It's difficult to make a real time stoppage detection methods based on a point of joint for new type of powerful conveyor belt. The paper proposed novel method based on statistics and image processing technology for identifying joints of powerful conveyor belt of new type. First, this algorithm regards every line of the image as the overall sample. Then it calculates the degree of smoothness of the pixel according to each line's mean value and every pixel' variance of the line. On the basis of threshold of smoothness to find point of joint area can form a rectangle zone to apply erosion and dilating. For recognition joint, we can use two features that variance in this rectangle zone and width-height ratio for this rectangle zone. Experimental results show that this algorithm can better identify joints of powerful conveyor belt.