摘要
针对目前基于RGB模型的干枣彩色图像研究中的一系列问题,提取干枣图像大小与实际大小相差较大,算法复杂、速度慢,很难达到在线检测的要求,提出一种利用机器视觉技术以提取干枣图像目标的新方法,分别统计R、G、B各分量的灰度值,绘制相应直方图;通过分析比较,选取R-B分量灰度直方分布图作为确定阈值分割的依据,利用大津法自适应求出阈值,然后提出区域阈值分割方法,并对干枣图像进行阈值分割。多次实验证明,利用该算法提取干枣大小与实际大小非常接近,稳定性高,检测速度快,可以满足进行干枣的形状、表面缺陷等检测要求,能够满足机器视觉技术在线检测要求。
Recently in the study of dry jujube＇s colorful images,which is based on the RGB model,there exist some problems： the big differences between the size of dry jujube＇s image and the real size of it,the complicated calculating methods and the low speed of calculating,and all of these problems make the study hard to approach the requirements of the on-line inspection.To solve them,a new method of extracting dry jujube＇s image with machine vision technology has been put forward,to account the gray value of R,G and B separately,and draw the relative histogram;through the analysis and comparing,choose the gray value histogram of R-B to confirm the threshold segmentation,to get thresholds by Otsu method,and then make threshold segmentation by the Regional threshold segmentation algorithm.Series experiments have proved that using this new algorithm,the size of extracted dry jujube and the real size are very close,more stable and faster,what＇s more,it can meet the inspection requirements of dry jujubes＇ shape and surface imperfection,and achieve the online inspection requirements of machine vision technology.
出处
《现代仪器》
2012年第6期44-47,共4页
Modern Instruments
基金
科技部国家中小型企业创新基金(项目编号:12C26213202510)
江苏省科技支撑计划(项目编号:BE2008353)
科技型中小企业技术创新基金(项目编号:07C26213200355)
关键词
干枣
直方图
阈值分割
机器视觉
Dry jujube Histogram Threshold segmentation Machine vision
