Track Area Detection Algorithm for Obstacle Detection of Fully Automatic Driverless Train

Abstract:

A non-contact detection algorithm based on vision sensors is proposed to address the issue of track area segmentation in the context of fully automatic driverless trains for rail transit. The algorithm uses frame difference threshold and grayscale distribution feature extraction methods to perform scene recognition and labeling of video image data. Image preprocessing and edge detection of the track contour are completed by an adaptive edge detection module， which adjusts the parameter input based on the results of image scene recognition. The track area boundary search module consists of two submodules： the sliding pane search submodule and the passband search submodule， based on the sliding pane-like approach， in order to extract the track outline curve. Finally， a Kalman filter is used to improve the accuracy and robustness of the detection results. The experimental results show that the algorithm exhibits strong detection performance on track boundaries.