基于先验信息的光照鲁棒性水下视觉定位方法  

A robust underwater visual positioning method for brightness based on prior information

作　　者：王嘉 关夏威 付少波 马哲松 WANG Jia;GUAN Xia-wei;FU Shao-bo;MA Zhe-song(Wuhan Second Ship Design and Research Institute,Wuhan 430205,China)

机构地区：[1]武汉第二船舶设计研究所,湖北武汉430205

出　　处：《舰船科学技术》2024年第10期98-101,共4页Ship Science and Technology

摘　　要：本文提出一种结合声学和光学定位手段的UUV水下回收方法,并重点对视觉定位进行研究。针对光照剧烈变化情况下视觉定位稳定性差的问题,设计一种基于先验信息的目标轮廓提取算法,充分利用环境亮度信息和合作目标的物理属性,通过筛选兴趣区(ROI)、动态阈值等方法,有效解决图像背景亮度过高或亮度不均衡导致的目标漏检、误检等问题,提升对目标轮廓的分割精度,从而优化目标图像坐标及空间坐标的计算准度。在湖验中应用该算法,实现全天时、全天候对UUV目标的可靠检测与准确定位,验证了该算法对环境亮度变化的鲁棒性。This paper proposes a method for UUV underwater recovery that combines acoustic and optical positioning methods,with a focus on visual positioning.Aiming at the problem of poor stability of visual positioning in the case of dra-matic change of brightness,a target contour extraction algorithm based on prior information is designed.It makes full use of the environmental brightness information and the physical attributes of the cooperative target.Solving the problems of target missing and false detection caused by the high or uneven brightness of the background by filtering the region of interest(ROI),dynamic threshold and other methods,so as to optimize the accuracy of target image coordinates and spatial coordin-ates.The algorithm is applied in the lake experiment to realize the reliable detection and accurate positioning of UUV targets,all-day and all-weather,which verifies that this algorithm is robust to change of environment brightness.

关 键 词：水下回收 视觉定位 先验信息 目标轮廓 环境光照 

分 类 号：TP242[自动化与计算机技术—检测技术与自动化装置]