摘要
为了快速定位玉米植株位置,以苗期4~6叶玉米为研究对象,提出了一种苗期玉米冠层识别与质心定位方法。首先,在大田环境下获取农田作物视频数据,基于FasterR-CNN对玉米冠层进行识别;其次,用差分内积线性特性改进质心检测算法,对玉米冠层和杂草进行分割,并对玉米冠层识别区域进行质心定位计算,得到玉米苗质心的像素坐标;最后,通过农田实验对本文冠层识别与质心定位方法进行验证。结果表明,苗期玉米冠层识别方法的平均识别率达92.9%,检测一帧图像的平均时间为0.17s,玉米冠层质心定位误差不超过1像素。
Identification and location of maize canopy at seedling stage is an important basis for plant operation and fertilization.Aiming at the problem that fertilizer utilization rate is not high,strip and broadcast fertilizer are the main fertilization methods in China,a new method for identification and location of maize canopy in seedling stage was proposed,which could identify and locate the canopy centroid of maize in seedling stage under farmland environment.In order to locate the maize plant position quickly,the maize canopy was identified by the method of deep learning,and then the centroid location of the identified area was calculated.Firstly,Faster R-CNN was used to identify the seedling maize canopy in the field environment.Then the centroid detection algorithm based on the linear property of differential inner product improved the centroid location method of maize canopy.After the corn canopy and the weeds were segmented,the centroid of maize canopy was located and the pixel coordinates of maize seedling centroid were obtained.Through the verification in the actual farmland environment,the accuracy of identification and positioning of the canopy centroid reached 92.9%.The average detection time of a frame image was 0.17 s.The positioning error of the canopy centroid was less than 1 pixel.Through the analysis of the test results,the research results can provide information support for the positioning fertilization and research basis for the follow-up variable fertilization operations,so as to achieve the production goal of saving fertilizer and increasing efficiency.
作者
宗泽
赵硕
刘刚
ZONG Ze;ZHAO Shuo;LIU Gang(Key Laboratory of Modern Precision Agriculture System Integration Research,Ministry of Education,China Agricultural University,Beijing 100083,China;Key Laboratory of Agricultural Information Acquisition Technology,Ministry of Agriculture and Rural Affairs,China Agricultural University,Beijing 100083,China)
出处
《农业机械学报》
EI
CAS
CSCD
北大核心
2019年第B07期27-33,共7页
Transactions of the Chinese Society for Agricultural Machinery
基金
国家重点研发计划项目(2016YFD0200600-2016YFD0200605)
关键词
玉米冠层
苗期
识别
质心定位
maize canopy
seedling stage
identification
centroid positioning
