Abstract: An efficient automatic detection system for multi-pass welding seam profile parameters based on band laser CCD(charge-coupled device) visual sensor is presented.By analyzing the characteristics of the image of multi-pass weld laser stripes,an image processing approach to characterize irregular welding groove is proposed,which can be used to identify the characteristics of multi-pass welding seam.The developed image processing procedures is threshold processing,image denoising,image centerline extraction,straight-line fitting and the feature point identifying.The proposed detecting system can provide the accurate signal for seam tracking,which is tremendously significant to improve the welding quality and automatization for thick plates welding.