Steel rail profile outlier detection and effective profile recognition method

Abstract

The invention discloses a steel rail profile outlier detection and effective profile recognition method. In view of a problem that seriously erroneous abrasion calculation is caused by the mismatch between a measured profile and a standard profile due to the outlier and profile diversity in an actual test in a railway line, an outlier detection and fast effective profile recognition algorithm based on profile coarse registration is provided. The algorithm comprises first splitting a profile curve, and eliminating most of sparse outliers in the curve; merging the remaining profile segments andmerging the profile segments belonging to the same region into one segment; performing a regional convexity-concavity test, determining a potential rail-head rail-web area according to continuous convexity-concavity and a maximum point number principle, and realizing the coarse registration of the potential area and the standard profile; finally recognizing the effective profile according to the rail web similarity between the two profiles after registration. On this basis, the method re-registers the original measured profile with the standard profile, and removes the outliers according to the relative distance between the profiles.

Description

technical field [0001] The invention relates to the field of rail transit detection, in particular to a method for outlier point detection and effective contour recognition of rail contours. Background technique [0002] By using the 2D displacement laser sensor to measure the rail profile, it is found that the collected original profile curve is greatly affected by external disturbances. The specific performance is the following two points: First, due to the influence of the reflection of the oil layer on the surface of the rail, the bright area, and the irrelevant areas such as the roadbed and fasteners, there are many outliers in the curve and the jump range is large, especially on the bright tread on the top of the rail. The performance is more obvious on the gable belt. The existence of outliers will affect the positioning of the double-arc area of ​​the rail waist, resulting in contour registration errors; secondly, due to the diversity of railway line shapes, except ...

AI Technical Summary

Problems solved by technology

It can be seen that the statistical distribution model is effective in identifying outliers at the upper and lower ends of the contour, but it is difficult to identify outliers consistent with the longitudinal amplitude of the contour; the distance model is effective in identifying sparsely distributed outliers, but it is difficult to identify outliers Densely distributed outliers; the density model may classify normal contour segments and outliers into one category, and there are more uncertainties in practical applications

None of the above three conventional detection methods can effectively remove outliers.

Images