摘要
使用SLIC(简单线性迭代聚类)超像素图像分割方法将木材表面缺陷图像预分割,并从提高算法速度和自适应阈值2方面对超像素合并算法进行改进;分析了DBSCAN(具有噪声的基于密度的聚类方法)聚类用于该类超像素合并中算法的复杂度,提出了自适应阈值的快速DBSCAN超像素合并算法来取得缺陷分割图像。结果表明:改进后的算法对于3类缺陷都能很好的分割,并且算法复杂度低;分割及合并的总时间为0.35 s左右,能满足在线分选的要求。
We used SLIC super pixel image segmentation method to pre-segment wood surface defect images,and improved the algorithm by speed and adaptive threshold. We analyzed the complexity of DBSCAN clustering algorithm,and proposed a fast DBSCAN super pixel merging algorithm based on adaptive threshold to obtain the image of defect segmentation. The improved algorithm can perform well in three kinds of defects with the low algorithm complexity. The total time of the segmentation and merging was about 0. 35 s,which met the requirements of on-line sorting.
出处
《东北林业大学学报》
CAS
CSCD
北大核心
2015年第10期97-102,共6页
Journal of Northeast Forestry University
基金
中央高校基本科研业务费专项资金项目(2572015BB11)
黑龙江省自然科学基金项目(QC2015080)
关键词
木材表面缺陷
超像素
图像分割
Wood surface defect
Super pixel
Image segmentation