基于侧扫声纳图像的AUV侧向速度估计与DVL粗差检测  

AUV Lateral Speed Estimation Based on Side Scan Sonar Image and DVL Gross Error Detection

作　　者：李凯[1] 吴文启[1] 张鹭[1] LI Kai;WU Wen-qi;ZHANG Lu(College of Mechatronics and Automation,National University of Defense Technology,Changsha 410073)

机构地区：[1]国防科学技术大学机电工程与自动化学院,长沙410073

出　　处：《导航与控制》2018年第4期72-80,共9页Navigation and Control

摘　　要：复杂水底环境和自主潜航器(AUV)机动会造成Doppler测速仪(DVL)速度测量结果中出现粗差,提出一种检测DVL速度粗差数据的方法。该方法利用侧扫声纳图像数据,结合SIFT特征匹配算法求取声纳图像的边缘特征点。根据声纳图像成像原理,计算出每一帧图像相对上一帧图像对应特征点像素的横向平移量;结合特征像素平移量与AUV实际横向位移的比例尺度关系以及声纳侧扫周期推算出AUV侧向速度。该侧向速度用于检测水下组合导航中DVL的速度粗差,在几十米水深的浅水区,该方法能够检测并剔除出大于0.2m/s的DVL侧向速度粗差,显著减小DVL前向速度误差。算法可实时实现,改善AUV水下组合导航性能。Gross error may occur in Doppler velocity log（DVL） speed data due to complicated water bottom environments and autonomous underwater vehicles（AUV） maneuver.This paper presents a method to detect the gross error in AUV speed data from DVL. The method uses the side scan sonar image data and the SIFT feature matching algorithm to obtain the edge feature points in the sonar image. According to the sonar imaging principle, lateral translation of pixels in corresponding feature between each adjacent sonar image frame can be obtained. In this way, the lateral AUV speed can be calculated based on the scale factor between pixel translation and AUV lateral displacement in each side scan cycle. The calculated lateral velocity can be used to detect the gross error of the AUV speed from the DVL in the integrated navigation system. In tens of meters deep shallow water area, this method can detect and eliminate the gross error greater than 0. 2 m/s in DVL speed data,and significantly reduce the forward velocity error of DVL forward velocity. The algorithm can be realized in real-time to improve the AUV underwater integrated navigation performance.

关 键 词：侧扫声纳图像 AUV侧向速度估计 SIFT特征匹配算法 DVL速度粗差检测 

分 类 号：U666.11[交通运输工程—船舶及航道工程]