Displacement fault detection of bearing weight saddle in TFDS based on hough transform and symmetry validation

Bearing weight saddle is a important part in railway freight car. In this paper, a novel method of automatic image recognition for displacement fault detection of bearing weight saddle is proposed. Due to the translation invariance and rotation invariance of wheel and bearing, Hough circle transform is applied to locate the center of wheel which can also determine the center of bearing weight saddle. Inspired by the symmetry of bearing weight saddle with normal position, template matching is performed between the area of bearing weight saddle and the mirror of the sub image of the left part. If the displacement fault occurs, the symmetry of bearing weight saddle is broken which will cause that the center point of matching area is far from the center point of the sub image of the left part. In the case of bearing weight saddle with normal position,the two points almost coincide. This method can ensure no omission in principle. It also greatly reduces false alarm rate and computation time. Test on large number of samples is done to validate this method.