互参考激光光条图像质量评价  

Cross-referenced image quality assessment for laser stripes

作　　者：贾振元[1] 王灵丽[1] 刘巍[1] 刘阳[1] 樊超楠[1] 

机构地区：[1]大连理工大学机械工程学院,辽宁大连116024

出　　处：《光学精密工程》2015年第11期3041-3050,共10页Optics and Precision Engineering

基　　金：国家自然科学基金仪器专项资助项目(No.51227004);国家自然科学基金面上项目(No.51375075);国家自然科学基金创新研究群体项目(No.51321004);国家973重点基础研究发展计划资助项目(No.2014CB046504)

摘　　要：提出了互参考激光光条图像质量评价方法用于检测激光辅助立体视觉测量中的图像质量。分析了立体视觉测量中单幅激光光条图像的灰度分布特性以及左右激光光条图像灰度分布的相关性。利用光条截面灰度分布梯度表征图像清晰度,并建立了综合考虑单幅图像质量与左右图像相关质量的互参考激光光条图像质量评价方法。然后,确定质量系数的阈值,通过分析噪声对光条提取误差的影响,确定提取误差不高于1个像素的单幅图像质量系数QL,QR∈[0.2,1),通过不同表面激光光条投影实验,确定左右图像相关质量系数QLR∈[0.8,1)。最后,在锻造车间进行了现场实验,重建方形锻件高度为1 747.0mm和491.3mm。实验结果显示,清晰的光条图像质量系数符合确定阈值,受噪声干扰的光条图像质量系数偏离设定阈值,验证了本文提出的质量评价方法能较好地反映激光光条图像缺陷。A cross-referenced image quality assessment method for laser stripes was proposed to detect image quality in laser-aided stereo vision measurement.The intensity profile of laser stripes in a single image and the correlation of the same laser stripes acquired in left and right images were analyzed.Then,the definition of the images was characterized by grayscale profile gradients and a model of cross-referenced image quality assessment was proposed by taking both the single image quality and the correlation quality of left and right images into account.To determine thresholds of quality coefficients,the effect of noise on laser stripes＇extraction error was analyzed and simulated,and the quality coefficients of single image QL,QRbelongs to[0.2,1）with the extraction error of less than 1pixel were determined.By projecting the lasers onto different surfaces,the correlation quality coefficients of left and right images QLR were determined to be [0.8,1）.Finally,experiments were performed in a forge workshop,and the reconstructed results of forgings＇heights are 1 747.0mm and491.3mm.Experimental results show that the image quality coefficients of clear laser stripes conform to the thresholds,while the image quality coefficients of the laser stripes interfered by noise are deviated from the threshold.It concludes that the proposed method of image quality assessment reflects the defects in images of laser stripes very well.

关 键 词：立体视觉 图像质量检测 互参考图像质量评价 激光光条图像 阈值 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术] TP242.62[自动化与计算机技术—计算机科学与技术]