基于机器视觉的船舶舱室液体泄漏监测方法  

Monitoring method for liquid leakage in ship cabins based on machine vision

作　　者：余永华[1] 张佳明[1] 胡磊[1] YU Yonghua;ZHANG Jiaming;HU Lei

机构地区：[1]武汉理工大学船海与能源动力工程学院,湖北武汉430063

出　　处：《中国修船》2022年第4期32-36,共5页China Shiprepair

基　　金：智能中速柴油机关键技术研究(工信部装函[2019]360号)。

摘　　要：基于机器视觉的舱室液体泄漏监测,是实现船舶机舱朝无人化方向发展的必然需求。文章针对目前低照度下船舶舱室液体泄漏的图像质量差和识别精度低等问题,提出了一种基于机器视觉的船舶舱室液体泄漏监测方法。首先,采用改进的多尺度Retinex算法对低照度下船舶舱室液体泄漏的图像进行处理,使图像对比度增强、细节更丰富、边缘保留完整。然后,对增强后的液体泄漏图像进行特征提取,通过机器学习对船舶舱室液体的泄漏故障进行判断。结果表明,增强后的泄漏图像识别精度得到了明显提升,使用k-近邻(k-Nearest Neighbor,k NN)算法可达到95%的分类精度,能够很好地监测低照度下的船舶舱室液体泄漏。Liquid leakage monitoring of cabins based on machine vision constitutes an inevitable demand for the development of an unmanned ship engine room.This paper proposes a monitoring method for liquid leakage in ship cabins based on machine vision to solve the problems of poor quality and low recognition accuracy of liquid leakage images under low illumination.Specifically,an improved multiscale Retinex algorithm is used to process images of liquid leakage in ship cabins under low illumination for enhanced contrast,richer details,and better integrity of the edge.Then,the features of the enhanced liquid leakage images are extracted,and the liquid leakage fault of the ship cabin is identified by machine learning.The results show that the recognition accuracy of the enhanced leakage images is significantly improved and that classification accuracy can reach 95%after the k-Nearest Neighbor(k NN)algorithm is employed.The proposed method can well monitor the liquid leakage in ship cabins under low illumination.

关 键 词：图像增强 船舶舱室 多尺度RETINEX算法 机器视觉 液体泄漏 

分 类 号：U672[交通运输工程—船舶及航道工程]