运动船舶参数视频检测算法  

A parameter detection algorithm of moving ship based on video sequence

作　　者：王江[1] 廖娟[1] 陈星明[1] 李勃[1] 陈启美[1] 

机构地区：[1]南京大学电子科学与工程学院,南京210023

出　　处：《南京大学学报（自然科学版）》2015年第2期227-233,共7页Journal of Nanjing University（Natural Science）

基　　金：国家自然科学基金(61401239;61105015);国家船联网专项科技项目:船舶实时视频图像监测识别系统(2012-364-641-209)

摘　　要：水路运输因其价廉,承载量大等,在我国综合运输体系中占有重要地位.然而由于船舶构造尺寸、航行操作不规范,缺乏有效监控手段,撞桥搁浅、两船相碰等现象时有发生.而借鉴公路交通的视频监控技术,存在种种困难,诸如,航道没有标线,摄像机难以标定;船尾波纹与船行同步,难以区分;加之水面波光粼粼,不似黑色柏油道路对比度高等等.为此,提出了运动船舶参数视频检测的系列算法:提出网格交互式摄像机标定算法,以获取图像坐标系与世界坐标系的映射;优化Vibe视觉背景建模算法,去除船尾波纹干扰,从而提取船舶轮廓;建立船舶长宽、速度、流量的检测模型,以分别获取船舶尺寸与运行参数.该系统现已鉴定,在京杭运河运行示范.经检测:长宽、速度和流量的检测准确率分别优于90%、95%和98%.Because of its low cost,carrying capacity,etc.,waterway transport plays an important role in our comprehensive transmission system.However,due to the different size of ship construction,nonstandard navigation operations and the lack of effective control measures,accidents such as ships colliding,hitting the bridge or going aground often occur.When referring to technology from road traffic video surveillance,difficulties occur,such as the camera is difficult to calibrate without markings on water,it＇s hard to distinguish ships from their stern ripple which move together,sparkling water doesn＇t have high contrast like the black asphalt road and so on.So this paper proposes a series of ship parameters detection algorithm based on video：proposed grid interactive camera calibration algorithm to obtain the mapping image between coordinate system and the world coordinate system;optimization Vibe algorithm to remove the stern ripple interference,thereby extracting the ship contour;establish detection model of ship length and width,speed,traffic,in order to get the ship size and operating parameters,respectively.The system has been identified in the Beijing-Hangzhou Grand Canal.It is tested with width,speed,and flow detection accuracy,respectively,better than 90%,95% and 98%.

关 键 词：视频分析 交互式标定 视觉背景建模算法 船舶参数检测 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]