A Comparison of Hough Transform and Deep Neural Network Methods on Road Segmentation

Thanks to developments in the computer hardware systems, deep learning has been an attractive field for many researchers in different disciplines. Aim of deep learning is to extract the desired features of raw data as a learning method by operating many hidden layers. Accomplished results of learning methods on complex issues as face recognition, object detection, motion recognition etc. led researchers to think about applying deep learning methods to road lane detection-segmentation which is one of the very important issues of Advanced Driver Assistance Systems (ADAS). Considering main limitations of conventional methods for lane detection, deep learning approach can provide more robustness than existing approaches. The objective of work is to compare the effectiveness of conventional and deep learning applications to improve accuracy of the road segmentation.