Gauge Point Detection Algorithm Research In Track Geometry Deformation Detection

The gauge point refers to the top of rail vertical downward position 16 mm. In the track geometry deformation detection, whether can positiongauge point accurately affects the measurement results of track gauge directly. The gauge value is an important parameters to judge the geometry state of the track. If gauge value exceeds the critical value will increase lateral thrust between the locomotive wheel flange and rail, causing and aggravating rail flange side abrasion, and changing relations between the wheel and rail, resulting the life of track and rolling stock greatly reduced; What is more serious is that the locomotive leave off orbit or stuck in orbit. Therefore, it is vital to carry out research on detecting rail gauge point accurately in track geometry deformation detection.In recent years, with the rapid development of high-speed railway, passenger traffic and freight traffic are increasing significantly. The quality requirements of the railway is also rising. By learning track geometry deformation detection methods both at home and abroad, decide to get the track information by using the machine vision technology to detect the rail gauge point precisely.The gauge point detection algorithm has been discussed in two different conditions,those are ideal environment and actual environment. Under ideal condition which can obtain the complete rail profile laser information detect the gauge point. Under practice condition, because of the track geometry deformation and other external factors lead to the track inspection device vibration and deflection, which resulting the collected rail profile laser information exist a few defects. Under practice condition, there are two kinds of defects, top of rail laser information defect and rail head laser information defect. Track point positioning algorithm has been discussed respectively based on the above two kinds of light defect.Firstly, the track geometry deformation detection system based on machine vision has been discussed and its detection principle and system are expounded, At the same time the laboratory gauge points detection device has been designed. Secondly. gauge point has been detected by solving the fitting curve and gauge point ordinate constant linear intersection,which will enhance the gauge point positioning accuracy than the traditional pixel-level positioning accuracy to the sub-pixel level. Thirdly, study gauge point positioning under practice condition, if the top of the laser light defected, use curve fitting to repair laser information and complete the gauge point positioning, to some extent, improving the efficiency of the collection data and equipment; If the rail head laser light defected, using Elman neural network do rolling forecasts to repair defects and realize the gauge point positioning, also improving the efficiency of the collection data and equipment. Finally, analyze the influence factors of the accuracy of the gauge point positioning and do the experimental verification. Concluded that the locomotives body vibration has no effect on the gauge positioning accuracy, but the deflection of the body have a certain bad effect on gauge point positioning and need to be amended. Compared to traditional pixel-level gauge point positioning, the article gauge point positioning algorithm to achieve a sub-pixel positioning, which greatly improves the positioning accuracy.