为避免船舶在航行过程中的航线冲突，设计基于无线网络的船舶多目标检测系统。该系统由遥感图像采集单元和船舶多目标检测单元构成，利用遥感图像采集单元内的遥感平台获取可见光遥感图像，用户端利用ZigBee无线网络从遥感平台读取可见光遥感图像并缓存；船舶多目标检测单元利用遥感图像预处理子单元读取可见光遥感图像后，对其进行降采样和高斯滤波处理后，输送至海陆分离子单元；海陆分离子单元对可见光遥感图像实时海陆分离处理，然后使用特征提取子单元提取可见光遥感图像特征；以该可见光遥感图像特征为基础，使用基于卷积神经网络的船舶多目标检测方法完成船舶多目标检测。实验结果表明：该系统具备较强的无线通信能力的同时，提取船舶可见光遥感图像特征精度高达98.91%，且其检测船舶多目标不受船型大小和分布位置影响，具备显著的应用效果。
In order to avoid the route conflict of ships during navigation, a ship multi-target detection system based on wireless network is studied. The system is composed of remote sensing image acquisition unit and ship multi-target detection unit. The remote sensing platform in the remote sensing image acquisition unit is used to obtain the visible remote sensing image, and the client uses ZigBee wireless network to read and cache the visible remote sensing image from the remote sensing platform. The ship multi-target detection unit reads the visible remote sensing image by using the remote sensing image preprocessing sub unit, and then transmits it to the sea and land ion separation unit after down sampling and Gaussian filtering. The sea land ion separation unit separates the visible light remote sensing image in real time, and then uses the feature extraction sub unit to extract the features of the visible light remote sensing image. Based on the characteristics of the visible remote sensing image, the ship multi-target detection method based on convolutional neural network is used to complete the ship multi-target detection. The experimental results show that while the system has strong wireless communication ability, the accuracy of extracting the features of ship visible light remote sensing image is as high as 98.91%, and its detection of ship multi-target is not affected by the ship size and distribution position, so it has significant application effect.
2022,44(12): 137-140 收稿日期：2021-12-30
DOI：10.3404/j.issn.1672-7649.2022.12.027
分类号：TP391
基金项目：中国职业技术教育学会资助项目（1710434）
作者简介：陈天文(1979-)，男，本科，高级实验师，研究方向为软件工程、软件逆向工程、模式识别及算法、网络构建
参考文献：
[1] 颜军, 冯素云, 鹿琳琳, 等. 基于目标多特征的SAR影像舰船检测优化算法[J]. 计算机科学, 2021, 48(z1): 132–136,157
[2] 刘松涛, 姜康辉, 刘振兴. 基于区域协方差和目标度的航空侦察图像舰船目标检测[J]. 系统工程与电子技术, 2019, 41(5): 972–980
[3] 仇荣超, 娄树理, 李廷军, 等. 多波段红外图像的海面舰船目标检测[J]. 光谱学与光谱分析, 2019, 39(3): 698–704
[4] 董超, 刘晶红, 徐芳, 等. 光学遥感图像舰船目标快速检测方法[J]. 吉林大学学报(工学版), 2019, 49(4): 1369–1376
[5] 周慧, 刘振宇, 陈澎. 利用改进特征金字塔模型的SAR图像多目标船舶检测[J]. 电讯技术, 2020, 60(8): 896–901
[6] 段敬雅, 李彬, 董超, 等. 基于YOLOv2的船舶目标检测分类算法[J]. 计算机工程与设计, 2020, 41(6): 1701–1707
[7] 董众, 林宝军, 申利民. 一种基于图像显著性的离岸船舶目标检测效率优化方法[J]. 燕山大学学报, 2020, 44(4): 418–424
[8] 李健伟, 曲长文, 彭书娟, 等. 基于生成对抗网络和线上难例挖掘的SAR图像舰船目标检测[J]. 电子与信息学报, 2019, 41(1): 143–149
[9] 王明春, 张嘉峰, 杨子渊, 等. Beta分布下基于白化滤波的极化SAR图像海面舰船目标CFAR检测方法[J]. 电子学报, 2019, 47(9): 1883–1890
[10] 孙元辉, 徐智勇, 张建林, 等. 基于深度学习的单步目标检测器特征增强算法[J]. 半导体光电, 2019, 40(1): 108–111
[11] 李晖晖, 周康鹏, 韩太初. 基于CReLU和FPN改进的SSD舰船目标检测[J]. 仪器仪表学报, 2020, 41(4): 183–190
[12] 顾佼佼, 李炳臻, 刘克, 等. 基于改进Faster R-CNN的红外舰船目标检测算法[J]. 红外技术, 2021, 43(2): 170–178
