近年来，越来越多国家和地区将发展的眼光聚焦于资源丰富的海洋。加快无人船相关技术的研究对我国提升海洋装备水平有着重大的战略意义。在无人船运行过程中，经常会遇到低对比度的水面图像，在对其进行增强时，传统的图像增强算法需要人工设置参数，且运行效率低。针对以上问题，本文用卷积神经网络重构了MSR算法，提出MSRN模块，使用深度学习方法替代人工参数。同时引入通道级视觉注意力机制与编解码结构，提出MSSEN模型。实验表明，MSSEN模型对低对比度水面图像增强效果明显。除此之外，基于MSSEN网络，本文提出低级视觉任务与高级视觉任务结合的框架，以端到端的形式完成图像增强与水面目标检测识别的任务，提升了整个算法的性能与效率。
In recent years, more countries have focused on the ocean area. Accelerating the research on unmanned boat-related technologies has great significance for our country to improve the level of marine equipment. During the sailing of the unmanned boat, low-contrast water surface images are often encountered. When enhancing the low-contrast image, the traditional algorithm needs to manually set parameters with low efficiency. To solve the problems above, this paper reconstructs MSR algorithm with convolutional neural network and proposes MSRN model. This paper introduce channel-level visual attention mechanism and encoder-decoder module to propose MSSEN. Experiments show that MSSEN model has obvious effect on low-contrast water surface image enhancement. In addition, based on MSSEN network, this paper proposes a framework for low-level and high-level visual tasks combination. This framework completes image enhancement and water surface target detection in an end-to-end form. It improves the performance and efficiency of the entire algorithm.
2023,45(14): 83-89 收稿日期：2022-7-28
DOI：10.3404/j.issn.1672-7649.2023.14.015
分类号：TP751
作者简介：赵瑞祥(1993-),男,硕士,助理工程师,主要从事航海导航及设备研发工作。
参考文献：
[1] 王俊平, 李锦. 图像对比度增强研究的进展[J]. 电子科技, 2013, 26(5): 160–165
WANG J P, LI J. Research progress of image contrast enhancement[J]. Electronic Science and Technology, 2013, 26(5): 160–165
[2] 李力, 金伟其, 黄有为, 等. 一种低对比度可见光超视距成像的实时图像增强方法[J]. 兵工学报, 2010, 31(2): 242–247
LI L, JIN W Q, HUANG Y W, et al. A real-time image enhancement method for low-contrast visible light over-the-horizon imaging[J]. Journal of Military Engineering, 2010, 31(2): 242–247
[3] SZEGEDY C, LIU W, JIA Y, et al. Going deeper with convolutions[C]// Boston: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2015: 1–9.
[4] HE K, ZHANG X, REN S, et al. Deep residual learning for image recognition[C]// Las Vegas: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2016: 770–778.
[5] LAND, EDWIN H. The retinex theory of color vision[J]. Scientific American, 1977, 237(6): 108–128
[6] RUMELHART D E, HINTON G E, WILLIAMS R J. Learning representations by back-propagating errors[J]. Nature, 1986, 323(6088): 533–536
[7] HU J, SHEN L, SUN G. Squeeze-and-excitation networks[C]// Salt Lake City: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2018: 7132–7141.
[8] SIMONYAN K, ZISSERMAN A. Very deep convolutional networks for large-scale image recognition[J]. Computer Science, 2014.
[9] LONG J, SHELHAMER E, DARRELL T. Fully convolutional networks for semantic segmentation[C]// Boston: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2015: 3431–3440.
[10] MALIK J. Contour detection and hierarchical image segmentation[J]. IEEE Transactions on Pattern Analysis & Machine Intelligence, 2011, 33(5): 898–916
[11] GERALD SCHAEFER M S. UCID - an uncompressed colour image database[J]. Storage & Retrieval Methods & Applications for Multimedia, 2003
[12] KINGMA D P, BA J. Adam: a method for stochastic optimization[J]. Computer Science, 2014.
[13] 苏正青, 马巧梅. 基于卷积神经网络的交通标志模糊影像识别[J]. 计算机仿真, 2020, 37(1): 117–120
SU Z Q, MA Q M. Traffic sign blurred image recognition based on convolutional neural network[J]. Computer Simulation, 2020, 37(1): 117–120
