ROAD LANE DETECTION USING DECISION TREE

Abstract

Lane detection is an extremely difficult issue to solve. For decades, it has been a focal point of research in computer vision. Computer vision and machine learning algorithms are facing a formidable obstacle in the form of lane detection, which is fundamentally a multi-feature detection issue. Despite the prevalence of machine learning techniques in lane identification, their primary use is in classification, not feature creation. However, successful feature detection tests may be uncovered by using current machine learning algorithms to determine which features are rich in recognition. Unfortunately, lane identification efficiency and accuracy have not been completely optimised using these approaches. A novel approach to its resolution is presented in this work. For pre-processing and ROI selection, we provide a novel approach. In the pre-processing step, we want to apply the HSV colour transformation to extract white features and include preliminary edge feature recognition. Then, we will choose the ROI according to the suggested pre- processing. The lane is detected using this new pre-processing approach. Compared to the current preprocessing and ROI selection methods, the results achieved from evaluating the suggested approach using the standard KITTI road database are better.