+255.
[9] DANG Z, SUN B, LI C, et al. CA - LSTM: An improved LSTM trajectory prediction method based on infrared UAV target detection[J]. Electronics, 2023, 12(19): 4081.
[10] ZHANG Y, WANG X, LIU X, et al. Multi - object tracking with LSTM-based post-processing for UAV-based video surveillance[J]. Journal of Intelligent & Robotic Systems, 2021, 102(1-2): 1-14.
