人造板表面缺陷检测图像自适应快速阈值分割算法
|
摘要
【目的】提出一种自适应快速阈值图像分割算法,为人造板表面缺陷在线检测提供支持。【方法】首先将整幅图像划分成若干子区域,通过计算子区域的方差对缺陷进行定位,提取出缺陷所在区域,只对缺陷区域进行图像分割,解决小面积目标难以准确分割的问题。然后对缺陷区域的一维灰度直方图进行处理,直方图平滑后去除掉不显著波峰,根据处理后保留的主要波峰数量和位置自适应地确定分割阈值个数以及每个阈值的分割区间,实现当图像中出现多种类型缺陷时算法自动确定分割阈值个数。最后,通过分析Otsu算法,将阈值穷举搜索改进为条件搜索并限定搜索方向,在每个分割区间内使用改进的Otsu算法对阈值进行搜索,提高搜索速度。【结果】对板面存在油污、大刨花、胶斑、杂物、松软5种类型缺陷的人造板表面图像进行分割,在板面缺陷数量、类型不固定的情况下,算法可以自适应地确定分割阈值个数,在15 ms内将各种类型缺陷从人造板表面图像中分割出来,平均分割准确率达97%。【结论】自适应快速阈值分割算法能够快速、准确将缺陷从人造板表面图像中分离出来,在执行速度和分割效果上均满足在线缺陷检测系统的要求,可为人造板表面缺陷在线检测提供新思路。
【Objective】 An adaptive fast thresholding image segmentation algorithm was proposed in this paper, which could quickly and accurately separate the defects from the surface images of wood-based panels,and provide support for on-line detection of wood-based panel surface defects.【Method】 Firstly,the algorithm divided the whole image into several sub-regions. Secondly,the defect areas were located by calculating the variance of each sub-region. And then,the image segment was only done in defect areas for solving the problem of accurate segmentation of small targets.The one-dimension gray scale histogram of extracted defect area was processedusing histogram smoothing to remove the non-significant peaks.According to the main wave peaks reserved in the histogram after the processing,the number of the thresholds and segmentation interval for each threshold were determined adaptively. At each interval,the threshold was searched using an improved fast Otsu segmentation algorithm.Through the analysis of the Otsu algorithm,the threshold was found by a conditional search instead of the exhaustive search and the search direction was specified. In each segmentation interval, the improved fast Otsu segmentation algorithm was used to search the threshold,which improved the search speed.【Result】 The segmentations of surface images of wood-based panels with five types of defects such as oil stains,big wood shavings,glue spots,sundries and loose regions were done using the adaptive fast algorithm proposed in this paper. Although the number and type of the defects were not fixed,this algorithm still could determine the number of the thresholds automatically. All kinds of defects were separated from the surface images in 15 ms with a above 97% segmentation accuracy rate.【Conclusion】 The adaptive fast threshold segmentation algorithm presented in this paper can quickly and accurately separate the defects from the surfaces of the panels,and the execution speed and the segmentation effect meets the requirements of the on-line defect detection system. It provides a new approach for automatic on-line detection of surface defects on wood-based panels.
【Objective】 An adaptive fast thresholding image segmentation algorithm was proposed in this paper, which could quickly and accurately separate the defects from the surface images of wood-based panels,and provide support for on-line detection of wood-based panel surface defects.【Method】 Firstly,the algorithm divided the whole image into several sub-regions. Secondly,the defect areas were located by calculating the variance of each sub-region. And then,the image segment was only done in defect areas for solving the problem of accurate segmentation of small targets.The one-dimension gray scale histogram of extracted defect area was processedusing histogram smoothing to remove the non-significant peaks.According to the main wave peaks reserved in the histogram after the processing,the number of the thresholds and segmentation interval for each threshold were determined adaptively. At each interval,the threshold was searched using an improved fast Otsu segmentation algorithm.Through the analysis of the Otsu algorithm,the threshold was found by a conditional search instead of the exhaustive search and the search direction was specified. In each segmentation interval, the improved fast Otsu segmentation algorithm was used to search the threshold,which improved the search speed.【Result】 The segmentations of surface images of wood-based panels with five types of defects such as oil stains,big wood shavings,glue spots,sundries and loose regions were done using the adaptive fast algorithm proposed in this paper. Although the number and type of the defects were not fixed,this algorithm still could determine the number of the thresholds automatically. All kinds of defects were separated from the surface images in 15 ms with a above 97% segmentation accuracy rate.【Conclusion】 The adaptive fast threshold segmentation algorithm presented in this paper can quickly and accurately separate the defects from the surfaces of the panels,and the execution speed and the segmentation effect meets the requirements of the on-line defect detection system. It provides a new approach for automatic on-line detection of surface defects on wood-based panels.
引文
崔丽群,黄殿平,宋 晓. 2017. 基于云模型鱼群算法的多阈值图像分割研究.计算机工程与应用,53(6): 204-208.(Cui L Q,Huang D P,Song X.2017. Multi-threshold method for image segmentation based on cloud model artificial fish swarm algorithm. Computer Engineering and Applications,53(6): 204-208.[in Chinese])
康杰红,马 苗. 2012. 基于蛙跳算法与 Otsu 法的图像多阈值分割技术. 云南大学学报: 自然科学版,34(6): 634-640.(Kang J H,Ma M.2012. Multilevel thresholding segmentation based on shuffled frog leaping algorithm and Otsu method. Journal of Yunnan University: Natural Sciences,34(6): 634-640. [in Chinese])
刘健庄,谢维信,高新波. 1995. 多阈值图像分割的遗传算法方法. 模式识别与人工智能,8(增): 126-132.(Liu J Z,Xie W X,Gao X B. 1995. Multi-thresholded image segmentation with genetic algorithms. Pattern Recognition and Artificial Intelligence,8(suppl): 126-132. [in Chinese])
刘 翔. 2017. 多阈值Otsu快速算法的研究. 长春: 吉林大学硕士学位论文.(Liu X.2017.Research on fast algorithm of multi-threshold Otsu. Changchun: MS thesis of Jilin University. [in Chinese])
刘 艳,赵英良. 2011. Otsu多阈值快速求解算法.计算机应用, 31(12): 3363-3365.(Liu Y,Zhao Y L. 2011. Quick approach of multi-threshold Otsu method for image segmentation. Journal of Computer Applications,31(12): 3363-3365. [in Chinese])
申铉京,刘 翔,陈海鹏. 2017. 基于多阈值Otsu准则的阈值分割快速计算. 电子与信息学报,39(1): 144-149.(Shen X J,Liu X,Chen H P. 2017. Fast computation of threshold based on multi-threshold Otsu criterion. Journal of Electronics & Information Technology,39(1): 144-149. [in Chinese])
王祥科,郑志强. 2006. Otsu多阈值快速分割算法及其在彩色图像中的应用.计算机应用, 26(S1): 14-15.(Wang X K,Zheng Z Q. 2006. Otsu multi threshold fast segmentation algorithm and application in color images. Journal of Computer Applications, 26(S1): 14-15. [in Chinese])
许向阳,宋恩民,金良海.2009. Otsu准则的阈值性质分析.电子学报,37(12): 2716-2719.(Xu X Y,Song E M,Jin L H. 2009. Characteristic analysis of threshold based on Otsu criterion. Acta Electronica Sinica,37(12): 2716-2719. [in Chinese])
阳树洪. 2014. 灰度图像阈值分割的自适应和快速算法研究. 重庆: 重庆大学博士学位论文.(Yang S H. 2014. Study on the adaptive and fast algrithm of gray scale image thresholding. Chongqing: PhD thesis of Chongqing University. [in Chinese])
詹 曙,梁植程,谢栋栋. 2017. 前列腺磁共振图像分割的反卷积神经网络方法.中国图象图形学报,22(4): 516-522.(Zhan S,Liang Z C,Xie D D. 2017. Deconvolutional neural network for prostate MRI segmentation. Journal of Image and Grophics,22(4): 516-522. [in Chinese])
Kamal H,Moussa D,Patrick S.2008. A multilevel automatic thresholding method based on a genetic algorithm for a fast image segmentation. Computer Vision & Image Understanding,109(2): 163-175.
Otsu N.1979. A threshold selection method from gray-level histograms. IEEE Transactions on System Man and Cybemetic,9(1): 62-66.
Raja N S M,Rajinikanth V,Latha K.2014. Otsu based optimal multilevel image thresholding using firefly algorithm.Hindawi Publishing Crop, 1-17.
康杰红,马 苗. 2012. 基于蛙跳算法与 Otsu 法的图像多阈值分割技术. 云南大学学报: 自然科学版,34(6): 634-640.(Kang J H,Ma M.2012. Multilevel thresholding segmentation based on shuffled frog leaping algorithm and Otsu method. Journal of Yunnan University: Natural Sciences,34(6): 634-640. [in Chinese])
刘健庄,谢维信,高新波. 1995. 多阈值图像分割的遗传算法方法. 模式识别与人工智能,8(增): 126-132.(Liu J Z,Xie W X,Gao X B. 1995. Multi-thresholded image segmentation with genetic algorithms. Pattern Recognition and Artificial Intelligence,8(suppl): 126-132. [in Chinese])
刘 翔. 2017. 多阈值Otsu快速算法的研究. 长春: 吉林大学硕士学位论文.(Liu X.2017.Research on fast algorithm of multi-threshold Otsu. Changchun: MS thesis of Jilin University. [in Chinese])
刘 艳,赵英良. 2011. Otsu多阈值快速求解算法.计算机应用, 31(12): 3363-3365.(Liu Y,Zhao Y L. 2011. Quick approach of multi-threshold Otsu method for image segmentation. Journal of Computer Applications,31(12): 3363-3365. [in Chinese])
申铉京,刘 翔,陈海鹏. 2017. 基于多阈值Otsu准则的阈值分割快速计算. 电子与信息学报,39(1): 144-149.(Shen X J,Liu X,Chen H P. 2017. Fast computation of threshold based on multi-threshold Otsu criterion. Journal of Electronics & Information Technology,39(1): 144-149. [in Chinese])
王祥科,郑志强. 2006. Otsu多阈值快速分割算法及其在彩色图像中的应用.计算机应用, 26(S1): 14-15.(Wang X K,Zheng Z Q. 2006. Otsu multi threshold fast segmentation algorithm and application in color images. Journal of Computer Applications, 26(S1): 14-15. [in Chinese])
许向阳,宋恩民,金良海.2009. Otsu准则的阈值性质分析.电子学报,37(12): 2716-2719.(Xu X Y,Song E M,Jin L H. 2009. Characteristic analysis of threshold based on Otsu criterion. Acta Electronica Sinica,37(12): 2716-2719. [in Chinese])
阳树洪. 2014. 灰度图像阈值分割的自适应和快速算法研究. 重庆: 重庆大学博士学位论文.(Yang S H. 2014. Study on the adaptive and fast algrithm of gray scale image thresholding. Chongqing: PhD thesis of Chongqing University. [in Chinese])
詹 曙,梁植程,谢栋栋. 2017. 前列腺磁共振图像分割的反卷积神经网络方法.中国图象图形学报,22(4): 516-522.(Zhan S,Liang Z C,Xie D D. 2017. Deconvolutional neural network for prostate MRI segmentation. Journal of Image and Grophics,22(4): 516-522. [in Chinese])
Kamal H,Moussa D,Patrick S.2008. A multilevel automatic thresholding method based on a genetic algorithm for a fast image segmentation. Computer Vision & Image Understanding,109(2): 163-175.
Otsu N.1979. A threshold selection method from gray-level histograms. IEEE Transactions on System Man and Cybemetic,9(1): 62-66.
Raja N S M,Rajinikanth V,Latha K.2014. Otsu based optimal multilevel image thresholding using firefly algorithm.Hindawi Publishing Crop, 1-17.