基于r相关不均匀双线性插值策略的四目视觉圆柱周标拼接研究OA

Aiming at solving problems such as difficult quality inspection and high false detection rate of cylindrical product circumferential labels,a new scheme of stable and high-quality stitching of cylindrical circumferential labels based on four-eye vision is proposed.Firstly,a two-layer fusion method based on four quasi-central axes is given to obtain the real position of the cylinder.Then,a novel strategy of r-correlated uneven bilinear interpolation is employed to achieve high-quality unfolding of cylindrical label images.On this basis,the feature matching of two adjacent images in the cylindrical label unfolded image is performed by normalized cross-correlation(NCC)algorithm,and the weighted fusion algorithm is used for image stitching and fusion.The experiments demonstrate that the real-time stitching time of the proposed scheme is less than 400 ms for circumferential labels of cylindrical products with different radii,which meets the requirements of industrial production for the accuracy and stability detection of the cylindrical circumferential label.