基于深度学习的海面垃圾检测系统
    点此下载全文
引用本文:杜叶挺,应泽光,马赛男,卓宏明,顾欣.基于深度学习的海面垃圾检测系统[J].计算技术与自动化,2023,(3):67-71
摘要点击次数: 219
全文下载次数: 0
作者单位
杜叶挺,应泽光,马赛男,卓宏明,顾欣 (浙江国际海运职业技术学院 海洋装备工程学院浙江 舟山 316021) 
中文摘要:舟山作为我国重要的海洋养殖城市与海洋旅游城市,在经济快速发展的同时如何解决日益严峻的海洋垃圾污染问题愈发突显。相比传统的雷达回波方式,基于深度学习的机器视觉检测方法,具备抗海浪干扰能力强,检测速度快,识别信息丰富等优点。采用DeepLabv3+图像语义分割模型,通过YOLOv5s目标检测算法对海面垃圾目标进行实时检测,最后由摄像机单目测距获取目标距离,可以实现检测效率25 fps,检测准确率87%,具备较好的工程应用价值。
中文关键词:深度学习  语义分割  机器视觉  目标检测
 
Detection System of Sea Garbage Based on Deep Learning
Abstract:Zhoushan as an important marine aquaculture city and marine tourism city in China, how to solve the increasingly serious sea garbage pollution problem was becoming more and more prominent while the economy was developing rapidly. Compared with the traditional radar detection methods, the machine-vision detection based on deep learning had more advantages, such as strong anti-wave interference ability, fast detection speed and rich recognition information. This system used DeepLabv3+ semantic segmentation model and YOLOv5s algorithm to realize the detection of the garbage target. At last, it can locate the garbage’s position by monocular ranging algorithm. It can achieve the detection efficiency of 25fps and the detection accuracy of 87%, and has certain engineering application value.
keywords:deep learning  semantic segmentation  machine vision  target detection
查看全文   查看/发表评论   下载pdf阅读器