LANE DETECTION ON ROADS USING ARTIFICIAL NEURAL NETWORKS

Abstract

For efficient and safe driving, autonomous and advanced driver assistance systems (ADAS) depend upon lane recognition as its basis. Lane edge detection with accuracy is needed for such systems. Model fitting, edge detection, and Hough transformations
are some of the more classical methods of lane recognition in computer vision. Varying illumination, shadows, occlusions, and vanishing lane markings are challenges to their function. An example of a competitor is an artificial neural network (ANN), which can learn features of lane patterns from vast databases with better flexibility and accuracy. To detect lane edges in a neural network-based system effectively, this paper proposes an integrated method consisting of picture preprocessing, deep feature extraction, and ANN-based classification. The introduced model demonstrates its flexibility by performing in a variety of environmental conditions with trained and deployed real-world road datasets. ANNs are superior to classical lane identification systems with greater robustness against road surface changes, lighting, and obstructions.