基于海天线检测的船载摄像机电子稳像技术  

Electronic image stabilization technology of shipborne camera based on sea-sky line detection

作　　者：单雄飞 赵德鹏[1] 潘明阳[1] 王德强[1] SHAN Xiong-fei;ZHAO De-peng;PAN Ming-yang;WANG De-qiang(Navigation College,Dalian Maritime University,Dalian 116026,China)

机构地区：[1]大连海事大学航海学院,辽宁大连116026

出　　处：《大连海事大学学报》2020年第3期76-86,共11页Journal of Dalian Maritime University

基　　金：国家自然科学基金资助项目(61772102);中央高校基本科研业务费专项资金资助项目(3132019400)。

摘　　要：在海天背景下,由于复杂海况以及船舶六自由度运动,船载摄像机难以获取稳定的视频序列,不利于海上目标探测.针对这一问题,通过将船载摄像机与惯性传感器数据相融合,提出了一种基于海天线检测的电子稳像算法.算法主要包括三个步骤,即海天线估算、海天线检测以及点线模型.首先,通过构建摄像机运动模型,估算出海天线在图像中的候选区域;然后,在候选区域内应用边缘检测以及霍夫变换提取最优海天线;最后,在海天线附近区域应用点线模型对原始视频进行稳像,获取稳定的视频序列.试验中,构建了与摄像机运动姿态相匹配的实船数据集,通过对比试验验证了海天线检测算法的有效性.试验结果表明,本文算法的稳像效果明显优于其他算法,这为无人驾驶船舶的视觉发展提供了可靠的技术支持.In the background of sea and sky,due to the complex sea conditions and the ship’s six-degree-of-freedom motion,it is difficult for the shipborne camera to obtain stable video sequence,which is not conducive to the detection of sea targets.Aiming at this problem,an electronic image stabilization algorithm based on sea-sky line(SSL)detection was proposed by fusing data of shipbrone cameras and inertial sensors.The algorithm consists of three steps,namely SSL estimation,SSL detection and point-line model(PLM).Firstly,the camera motion attitude model was constructed to estimate the candidate region of SSL in the image;then,the edge detection and Hough transform were applied to extract the optimal SSL in the candidate region;finally,the PLM was applied to stabilize the original video in the area near the SSL,and the stable video sequence was obtained.In the experiment,the real ship data set matched with the camera’s motion attitude was constructed,and the effectiveness of the SSL detection algorithm was verified by comparative experiments.The experiment results show that the image stabilization effect is significantly better than other algorithms,which provides reliable technical support for the visual development of unmanned ships.

关 键 词：船舶 六自由度运动 船载摄像机 惯性传感器 电子稳像技术 运动姿态模型 边缘检测 霍夫变换 点线模型(PLM) 

分 类 号：U654[交通运输工程—港口、海岸及近海工程]