摘要
在基于机器视觉技术的枸杞外部缺陷识别过程中,样品数量过多且摆放密集,导致边界阴影与外部缺陷相互混淆,从而影响了识别的准确性。针对此问题,提出一种基于形态学边缘检测的枸杞外部缺陷识别方法。首先,通过二值化进行背景分割,提取存在边界阴影的果实图像。然后,以改进的自适应形态学梯度算子提取精确果实边界,进而对其填充并与样本图像做差运算,去除边界阴影。最后,对图像去噪并增强对比度,通过二值化筛选缺陷部位完成识别。实验结果表明,提出的方法正确率达到90%以上。该方法满足枸杞外部缺陷识别的实际需求,为进一步进行枸杞质量评价提供了有利条件。
In the recognition of the wolfberry appearance defects based on the technology of machine vision,samples are excessive and intensive,which causes a mutual confusion between border shadow and defect region and accordingly affects the recognition accuracy.To solve this problem,this paper puts forward a recognition method for wolfberry appearance defects by edge detection of morphology.Firstly,the binarization is applied to segment the background and extract the fruit region with boundary shadow.Secondly,a modified adaptive morphological gradient operator is used to obtain the precise boundaries of fruit.Thirdly,different operation is performed between the filled image and the sample image to extract precise fruit image for removing the boundary shadow.Finally,the resultant image is denoised and enhanced to improve its contrast.And the binarization is used to locate the appearance defects region for its recognition.The experiment results indicated that the correct rate is above 90%.The proposed method meets the practical demand of the recognition of wolfberry appearance defects and offers favorable conditions for its further quality evaluation.
出处
《湖北工程学院学报》
2015年第6期32-37,共6页
Journal of Hubei Engineering University
基金
国家自然基金项目(61363054)
关键词
枸杞外部缺陷
边界阴影
缺陷区域识别
形态学边缘检测
wolfberry appearance defects
boundary shadow
recognition of defects region
morphology edge detection
