Steel rail damage B-display image recognition method based on deep convolutional neural network

• Main Authors: ZENG CHUQI, WANG PING, FU BIN, WANG QIHANG, YANG KANGHUA, YAO JIDONG, WANG XIAOMING, CHEN ZHENGXING, LIU YONG, HE QING
• Format: Patent
• Language: Chinese; English
• Published: 08.01.2021
• Subjects: CALCULATING; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS; COMPUTING; COUNTING; HANDLING RECORD CARRIERS; IMAGE DATA PROCESSING OR GENERATION, IN GENERAL; INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIRCHEMICAL OR PHYSICAL PROPERTIES; MEASURING; PHYSICS; PRESENTATION OF DATA; RECOGNITION OF DATA; RECORD CARRIERS; TESTING

The invention relates to the technical field of steel rail flaw detection, in particular to a steel rail flaw B-display image recognition method based on a deep convolutional neural network, which comprises the following steps of: 1, searching and storing four types of wave generation conditions of a weld joint, a normal screw hole, an abnormal screw hole and a surface flaw; 2, filling an originalpicture, and expanding the size of the picture; 3, cutting the filling data; 4, conducting transverse and symmetric cutting to form two parts up and down along the symmetry axis of the left and rightsteel rails, conducting average cutting to form four parts in the longitudinal direction, and finally obtaining 416*416 standard pictures; 5, conducting labeling operation, and making the label filesand the picture files into a standard data set; 6,constructing a YOLO-UAV+min model; 7, performing clustering analysis on the data set to obtain a priori box, and inputting the priori box into the model; 8, configuring network