摘要
为解决凭借人工经验方式对小桐子种子品质进行筛选效率较低、主观性强、错误率高且实时性差等问题,采用数字图像处理技术对小桐子种子图像进行分析,针对其特征对图像进行了R,G,B彩色分量的算术运算融合,采用形态学开运算进行消噪处理,并根据OTSU方法进行自适应的阈值选取将图像处理成二值图像,利用LOG算子实现了图像的边缘提取,且通过计算得到了小桐子种子的部分形态特征参数值.结果表明,与人工分割的方法相对比,90幅具有不同摆放方式的小桐子种子图像的平均分割误差不超过0.63%,最大分割误差为1.07%,均方误差σ不超过0.006 4,能较为准确地实现小桐子种子图像的背景分割和参数检测.
In order to solve the problems of low efficiency, strong subjectivity, high error rates and poor real time that when selecting the quality of Jatropha seeds by experience, the images of Jatropha seeds were analyzed by digital image processing technology in this paper. Aiming at the characteristics of Jatropha seeds images, the R,G,B color components by arithmetic were firsty fused, then the arithmetic results were processed for noise cancellation by morphological opening operation, the adaptive threshold selection method was used for these images which were converted into binary images according to the method of OTSU and the edge extraction of images was realized by way of LOG operator. The portion of Jatropha seeds morphology parameters with the procedure of image processing finally was also calculated and got. Compared with the artificial segmentation method, the experimenal results show that although the background segmentation of Jatropha seeds images produces certain errors because the Jatropha seeds which are irradiated by light generate unequal shaded area in various display modes, overall average error of segmentation is less than 0.63%, the maximum error is 1.07% and the mean square error σ is not over 0.006 4 in 90 images of Jatropha seeds. Above all, this algorithm can achieve the accurate and reliable background segmentation and parameter detection of Jatropha seeds images.
作者
王焕昌
杨启良
杨莹
WANG Huanchang;YANG Qiliang;YANG Ying(College of Modern Agricultural Engineering,Kunming University of Science and Technology,Kunming,Yunnan 650500,Chin)
出处
《排灌机械工程学报》
EI
CSCD
北大核心
2018年第8期713-718,共6页
Journal of Drainage and Irrigation Machinery Engineering
基金
国家自然科学基金资助项目(51379004
51009073
51109012)
关键词
小桐子种子
背景分割
参数检测
摆放方式
形态特征
Jatropha seeds
background segmentation
parameter detection
display mode
morphological characteristic
