Abstract: In automated welding process based on vision method,images containing information of welding seam are obtained in series through vision sensor and processed real time to get central position of welding seam.However,because welding seam images are complicated,it is very important to choose suitable and effective processing algorithm in order to get accurately the information.This paper put forward a series of processing steps for welding seam images which include median filtering,creating binary image by adaptive threshold,acnode filtering,and edge detecting,searching central position of welding seam.Creating binary image by adaptive threshold and acnode filtering are focused.In each processing step four different welding seam images are used for observation and comparison.The result shows these processing algorithms arc relatively effective.