摘要
数字图像在作物信息采集方面具有信息量大、速度快、精度高等显著的特点和优势,被广泛应用于作物生长诊断方面。以往研究集中在作物图像分析技术,对图像获取手段和获取系统的技术要求等论述较少。从建立作物图像获取系统的角度出发,在分析作物图像颜色、分辨率、格式和文件存储空间大小的基础上,总结了常用作物图像获取及分析技术在作物监测方面的应用。结果表明,作物水分及营养状况诊断多采用图像颜色特征分析的方法,常用RGB和HIS颜色模型,作物图像可以在自然光照条件下获取,镜头距离地面或作物的距离、拍摄角度、图像分辨率和图像存储格式并无定规,可用JPEG格式来存储图像,节省大量的图像存储空间。土壤和杂草等背景的识别、病虫害特征提取等方面采用颜色特征、多光谱图像、纹理特征和形状特征等方法,图像颜色分辨率从288×352像素到3 072×2 304像素可使识别目标达到80%～95%。图像分辨率高的识别效果好,但是程序用时长,图像文件占用空间大,影响图像传输速度和诊断的实时性。因此获取图像时要根据图像应用的具体情况来选择合理的图像分辨率。
Image acquisition is a significant step for the research of monitoring crop conditions using computer vision technology.Many research works have concentrated on image process methods for diagnosing crop conditions.However,few researches have been conducted on the technology of crop image acquisition.For this purposes,this paper analyzes the color,resolution,format and size of crop image file firstly.Then,the methods on crop image process and crop conditions diagnoses have been summarized.It is concluded that the color characters analysis models of RGB and HIS usually are applied for crop water stress and nutrients condition diagnose.The crop image used for this diagnose could be obtained under filed condition.The format and resolution of the crop image have no significant influence on diagnose.Both the image-taking distance away from the plant and picture taking angle to the plant has no significant influence on diagnose.Therefore,the JPEG format of crop image is recommended in order to save computer storage space.Methods for distinguish of crop,weeds,plant diseases and insect pests includes color analysis,multiple spectroscopy image process,image venation and shape features analysis.The pixel of image from 288×352 to 3072×2304 will result with the distinguish precision from 80% to 95%.The more resolution of the crop image,the higher distinguish precision,the long running time and the more storage space needed for the computer.This will limit the image transmitting speed and real time image processing.Therefore,the selection of reasonable image resolution should consider the application purpose of image processing.
出处
《农机化研究》
北大核心
2012年第6期1-6,共6页
Journal of Agricultural Mechanization Research
基金
国家“十二五”科技支撑计划项目(2011BAD29B08)
高等学校创新引智项目(111-2-16)
关键词
计算机视觉
图像获取
图像分辨率
图像特征信息
作物生长诊断
computer vision
image obtaining
image resolution
image feature information
crop condition diagnose
