Fusion Technology Of Intertial And Vision Sensors In Track Inspection System

The long wave irregularity of track is caused by the track structure or inhomogene-ous distribution of the elastic under the rail foundation. It affects the comfortaible of train,even worse, it will affects the safety of the train. With the rapid growth of rail mileage,researching a new high-speed non-contact detection of track irregularity methods andtechniques is an urgent task in science and engineering.Nowadays, the coordinate system which is obtained by the dynamic detectionmethod of the long wave irregularity of track is obtained by intergrating the data of theIMU. The dynamic detection method of the long wave irregularity of track is theoreticalbasis to how to get the long wave irregularity of track, but the long wave irregularity oftrack obtained by this method can’t express the absolute position of the track distribution.The key to express the long wave irregularity of track in absolute coordinate system is toget the relationship between relative coordinate and absolute coordinate system. In thispaper, we get the conversion relation between those two sensors. We use step-solvingmethod to calibrate the relative position and orientation between inertial and visual sensors,the main contents including:(1) The relative orientation calibration between inertial and visual sensors. In theprocess of calibration, do some research on vision sensor calibration, obtain corners ofcalibration bar and how to get the vertical vector; in particular, do some reaearch on how toget the relative orientation between inertial and visual sensors.(2) Solving the relative position between the inertial and visual sensors. The relativeposition between the inertial and vision sensors is seemed as the state vector. And then,establish the state equation and measurement equation about the state vector. We can getthe relative position between the inertial and vision sensors by sloving those equation.Simulation results show that the method can accurately calibrate the relative position andorientation of the inertial and visual sensors, the method is simple and easy to implement. (3) The experiment and simulation of the relative pose between inertial and visualsensors. Using Matlab to solve the simulation system. Simulation results show that themethod can accurate calibration of the relative position and orientation between inertialand visual sensors. And the experiment show that this method is capable of calibrating therelative position and orientation between inertial and visual sensors.