摘要
构建了基于计算机视觉的齿形链链板的测量系统,对直线和圆的边缘定位,提出了新的亚像素算法,计算了齿形链链板的销孔直径及圆度误差、链板节距和两直边夹角,对测量结果进行了误差分析。将图像处理、参数计算、标定等整个测量过程集成起来,通过标准样件对被测零件图像进行坐标变换,确定待测区域并完成标定。实验表明,改进最小二乘直线拟合定位误差是常规最小二乘拟合误差的1/4。而圆的亚像素定位可使CCD的分辨率提高42倍,大大提高图像边缘的定位精度。利用亚像素方法测量时,销孔直径为2.563mm,链板节距为6.296mm,两直边夹角为60.453°,圆度误差为14.7μm。单一链板的整个测量过程不足1s。
The measurement system on chain board of tooth shape chain was established based on computer vision,and a new kind of sub-pixel algorithm for edge orientation of beeline and circle was put forward.Pinhole diameter,roundness error,chain board pitch and angle between two right sides of chain board were calculated,then the error analysis to measurement results was performed.The whole course containing image process,parameter calculation,and calibration etc.was integrated.Coordination transmission was performed,measurement area was confirmed,and the calibration for part image to be measured was finished through standard sample.Experimental results show that the orientation error of improved least-square linear fitting method is one quarter of ordinary least-square linear fitting error under small difference of orientation time.The sub-pixel orientation of circle makes resolution of CCD increase 42 times,which enhances greatly orientation precision of image edge.Pinhole,pitch,angle between two right sides and roundness are respectively 2.563mm,6.296mm,60.453° and 14.7μm,while taking sub-pixel method.The whole measurement course for one chain board spends less than one second.
出处
《中国机械工程》
EI
CAS
CSCD
北大核心
2008年第13期1539-1543,共5页
China Mechanical Engineering
基金
吉林省科技发展计划资助项目(20040534)
关键词
CCD
齿形链
亚像素
集成
误差
CCD
tooth shape chain
sub-pixel
integration
error
