基于Ward反射模型的航天密封圈表面缺陷检测
|
摘要
为有效检测航天系统密封圈表面的缺陷,提出一种基于Ward反射模型的检测方法。根据Ward反射模型计算密封圈曲面在不同光源方向和不同观测方向下的辐射强度。通过计算得到密封圈表面的灰度阈值,提取出相机采集的密封圈表面高亮区域。对于不带缺陷的密封圈,当某区域产生缺陷后,该区域各像素点的表面法向发生改变,造成在相同的光源方向与观测方向下,表面灰度图像中高亮度区域与基于Ward模型的表面高辐射区域在数量和位置上不对等。由基于Ward模型的辐射图像确定比值ks,提取出该密封圈不带缺陷时其灰度图像中的高亮度区域。在提取出的相机采集的密封圈表面高亮区域中,结合图像噪声、密封圈表面细微粉尘和若该密封圈不带缺陷时的灰度图像中的高亮度区域这三者面积,筛选出密封圈缺陷区域。实验结果表明,该方法能够有效地提取密封圈表面的凹痕、飞边等缺陷,并能给出占据的像素面积。
In order to detect the surface defects of the sealing rings more effectively,a detection method is proposed based on Ward reflection model. It calculates radiation intensity of sealing ring surface under different light source directions and different observation directions according to Ward reflection model. The gray threshold value of the sealing surface is calculated, and the highlighted region of the sealing ring surface is extracted by the CCD camera. For sealing rings without defects,when a region is defective,the surface normal direction of each pixel in the region is changed compared with the original one,this results the high brightness region in the surface gray image is not equal to the number and position of the surface high radiation area based on the Ward model in the same direction of the light source and the direction of observation. The ratio ks is determined by the radiation image the Ward model based on,and the high brightness region of the gray image is extracted when the sealing ring is not defective. In the highlight area of the sealing ring surface which is acquired by the CCD camera,combined with the image noise,the fine dust on the surface of the sealing ring and the high brightness area in the gray image of the sealing rings without defects,the defect area of the sealing rings is selectell. Experimental results show that the proposed method can effectively extract the dent,flash and other defects on the sealing rings, and the pixel area of the defects is given.
In order to detect the surface defects of the sealing rings more effectively,a detection method is proposed based on Ward reflection model. It calculates radiation intensity of sealing ring surface under different light source directions and different observation directions according to Ward reflection model. The gray threshold value of the sealing surface is calculated, and the highlighted region of the sealing ring surface is extracted by the CCD camera. For sealing rings without defects,when a region is defective,the surface normal direction of each pixel in the region is changed compared with the original one,this results the high brightness region in the surface gray image is not equal to the number and position of the surface high radiation area based on the Ward model in the same direction of the light source and the direction of observation. The ratio ks is determined by the radiation image the Ward model based on,and the high brightness region of the gray image is extracted when the sealing ring is not defective. In the highlight area of the sealing ring surface which is acquired by the CCD camera,combined with the image noise,the fine dust on the surface of the sealing ring and the high brightness area in the gray image of the sealing rings without defects,the defect area of the sealing rings is selectell. Experimental results show that the proposed method can effectively extract the dent,flash and other defects on the sealing rings, and the pixel area of the defects is given.
引文
[1]GUO W,CUI W,YU T,et al.O-Ring static sealing reliability model and influence factors analysis[C]//Proceedings of 2013 International Conference on Quality,Reliability,Risk,Maintenance,and Safety Engineering.Washington D.C.,USA:IEEE Press,2013:116-121.
[2]国家质量检督检验检疫总局.液压气动用O形橡胶密封圈第2部分:外观质量检验规范:GB/T 3452.2-2007[S].北京:中国标准出版社,2007.
[3]李晓舟,于化东,于占江,等.微小尺寸零件表面缺陷光学检测方法[J].兵工学报,2011,32(7):872-877.
[4]叶声华,秦树人.现代测试计量技术及仪器的发展[J].中国测试,2009,35(2):1-6.
[5]朱红军,高潮,郭永彩.基于计算机视觉的非朗伯表面三维重构[J].强激光与粒子束,2014,26(1):289-299.
[6]OREN M,NAYAR K.Generalization of the lambertian model and implications for machine vision[J].International Journal of Computer Vision,1995,14(3):227-251.
[7]IKEHATA S,WIPF D,MATSUSHITA Y,et al.Photometric stereo using sparse Bayesian regression for general diffuse surfaces[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2014,36(9):1816-1831.
[8]尹芳,陈田田.均场退火算法在单幅灰度图像高光检测与恢复中的应用[J].计算机辅助设计与图形学学报,2017,29(5):829-837.
[9]葛文谦,赵慧洁.高曲率微小零件表面缺陷及外形尺寸测量系统研究[J].红外与激光工程,2008,37(s1):274-278.
[10]吴晓波,杨永琴.图像测量技术的新应用[J].光学精密工程,1998,6(3):10-15.
[11]袁小翠,吴禄慎,陈华伟.钢轨表面缺陷检测的图像预处理改进算法[J].计算机辅助设计与图形学学报,2014,26(5):800-805.
[12]刘蕴辉,刘铁,王权良,等.基于图像处理的铁轨表面缺陷检测算法[J].计算机工程,2007,33(11):236-238.
[13]WARD G L.Measuring and modeling anisotropic reflection[J].Computer Graphics and Interactive Techniques,1992,26(2):265-272.
[14]郑作勇.基于图像的带高光物体的形状和反射属性建模技术研究[D].上海:上海交通大学,2009.
[15]特霍尔德-霍恩.机器视觉[M].王亮,蒋欣兰,译.北京:中国青年出版社,2014.
[16]李云飞,司国良,郭永飞.科学级CCD相机的噪声分析及处理技术[J].光学精密工程,2005,13(z1):158-163.
[2]国家质量检督检验检疫总局.液压气动用O形橡胶密封圈第2部分:外观质量检验规范:GB/T 3452.2-2007[S].北京:中国标准出版社,2007.
[3]李晓舟,于化东,于占江,等.微小尺寸零件表面缺陷光学检测方法[J].兵工学报,2011,32(7):872-877.
[4]叶声华,秦树人.现代测试计量技术及仪器的发展[J].中国测试,2009,35(2):1-6.
[5]朱红军,高潮,郭永彩.基于计算机视觉的非朗伯表面三维重构[J].强激光与粒子束,2014,26(1):289-299.
[6]OREN M,NAYAR K.Generalization of the lambertian model and implications for machine vision[J].International Journal of Computer Vision,1995,14(3):227-251.
[7]IKEHATA S,WIPF D,MATSUSHITA Y,et al.Photometric stereo using sparse Bayesian regression for general diffuse surfaces[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2014,36(9):1816-1831.
[8]尹芳,陈田田.均场退火算法在单幅灰度图像高光检测与恢复中的应用[J].计算机辅助设计与图形学学报,2017,29(5):829-837.
[9]葛文谦,赵慧洁.高曲率微小零件表面缺陷及外形尺寸测量系统研究[J].红外与激光工程,2008,37(s1):274-278.
[10]吴晓波,杨永琴.图像测量技术的新应用[J].光学精密工程,1998,6(3):10-15.
[11]袁小翠,吴禄慎,陈华伟.钢轨表面缺陷检测的图像预处理改进算法[J].计算机辅助设计与图形学学报,2014,26(5):800-805.
[12]刘蕴辉,刘铁,王权良,等.基于图像处理的铁轨表面缺陷检测算法[J].计算机工程,2007,33(11):236-238.
[13]WARD G L.Measuring and modeling anisotropic reflection[J].Computer Graphics and Interactive Techniques,1992,26(2):265-272.
[14]郑作勇.基于图像的带高光物体的形状和反射属性建模技术研究[D].上海:上海交通大学,2009.
[15]特霍尔德-霍恩.机器视觉[M].王亮,蒋欣兰,译.北京:中国青年出版社,2014.
[16]李云飞,司国良,郭永飞.科学级CCD相机的噪声分析及处理技术[J].光学精密工程,2005,13(z1):158-163.