| 基于梯度图像的玉米种胚褶皱识别 |
| |
| 引用本文： | 程洪,史智兴,周桂红,尹辉娟,李亚南.基于梯度图像的玉米种胚褶皱识别[J].中国农业大学学报,2014,19(2):196-200. |
| |
| 作者姓名： | 程洪 史智兴 周桂红 尹辉娟 李亚南 |
| |
| 作者单位： | 河北农业大学 信息科学与技术学院, 河北 保定 071001;中国农业大学 信息与电气工程学院, 北京 100083;河北农业大学 信息科学与技术学院, 河北 保定 071001;河北农业大学 信息科学与技术学院, 河北 保定 071001;河北农业大学 信息科学与技术学院, 河北 保定 071001;河北农业大学 信息科学与技术学院, 河北 保定 071001 |
| |
| 基金项目： | 科技部“十二五”农村领域国家科技计划资助项目 （2011BAD16B08-3）；河北农业大学理工基金项目（LG20110601）；保定市科学研究与发展计划项目（13ZN010） |
| |
| 摘 要： | 针对玉米品种自动识别中玉米种胚特征的提取问题,提出一种根据玉米种胚图像梯度图,采用基于密度的K-均值聚类算法识别分析玉米种胚褶皱的方法。根据阈值识别出胚部区域,计算其长轴方向的梯度;利用Sobel算子与膨胀算子,去除梯度图像边缘噪声,利用中值滤波去除孤立点噪声;以像素点的R、G、B梯度为特征,组成特征向量空间;采用基于密度的初始中心点优化算法,启发式的找到初始聚类中心;K-均值聚类分析,得到褶皱区域,标记区域并统计褶皱个数。对300颗玉米籽粒进行胚部褶皱识别分析,结果表明:无论是深褶皱、还是浅褶皱的,该方法均能有效识别,与人工观测值的平均吻合率达82.7%。通过本方法获取的褶皱信息体现了玉米种胚纹理特性,为玉米籽粒品种自动识别中特征参数的选取提供了新的思路和方法。
|
| 关 键 词： | 玉米种子 胚部褶皱 图像梯度图 K-means 自动识别 |
| 收稿时间： | 2013/7/5 0:00:00 |
Corn kernel wrinkle recognition algorithm based on gradient image |
| |
| CHENG Hong,SHI Zhi-xing,ZHOU Gui-hong,YIN Hui-juan and LI Ya-nan.Corn kernel wrinkle recognition algorithm based on gradient image[J].Journal of China Agricultural University,2014,19(2):196-200. |
| |
| Authors: | CHENG Hong SHI Zhi-xing ZHOU Gui-hong YIN Hui-juan and LI Ya-nan |
| |
| Institution: | College of Information Science and Technology, Agricultural University of Hebei, Baoding 071001, China;College of Information and Electrical Engineering, China Agricultural University, Beijing 100083, China;College of Information Science and Technology, Agricultural University of Hebei, Baoding 071001, China;College of Information Science and Technology, Agricultural University of Hebei, Baoding 071001, China;College of Information Science and Technology, Agricultural University of Hebei, Baoding 071001, China;College of Information Science and Technology, Agricultural University of Hebei, Baoding 071001, China |
| |
| Abstract: | The distribution of the wrinkles of corn kernel embryo is closely related with the corn variety.It is very important to identify the embryo wrinkle automatically and accurately for promoting the automatic recognition corn varieties in accuracy.This paper presented a method to identify wrinkle distribution.This method was using Density-based K-means clustering algorithm based on the gradients of corn kernel embryo area.First of all,the region of the embryo was identified according to the threshold value,and then the gradient of the major axis direction of embryo region was calculated.Using sobel operator and dilation operator,the edge noises of the gradient image and the isolated point noises were removed by median filter.A feature vector space with R gradient,G gradient,B gradient of pixel as characteristics was formed.Optimization algorithm based on the density was used to optimize initial centers of K-means.The wrinkle regions,marked area and computed the number of wrinkles were found by K-means clustering analysis.300 corn kernels were selected for experiment and the results showed that this method was valid for the deep wrinkles and light wrinkles of the embryo.The average recognition rate was 82.7%.In a word,there were three innovations in this paper:Firstly,this paper presented the idea that it analyzed embryo winkle distribution by analyzing gradients of the maize kernels embryo R,G,B images.Secondly,it gave the basic definitions of the pixel density used in this study.Finally,this method of analysis of the embryo wrinkles not only was used to count the number of corn kernel embryo wrinkles and confirm general distribution location,but also could be used to analyze the depth of the wrinkles by the gradients of wrinkle if the research would need. |
| |
| Keywords: | corn kernel embryo wrinkle gradient image K-means automatic identification |
| 本文献已被 CNKI 等数据库收录！ |
| 点击此处可从《中国农业大学学报》浏览原始摘要信息 |
| 点击此处可从《中国农业大学学报》下载免费的PDF全文 |
|
