基于视觉数据驱动的棉花异性纤维检测
2017-10-15师红宇管声启
师红宇 管声启
摘要: 为了提高棉花异性纤维检测准确性及快速性,提出一种基于视觉数据驱动的棉花异性纤维检测新方法。首先,对采集棉花异性纤维图像进行小波金字塔多层分解,实现棉花异性纤维图像中各信息的分离;然后使用meanshift算法平滑图像,消除光照不均的干扰;在此基础上,采用中央-周边操作算子和融合操作构建显著图,从而提高异性纤维的显著度;最后,利用改进的区域生长法获得目标区域。文章采用视觉数据驱动构建的检测模型,结构简单、检测速度快、准确率高。实验结果表明,该检测方法能够有效实现棉花异性纤维的检测,提高检测准确性和快速性,为棉花异性纤维自动检测提供一种新方法。
关键词: 异性纤维;数据驱动;显著图;meanshift算法;改进的区域生长法
中图分类号: TS101.91;TP391.41
文献标志码: A
文章编号: 10017003(2017)05003607
引用页码: 051107
Abstract: In order to improve the accuracy and rapidity of cotton foreign fiber detection, a new method based on visual data driven to detect cotton foreign fibers is proposed. Firstly, cotton foreign fiber image was decomposed into multiple layers by wavelet Pyramid, to achieve the separation of all kinds of information in the cotton foreign fiber image. Secondly, meanshift method was applied to smoothen the image so as to eliminate the interference of uneven light. On this basis, saliency map was constructed by centersurround operator and fusion operator, to improve saliency degree of cotton foreign fibers. Finally, the target area was obtained by using the improved region growing method. This detection model has the advantages of simple structure, fast detection speed and high accuracy. The experimental results show that the detection method can effectively detect the foreign fiber, and improve the detection accuracy and rapidity. This paper provides a new method for automatic detection of cotton foreign fibers.
Key words: foreign fibers; data driven; saliency map; meanshift algorithm; improved region growing method
棉花是紡织企业加工处理的原材料,在原棉加工处理过程中由于很多原因混入了多种异性纤维(棉花中的异性纤维是指混入棉花中非棉纤维和非本色纤维,如棉籽、棉壳、棉叶、毛发、塑料绳、染色线等),这些混入棉花中的异性纤维,在纺织加工中被碎成单纤维,对棉纱和布面的质量造成了很大危害,难以清除[12]。因此,棉花异性纤维的检测和剔除是纺织企业的一个重要必备环节。然而,目前对异性纤维的检测主要还是采用人工检测,而这种方法通常检测效率低下,容易受检测环境的影响[34]。
随着机器视觉理论的发展,在棉花异性纤维检测中引入机器视觉,使得异性纤维检测的工作效率得到了改善,异性纤维检出率提高了[56]。……
