摘要:AbstractIn this paper a non-contact weld measurement method based on machine vision is proposed for quality inspection of weld height and width. The presented method is developed for yet not restricted to body-in-white (BIW) assembly of high-speed train carriages. The proposed measurement method increases the anti-noise ability by adopting median filtering algorithm in an adaptive window. The center extraction algorithm combining gray centroid and Hessian matrix are designed to realize the extraction of sub-pixel centerline stripes. The dimensional measurement of the weld features is completed by mapping relationship between the pixel coordinate system and the world coordinate system obtained by the Zhang's calibration method. The experimental results show that the measurement accuracy can reach 0.05mm while maintain high level of real-time performance.