An Efficient Object and Railway Track Recognition in Thermal Images Using Deep Learning

Abstract

AbstractTo avoid rail accidents, an efficient railway safety system is essential. The collision between the trains and obstacles on the railway track is one of the main reasons for accidents that result in terms reduced safety and higher financial burdens. Researchers are ceaselessly working to enhance railway safety for curtailing the accident rates. In this paper, a novel technique is discussed for recognizing the objects (obstacles) on the railway track in front of moving train. This methodology presents identification of the railway track along with deep network-based technique for recognition of obstacles on the railway track. The deep network gives the model understanding of real-world objects and enables obstacle recognition. The thermal image data is used for the training and validation of deep network. In this work, Faster R-CNN is utilized to effectively recognize obstacles on the railway tracks. This work can be an incredible assistance for railway to curtail mishaps and monetary burdens. The results demonstrate that the proposed work assists to boost railway safety with good accuracy.KeywordsThermal imagingRailway track detectionHough transformHSV color space segmentationFaster R-CNN