无人机遥感技术在多房棘球绦虫中间宿主监测中的应用  

Application of UAV remote sensing technology for intermediate host surveillance of echinococcus multilocularis

作　　者：张光葭 王奇[1] 王谦[1] 喻文杰[1] 何伟[1] ZHANG Guang-jia;WANG Qi;WANG Qian;YU Wen-jie;HE Wei(Sichuan Center for Disease Control and Prevention,Chengdu 610041,Sichuan Province,China)

机构地区：[1]四川省疾病预防控制中心,成都610041

出　　处：《预防医学情报杂志》2023年第1期1-5,12,共6页Journal of Preventive Medicine Information

基　　金：国家卫生健康委包虫病防治研究重点实验室项目(项目编号:2021WZK1005);四川省科技厅重点研发计划(项目编号:2023YFS0221)。

摘　　要：目的采用无人机遥感技术对多房棘球绦虫中间宿主分布和密度进行监测,探究多房棘球绦虫中间宿主洞穴航空影像解译方法,为多房棘球绦虫中间宿主监测提供新方法。方法采用大疆创新公司(DJI)的M200(Matrice 200)飞行平台搭载禅思Z30(Zenmuse Z30)云台相机获取航空影像,选取4个简单场景区域影像以及4个复杂场景区域影像,利用深度学习的方法,对8个影像中所存在的多房棘球绦虫中间宿主洞穴进行检测,并评估此方法精度。结果4个简单场景区域的召回率、精确率分别为90.37%、84.60%,4个复杂场景区域的召回率、精确率分别为80.96%,77.40%,4个简单场景区域和4个复杂场景区域通过深度学习方法识别出多房棘球绦虫中间宿主的召回率之间差异有统计学意义(χ^(2)=1536.684,P<0.001),且精确率之间差异有统计学意义(χ^(2)=793.443,P<0.001)。结论简单场景区域的召回率和精确率均高于复杂场景区域;在简单场景区域,多房棘球绦虫中间宿主洞穴的自动检测效果较佳;在复杂场景区域,多房棘球绦虫中间宿主洞穴的自动检测效果较差。无人机遥感技术在多房棘球绦虫中间宿主监测中有较好的适用性和应用潜力,为今后采用遥感技术开展监测工作提供了参考。Objective To monitor the distribution and density of Echinococcus multilocularis intermediate hosts by using UAV remote sensing technology,and explore the interpretation method of aerial image of Echinococcus multilocularis intermediate host caves to provide a new method for Echinococcus multilocularis intermediate host monitoring.Methods DJI′s M200(Matrice 200)flight platform with Zenmuse Z30 gimbal camera was used to acquire aerial images.Four images of simple scene areas and four images of complex scene areas were selected to detect the presence of intermediate host burrows of Echinococcus multilocularis by using a deep learning method.We also evaluated the accuracy of this method.Results Recalls and accuracy rates for the four simple scene regions were 90.37%and 84.60%,respectively.Recalls and accuracy rates for the four complex scene regions were 80.96%and 77.40%,respectively.The differences in recall rates between the four simple scene regions and the four complex scene regions identified by the deep learning method for the intermediate hosts of Echinococcus multilocularis were statistically significant(χ^(2)=1536.684,P<0.001),and the differences in accuracy rates were also statistically significant(χ^(2)=793.443,P<0.001).Conclusions Recall and accuracy rates were higher for simple scenario regions than for complex scenario regions.Automatic detection of intermediate host burrows of Echinococcus multilocularis performs better in simple scene areas,while this method performs poor in complex scene areas.There is good application potential of UAV remote sensing technology in the intermediate host monitoring of Echinococcus multifasciatus,which provided a reference for future monitoring work using remote sensing technology.

关 键 词：无人机 遥感技术 多房棘球绦虫 监测 

分 类 号：R532.32[医药卫生—内科学]