对船舶周围环境实时监控有利于提升航行的安全性，立体视觉技术通过计算图像之间的视差来重建目标的三维结构，但是三维视觉图像的构建严重依赖于图像质量。本文对暗通道分析技术和深度神经网络技术进行研究，分析图像去雾以及通过绘制环境场景进行还原的基本流程，提出一种基于MobileNet V2的轻量级雾浓度分类器，得到了不同训练周期时图像分类的训练准确率和验证准确率，对船舶监控三维视觉图像模糊目标清晰化处理进行验证试验，结果表明清晰化处理后的环境空间图像质量得到了大幅度提升。
Real-time monitoring of the surrounding environment is beneficial for enhancing the safety of navigation. Stereo vision technology reconstructs the three-dimensional structure of the target by calculating the disparity between images. However, the construction of three-dimensional visual images heavily relies on image quality. This paper studies dark channel analysis technology and deep neural network technology, analyzes the basic process of image defogging and restoration , and proposes a lightweight fog density classifier based on MobileNet V2. It presents the training accuracy and validation accuracy of image classification at different training epochs. Verification experiments are conducted on the clear processing of blurry targets in three-dimensional visual ship monitoring images, and the results show that the image quality has been significantly improved after the clarification treatment.
2024,46(12): 162-165 收稿日期：2023-12-27
DOI：10.3404/j.issn.1672-7649.2024.12.028
分类号：U667.65
基金项目：湖北省高等学校优秀中青年科技创新团队计划项目（T2018371）
作者简介：彭笛(1987-),女,硕士,讲师,研究方向为环境设计三维空间视觉图像制作及处理
参考文献：
[1] 钟思, 李碧青, 袁天然, 等. 视觉显著性和稀疏学习相融合的船舶图像目标检测[J]. 舰船科学技术, 2024, 46(8): 157-160.
ZHONG Si, LI Biqing, YUAN Tianran, et al. Research on Ship image target detection by integrating visual salience and sparse learning[J]. Journal of Ship Science and Technology, 2024, 46(8): 157-160.
[2] 苏白华宁, 刘畅, 王超. 基于参数估计与子孔径提取的船舶ISAR实时成像算法[J]. 中国科学院大学学报, 2023, 40(5): 647-657.
SU Baihuaning, LIU Chang, WANG Chao. Research on real-time imaging algorithm of ship ISAR based on parameter estimation and subaperture extraction[J]. Journal of the Graduate School of the Chinese Academy of Sciences, 2023, 40(5): 647-657.
[3] 张雪. 基于PSD和改进YOLOv5的雾天船舶检测方法研究[D]. 大连: 大连海事大学, 2023.
[4] 姜俐伶. 复杂场景船舶图像中的船名检测与识别研究[D]. 大连: 大连海事大学, 2023.
[5] 张梦瑶. 基于深度学习的复杂条件下船舶检测方法[D]. 哈尔滨: 哈尔滨师范大学, 2023.
[6] 周林宏, 杨戈, 李娜, 等. 基于自适应图像增强和图像去噪的水面航行船舶识别方法[J]. 船舶工程, 2021, 43(S2): 101-105.
ZHOU Linhong, YANG Ge, LI Na, et al. Research on recognition method of surface vessels based on adaptive image enhancement and denoising[J]. Ship Engineering, 2021, 43(S2): 101-105.
