南京理工大学学报（自然科学版）2023，Vol.47Issue(6)：782-789,8.DOI:10.14177/j.cnki.32-1397n.2023.47.06.007
基于声呐图像的鱼群识别与计数方法
Fish recognition and counting method based on sonar images
摘要
Abstract
In order to obtain the number of fishes accurately,a fish recognition and counting method based on an sonar images is proposed.The image sonar is fixed underwater to collect information,the sonar data is preprocessed with image construction,coloring and noise removal.The you-only-look-once v5(YOLOv5)neural network model is used to recognize and label fish automatically,a multi-target tracking algorithm is used based on recognition boxes to track and count fish.Then a historical recognition box sequence is used to solve the pairing problem of reappearing recognition boxes.This method is used for experimental detection in lakes,the results are compared with manual counting.The results show that the fish recognition and counting method based on an sonar images has high accuracy,with a counting error within 15%.关键词
声呐图像/目标识别/鱼群计数/图像构建/上色/噪声去除/多目标跟踪/历史识别框序列Key words
sonar images/target recognition/fish counting/image construction/coloring/noise removal/multi-target tracking/historical recognition box sequence分类
信息技术与安全科学引用本文复制引用
朱俊,封磊..基于声呐图像的鱼群识别与计数方法[J].南京理工大学学报（自然科学版）,2023,47(6):782-789,8.基金项目
江苏高校"青蓝工程" （）
金陵科技学院学术拔尖培养工程(2020) （2020）
