基于无人机遥感影像的剑麻株数识别  

Monitoring Sisal Plant Number Based on UAV Digital Images

作　　者：付虹雨 崔国贤[1] 崔丹丹 佘玮[1] 李绪孟[2] 苏小惠 王继龙 刘婕仪 王昕慧 刘皖慧 赵亮 全芮萍 周倩文 FU Hongyu;CUI Guoxian;CUI Dandan;SHE Wei;LI Xumeng;SU Xiaohui;WANG Jilong;LIU Jieyi;WANG Xinhui;LIU Wanhui;ZHAO Liang;QUAN Ruiping;ZHOU Qianwen(Ramie Research Institute of Hunan Agricultural University,Changsha,Hunan 410128,China;College of Agriculture,Hunan Agricultural University,Changsha,Hunan 410128,China)

机构地区：[1]湖南农业大学苎麻研究所,湖南长沙410128 [2]湖南农业大学农学院,湖南长沙410128

出　　处：《中国麻业科学》2020年第6期249-256,共8页Plant Fiber Sciences in China

基　　金：国家重点研发计划(2018YFD0201106);国家麻类产业技术体系(CARS-16-E11);国家自然科学基金(31471543);湖南省重点研发计划项目(2017NK2382)。

摘　　要：准确、快速、无损地获取剑麻株数信息对于剑麻产量估测至关重要。研究以广西壮族自治区钦州市东方农场为试验区,探讨面向对象多尺度分割算法提取剑麻株数信息的可行性。首先基于无人机遥感系统获取的剑麻冠层可见光谱影像,使用Pix4Dmapper拼接重建试验区的全幅正射影像;然后利用面向对象多尺度分割算法分割剑麻中心区域和其他区域,通过采样定义类别特征空间实现株数识别;将基于算法提取的株数与目视解译统计的株数进行对比分析,评估识别精度;最后,基于面对对象分类结果,利用主成分分析法提取表征剑麻主要性状信息的主成分,为剑麻冠层图像特征研究提供参考。结果表明:采用面向对象多尺度分割算法提取剑麻株数的精度在87.1%左右,该方法用于作物株数估测是可行的。It is important to acquire sisal plant number accurately,rapidly and nondestructively for the estimation of sisals’yield.Taking Dongfang Farm of Guangxi province as the research area,the feasibility of extracting sisal plant number information by object-oriented multi-scale segmentation algorithm was discussed.Firstly,based on the UAV remote sensing system,the sisal canopy visible spectrum images were obtained,and Pix4 Dmapper was used to reconstruct the full-amplitude orthophoto image of the test area.Then the multiscale segmentation algorithm was used to segment the central region of sisal and other regions,and the plant number recognition was realized by object oriented classification.The number of plants extracted based on algorithm was compared with the number of plants based on visual interpretation statistics to evaluate the recognition accuracy.Finally,the image features of the central region of sisal were extracted based on the classification results,and the principal components were extracted to represent the principal components of the main character information of sisal,so as to provide reference for the classification research of sisal.The results showed that the accuracy of sisal plant number extracted by object-oriented multi-scale segmentation algorithm was about 87.1%,which indicated that it was feasible for plant number estimation.

关 键 词：无人机 剑麻 株数 主成分 

分 类 号：S563.8[农业科学—作物学]