摘要
为实现镀膜金属带表面缺陷在线实时准确识别,提出一种基于二阶振荡粒子群优化的Otsu算法。通过加入二阶振荡粒子群优化Otsu算法最佳阈值选取,加快分割算法收敛速度,提高图像分割效率。针对图像分割结果,进行形态学分析与孔洞填充处理,得到精确分割结果。实验结果表明,该算法实现阈值分割仅需25 ms,比传统Otsu算法提高了约88%,比迭代最佳阈值法节省至少90%的时间,全局收敛性优于标准粒子群Otsu算法,在分割速度提高的同时能较好地实现缺陷准确分割。
To realize the real-time accurate identification of surface defects of coated metal strip,an Otsu algorithm based on the second-order oscillation particle swarm optimization was proposed.By optimizing the threshold of the second order oscillatory particle swarm Otsu algorithm,the converging of the segmentation algorithm was accelerated and the efficiency of the image segmentation was increased.Based on the image segmentation,morphological analysis and hole filling process were carried out and accurate segmentation results were obtained.Experimental results show that the threshold segmenting time using the proposed algorithm is only 25 ms,which is about 88%higher than that of the Otsu algorithm and saves at least 90%of the time compared with the iterative optimal threshold method.The global convergence is better than the Otsu algorithm based on particle swarm,and the proposed algorithm can improve both the segmentation efficiency and the accuracy.
作者
赵梦超
孔令成
谭治英
ZHAO Meng-chao;KONG Ling-cheng;TAN Zhi-ying(College of Information Science and Engineering,Changzhou University,Changzhou 213164,China;Hefei Institutes of Physical Science,Chinese Academy of Sciences,Hefei 230031,China;Institutes of Advanced Manufacturing Technology,Chinese Academy of Sciences,Changzhou 213164,China)
出处
《计算机工程与设计》
北大核心
2018年第9期2811-2816,共6页
Computer Engineering and Design
基金
国家自然科学基金项目(61401437)
关键词
缺陷检测
二阶振荡粒子群
OTSU算法
阈值分割
适应度函数
defect detection
second order oscillatory particle swarm
Otsu algorithm
threshold segmentation
fitness function
