abstract |
The invention provides a multi-source sensor fusion method and system for detecting and measuring rail surface defects, including: using a rail surface three-dimensional data acquisition system to collect rail three-dimensional data; For the severity of surface defects, establish a rail surface defect data set, which is divided into training set, validation set and test set; design the backbone network, neck module and head module, build a three-dimensional rail surface defect detection and measurement network, and build a network based on The rail surface defect detection and measurement network model based on multi-source sensor fusion uses the training set and validation set of the rail surface defect data set to train the network model, and uses the test set to test the detection performance index of the network model, and determines the trained multi-source sensor fusion based network model. rail surface defect detection and measurement network model; input the 3D data of the rail surface collected in the real line into the network model to determine the rail surface defect detection results. |