摘要
针对目前茧丝疵点检测自动化程度低、分割精度不高等问题,提出一种基于轮廓拟合与距离分析的茧丝疵点分割算法。背景糙点的外接矩形面积较小,据此将其剔除。针对疵点边缘分割易失真的问题,对茧丝曲线轮廓重新进行多线段拟合。根据采用多线段拟合疵点与丝干轮廓时两者相邻线段端点距离差异较大的特点,设置距离阈值先删除疵点轮廓,再与曲线轮廓作差得到疵点图像。研究结果表明,基于轮廓拟合与距离分析的茧丝疵点分割算法的平均交并比与平均F1分数分别达85.57%和91.50%,能较好地保留疵点边缘轮廓信息,同时显著提升疵点提取精度。
A cocoon silk defect segmentation algorithm based on contour fitting and distance analysis was proposed to address the current issues of low automation and segmentation accuracy in cocoon silk defect detection.The background rough points were removed based on the smaller area of the bounding rectangle of the background rough points.In response to the problem of distortion in defect edge segmentation,the cocoon curve contour was fitted with multiple line segments.Taking advantage of the significant difference in the distance between adjacent line segment endpoints when fitting the defect contour and the silk stem contour,the defect image was obtained by setting distance threshold to remove the contour of the defects first and then make the difference with the curve contour.The research results show that the average intersection to union ratio and average F1 score of the algorithm based on contour fitting and distance analysis are 85.57% and 91.50% respectively,which can better retain the edge contour information of the defects and improve the extraction accuracy of the defects significantly.
作者
胡涛涛
孙卫红
梁曼
邵铁锋
Hu Taotao;Sun Weihong;Liang Man;Shao Tiefeng(School of Mechanical and Electrical Engineering,China Jiliang University,Hangzhou 310018,Zhejiang,China;Institute of Cocoon and Silk Quality Inspection Technology,China Jiliang University,Hangzhou 310018,Zhejiang,China)
出处
《产业用纺织品》
2023年第4期47-53,共7页
Technical Textiles
基金
中国纤维质量监测中心项目(OITC-G190281374)
浙江省基础公益研究计划项目(LGG20E050014)。
关键词
茧丝疵点
疵点分割
轮廓拟合
距离分析
多线段拟合
端点距离
cocoon silk defect
defect segmentation
contour fitting
distance analysis
multiple line segment fitting
endpoint distance
