红外相机监测目标物种的一种自动化检测方法:以绿尾虹雉为例  

A Method for Automatic Detection of Target Species Using Infrared Cameras:A Case Study of Lophophorus lhuysii

作　　者：王翰霖 文帅 白俊 李东睿 罗概 林玉成[1,2] WANG Hanlin;WEN Shuai;BAI Jun;LI Dongrui;LUO Gai;LIN Yucheng(Key Laboratory of Bio-Resources and Eco-Environment(Ministry of Education),College of Life Sciences,Sichuan University,Chengdu 610065,China;Sichuan Key Laboratory of Conservation Biology on Endangered Wildlife,College of Life Sciences,Sichuan University,Chengdu 610065,China;College of Computer Technology and Science,Southwest University of Science and Technology,Mianyang,Sichuan Province 621010,China)

机构地区：[1]四川大学生命科学学院,生物资源与生态环境教育部重点实验室,成都610065 [2]四川大学生命科学学院,四川省濒危野生动物保护重点实验室,成都610065 [3]西南科技大学计算机科学与技术学院,四川绵阳621010

出　　处：《四川动物》2022年第4期361-369,共9页Sichuan Journal of Zoology

摘　　要：红外相机是目前在野生动物资源调查和监测中的一种重要手段和工具,但对其采集的海量影像数据的甄别和所拍获物种的鉴定工作费时费力。为解决红外相机影像数据量庞大、无法自动识别目标物种、人为检索繁琐、以及卷积神经网络方法的检测效率和鉴别正确率低等问题,本文对红外相机采集的11万余张图像进行筛选,以绿尾虹雉Lophophorus lhuysii为例,运用协同注意力机制,提出一种针对红外相机影像数据中目标物种的自动化检测方法。实验结果表明,该方法对绿尾虹雉图像与视频的识别准确度达到99.62%。本文提出的方法能够提高对检测目标物种的识别率,降低人力成本,有利于指导野生动物的监测和保护。Infrared camera is an important method in the investigation and monitoring of wildlife resources,but the processes of massive image data and species identification are time-consuming and laborious. Infrared camera is unable to automatically identify target species along with large amount of image data,cumbersome manual retrieval,and low detection efficiency and identification accuracy of convolution neural network. To solve these problems,this study proposed an object detection method embedding attention mechanism by using Chinese monal(Lophophorus lhuysii)as model. A total of 110 000 images were collected by infrared cameras,and 715 images were selected for further analysis. The results showed that the accuracy of the proposed network could achieve 99. 62%. The improved method can improve the efficiency of species identification,reduce labor costs,and promote the protection of target species.

关 键 词：红外相机 绿尾虹雉 注意力机制 目标检测 

分 类 号：S862[农业科学—野生动物驯养]