基于单目机器视觉的实时船舶测距方法及误差修正  

Real-time Ship Ranging Method and Error Correction Based on Monocular Machine Vision

作　　者：徐进 陈姚节[1] XU Jin;CHEN Yaojie(School of Computer,Wuhan University of Science and Technology,Wuhan 430081)

机构地区：[1]武汉科技大学计算机学院,武汉430081

出　　处：《计算机与数字工程》2022年第5期1006-1009,共4页Computer & Digital Engineering

摘　　要：测距功能是保障船舶在水上安全航行,极大程度上避免与周围障碍物及其他船舶之间碰撞的重要方式,而船舶摇摆一直是船舶测距的重难点。针对此问题,在传统单目机器视觉的基础上进行改进,提出一种实时船舶测距方法和误差修正方法。使用faster R-CNN算法画矩形框,ECO算法实现对目标的跟踪并将光电云台的光心对准目标的底部中心,通过陀螺仪采集船舶摇摆时的动态角度,三角几何原理进行实时测距。根据误差源对测距结果进行修正。实验表明,在100m内,船舶静态测距误差在2m之内;船舶摇摆角度在(-5°,5°)时,测距相对误差小于2.7%,验证了该方法的有效性。The distance measurement function is an important way to ensure ships safe navigation on the water and avoid collision with the surrounding obstacles and other ships to a great extent,however ship swaying has always been a difficult problem in ship ranging.To solve this problem,the traditional method of monocular distance measurement is improved,and a real-time ship ranging method and error correction method are proposed.By using the faster R-CNN algorithm to draw a rectangular frame,ECO algorithm can track the target and align the optical center of the photoelectric cloud platform to the bottom center of the target.The gyroscope is used to collect the dynamic angle of the ship's swinging,and the trigonometric geometry principle is used to carry out real-time ranging.Finally,the ranging result is corrected according to the error source.The experimental results show that the static ranging error is within 2 meters in 100 meters,and the relative ranging error is less than 2.7%when the ship swing angle is within(-5°,5°),which verifies the effectiveness of the proposed method.

关 键 词：机器视觉 单目测距 船舶摇摆 误差修正 

分 类 号：TP242[自动化与计算机技术—检测技术与自动化装置] U675[自动化与计算机技术—控制科学与工程]