沈阳农业大学学报2017，Vol.48Issue(5)：629-635,7.DOI:10.3969/j.issn.1000-1700.2017.05.017
基于无人机高清数码影像的水稻产量估算
Rice Yield Estimation Based on High-definition Digital Image of UAV
摘要
Abstract
Now the estimation of rice yield is mainly based on satellite remote sensing, however, the resolution of satellite remote sensing is low, resulting in lacks of the mechanism and error. In order to obtain the canopy information of rice and improve the resolution and accurately estimate rice yield, the rice canopy image can be taken from the heading stage to the mature stage by using the UAV platform equipped with high-definition digital camera. Firstly, the median filter algorithm was used to denoise the rice canopy image in RGB color space, then according to the color feature of the rice image, the image was converted from RGB color space to L*a*b* color space. K-means clustering algorithm was used to cluster analysis and image segmentation of the rice canopy image. We extracted the rice panicle and put the number of rice panicle into the rice yield estimation formula. There were 18 rice communities (8m long and 5m wide) in the experimental plots, and the images were taken four times between the heading stage and the mature stage.The data of the test records include the time, height and resolution of the shoots. At the same time, the number of rice panicles and yield were measured in the field,which provided the basis for the later evaluation and judgment of K-means clustering algorithm to extract the accuracy of rice panicles and yield estimation.The measured and estimated values of rice yield,panicles in field and extracted in the image were analyzed.The results showed that according to the rice canopy images taken by UAV on August 18th,the effect of rice panicle extraction was good, with a high estimated resolution. The root mean square error and mean absolute percentage error of the yield estimation were 9.08 and 22.8%, respectively; the root mean square error and mean absolute percentage error of the panicle numbers were 19.86 and 5.8%, respectively. It showed that UAV equipped with digital camera can quickly and without loss access to rice canopy information, and that the K-means clustering algorithm can accurately segment rice panicles from the canopy image. It is feasibleto estimate rice yield by digital image.关键词
无人机/颜色空间/K均值聚类/图像分割/水稻穗Key words
UAV/color space/K-means clustering/image segmentation/rice panicle分类
农业科技引用本文复制引用
李昂,王洋,曹英丽,于丰华,许童羽,肖文..基于无人机高清数码影像的水稻产量估算[J].沈阳农业大学学报,2017,48(5):629-635,7.基金项目
国家重点研发项目(2016YFD0200700,2017YFD0300706) （2016YFD0200700,2017YFD0300706）
辽宁省教育厅课题重点项目(LSNZD201605) （LSNZD201605）
