摘要
提出运用织物图像的直线纹理特征 ,由概率统计生成直方图 ,有效地提取织物图像的特征波形 ,经大量检测试验证明特征波可靠稳定。基于波形特征参数对比能准确定位织物纹理结构的异常位置 ,正确识别织物疵点。为简化识别算法可引入阈值过滤 ,允许正常波形参数通过过滤 ,而记录异常波形变化 ,用过滤单一的正常纹理来识别类型复杂的织物疵点。
Based on linear feature texture of fabric,achieved statistic histogram can abstract fabric characteristics perfectly.These feature waves show reliable and stable during numerous tests.With comparing wave features,abrupt or abnormal changes can be located precisely and fabric defects can be picked out correctly.To simplify identification,threshold filter should be introducd to overpass general wave and record abnormal status.Discrimination of various and complex defects can form a simple by filtering general unique texture.
出处
《纺织学报》
EI
CAS
CSCD
北大核心
2005年第2期121-123,共3页
Journal of Textile Research
基金
陕西省教育厅资助项目 (0 4JK183)
关键词
直方图
自适应
纹理识别
特征强化
阈值过滤
histogram
self-adaptive
texture identification
feature strengthening
threshold filtration
