点击次数：

发表时间：2022-10-04

发表刊物：光学精密工程

所属单位：机械工程学院

期号：11

页面范围：3041-3050

ISSN号：1004-924X

摘要：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 1 pixel were determined. By projecting the lasers onto different surfaces, the correlation quality coefficients of left and right images QLRwere determined to be [0.8, 1). Finally, experiments were performed in a forge workshop, and the reconstructed results of forgings' heights are 1 747.0 mm and 491.3 mm. 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. © 2015, SCIENCE PRESS. All right reserved.

备注：新增回溯数据