船舶航行环境复杂多变，多模态视觉相机采集的多源视觉信息融合过程繁杂，特征提取不够精细，为此研究光视觉信息自适应聚合的船舶自主避碰决策方法。利用Kinect相机采集多源图像，通过自适应聚合增强环境感知。结合Mask RCNN与ResNet-101，精确提取图像信息中的目标特征，经RPN生成候选区域。特征向量进入FCN与分类回归分支，识别目标并定位，转换坐标后获取运动态势。依据碰撞风险计算和船舶操纵经验，智能输出避碰指令。实验结果表明，该方法在复杂航行环境下展现出良好的避碰决策性能，光视觉信息聚合后，交并比、F1 score分别为92.7%、94.2%。
The navigation environment of ships is complex and ever-changing, and the fusion process of multi-source visual information collected by multimodal visual cameras is complicated. The feature extraction is not precise enough. Therefore, a ship autonomous collision avoidance decision-making method based on adaptive aggregation of optical visual information is studied. Use Kinect camera to capture multi-source images and enhance environmental perception through adaptive aggregation. Combining Mask RCNN and ResNet-101, accurately extract target features from image information and generate candidate regions through RPN. The feature vector enters the FCN and classification regression branches, identifies the target and locates it, and obtains the motion situation after coordinate transformation. Based on collision risk calculation and ship maneuvering experience, intelligently output collision avoidance commands. The experimental results show that this method exhibits good collision avoidance decision-making performance in complex navigation environments. After the aggregation of optical visual information, the intersection to union ratio and F1 score are 92.7% and 94.2%, respectively.
2024,46(19): 165-169 收稿日期：2024-6-17
DOI：10.3404/j.issn.1672-7649.2024.19.030
分类号：U675.96
作者简介：张博文(1997-),男,助理工程师,研究方向为航海保障与智能航运
参考文献：
[1] 杨琪森, 王慎执, 桑金楠, 等. 复杂开放水域下智能船舶路径规划与避障方法[J]. 计算机集成制造系统, 2022, 28(7): 2030-2040.
[2] 徐言民, 律建辉, 刘佳仑, 等. 基于CSSOA的多船智能避碰决策研究[J]. 中国舰船研究, 2023, 18(6): 88-96.
XU Yanmin, LV Jianhui, LIU Jialun, et al. Multi-vessel intelligent collision avoidance decision-making based on CSSOA[J]. Chinese Journal of Ship Research, 2023, 18(6): 88-96.
[3] 庹玉龙, 康彩霞, 王莎莎, 等. 未知干扰下多船相邻交叉耦合同步编队避障控制[J]. 交通运输工程学报, 2023, 23(6): 314-326.
TUO Yulong, KANG Caixia, WANG Shasha, et al. Adjacent cross-coupling synchronous formation control with collision avoidance for multiple ships under unknown disturbances[J]. Journal of Traffic and Transportation Engineering, 2023, 23(6): 314-326.
[4] 关巍, 崔哲闻, 罗文哲. 基于改进PPO算法的船舶自主避碰决策[J]. 大连海事大学学报, 2023, 49(4): 28-36.
GUAN Wei, CUI Zhewen, LUO Wenzhe. Ship autonomous collision avoidance decision based on improved PPO algorithm[J]. Journal of Dalian Maritime University, 2023, 49(4): 28-36.
[5] ZHENG Yisong, ZHANG Xiuguo, SHANG Zijing, et al. A decision-making method for ship collision avoidance based on improved cultural particle swarm[J]. Journal of Advanced Transportation, 2021, 2021(1): 1-31.
[6] 谢鸿伟, 张英俊, 邢胜伟, 等. 基于模型预测控制的船舶自主避碰方法[J]. 船舶工程, 2021, 43(8): 23-28+95.
XIE Hongwei, ZHANG Yingjun, XING Shengwei, et al. A method for ship autonomous collision avoidance based on model predictive control[J]. Ship Engineering, 2021, 43(8): 23-28+95.
[7] 徐茂竹, 李弘, 李亚光, 等. 单目视觉引导下的无人艇局部避障方法[J]. 重庆邮电大学学报(自然科学版), 2023, 35(4): 732-741.
[8] 张芯睿, 赵清华, 王雷, 等. 基于Mask R-CNN的雾天场景目标检测[J]. 电光与控制, 2022, 29(12): 83-88.
