苑玮琦,李绍丽,李德健.基于纹理脊线特征融合的木材表面裂缝检测[J].仪器仪表学报,2017,38(2):436-444
基于纹理脊线特征融合的木材表面裂缝检测
Wood surface crevice detection based on fusion of texture ridge line features
  
DOI:
中文关键词:  纹理脊线特征  特征融合  木材裂缝检测  模糊规则
英文关键词:texture ridge line feature  feature fusion  wood crevice detection  fuzzy rule
基金项目:国家自然科学基金(61271325)项目资助
作者单位
苑玮琦 1.沈阳工业大学 觉检测技术研究所沈阳110870;2.辽宁省机器视觉重点实验室沈阳110870 
李绍丽 1.沈阳工业大学 觉检测技术研究所沈阳110870;2.辽宁省机器视觉重点实验室沈阳110870 
李德健 1.沈阳工业大学 觉检测技术研究所沈阳110870;2.辽宁省机器视觉重点实验室沈阳110870 
AuthorInstitution
Yuan Weiqi 1.Computer Visition Group, Shenyang University of Technology, Shenyang 110870, China; 2. Key Laboratory of Machine Vision, Shenyang 110870, China 
Li Shaoli 1.Computer Visition Group, Shenyang University of Technology, Shenyang 110870, China; 2. Key Laboratory of Machine Vision, Shenyang 110870, China 
Li Dejian 1.Computer Visition Group, Shenyang University of Technology, Shenyang 110870, China; 2. Key Laboratory of Machine Vision, Shenyang 110870, China 
摘要点击次数: 491
全文下载次数: 483
中文摘要:
      裂缝是木材表面一种严重缺陷,对木材的加工和使用影响极大,然而,由于裂缝与木材表面的矿物线具有诸多相似之处,因此如何准确地将裂缝纹理识别出来是一个亟待解决的问题。提出了一种基于纹理脊线特征融合的检测方法,首先建立纹理脊线灰度和形态特征提取基本模型;然后分割出木材表面全部纹理区域,并根据模糊规则提取出条状纹理,包括裂缝和矿物线;最后根据建立的模型提取条状纹理的两种脊线特征,并进行特征信息融合得到复合判别因子,最终通过融合结果与预设阈值的数值关系识别裂缝纹理。在自建图库上进行了测试,结果显示所用方法对裂缝缺陷识别的等误率仅为4.64%,相对于其他经典特征提取和纹理识别方法其等误率最少降低了10.06%,表明了本方法的高效性,具有实际应用价值。
英文摘要:
      Crevice is a serious defect on wood surface which influences the process and usage of wood seriously. However, the mineral streak of the crevice on wood surface is similar. Hence, how to identify the crevice texture accurately is an urgent problem to be solved. A detection method based on texture ridge line features fusion is proposed. Firstly, the basic model of extracting texture ridges line grayscale characteristic and shape feature is formulated. Secondly, all wood surface texture regions are segmented and the strip texture is extracted according to the fuzzy rules, including crevice and the mineral streak. Finally, these two ridge line features of striped texture are extracted by the model established. The composite discriminant factor is obtained through feature information fusion. The crevice is identified by analyzing the relationship of fusion result and the predetermined threshold value. The proposed method is evaluated by the self built atlas. The experimental results show that the Equal Error Rate (EER) of crevice defection is 4.64%. Compared with the other classical feature extraction and texture recognition methods, the EER of the proposed method is decreased by 10.06% at least. The efficiency and the practical application value of this method is proved.
查看全文  查看/发表评论  下载PDF阅读器