Deviation monitoring method for conveying belt of belt conveyor based on deep learning

The invention provides a method for monitoring deviation of a conveying belt of a belt conveyor based on deep learning. A universal target detection network is used for detecting linear features of the edge of the conveying belt and effectively judging the deviation state. According to the method, the marking method with specific marking requirements is adopted to carry out data marking on the edge area of the conveying belt to obtain the conveying belt deviation data set, then the data set is utilized to train the universal target detection network, and then the trained network is used for predicting the edge area of the conveying belt. And calculating coordinate positions of four vertexes of the prediction frame to obtain diagonal positions and equations of the prediction frame in the edge area of the conveyor belt, and representing edge straight lines of the conveyor belt according to the diagonal positions and equations. The deviation state of the conveying belt is effectively monitored by comparing the o