皮棉表面多类异性纤维的高光谱图像检测
Detection of foreign materials on the surface of ginned cotton by hyper-spectral imaging
-
摘要: 为了考察高光谱成像技术检测多类异性纤维共存时的性能,也为其今后用于商业化皮棉质量评价,或在线皮棉杂质分拣提供研究基础,试验采集丙纶丝和毛发共存的样本高光谱图像。依据单一异性纤维分割的最佳波段图像,获得灰度平均和小波变换后融合图像。通过对比小波变换融合图像,灰度平均图像,以及全波段下主成分和独立成分得分图像,获得用于目标分割的最佳图像。采用最优特征集和分类树判别方法,判别最佳图像分割后的连通区域,剔除噪声点和假阳性。依据目标判定原则,训练集和独立验证集的异性纤维识别率分别为84.09%和75.86%。结果表Abstract: In order to investigate the potential for detecting multi-foreign materials on the surface of ginned cotton by hyper-spectral imaging, and also to provide the research basis for applying this imaging to evaluate cotton quality or to sort the trashes in li