Abstract: To the problems of low defection precision and self-adaption of high stress belt conveyor joint twitch detection algorithm，another high stress belt conveyor joint twitch detection algorithm was put forward，which combining improved Y-difference and SURF features matching together.First，the X-ray image belt joint points was detected，and then the SURF feature vectors of detection belt joints were extracted，and feature matching was proceeded according feature value，mapping relation between present image and standard image was seeking，normalization coefficient was calculated，belt joints twitch trend and twitch distance were analyzed according normalization coefficient.On the basis of application verification，the method precision was higher，and it has some self-adaption for speed change，migration and vibration，so safety transportation of high stress belt conveyor was guaranteed.