Lane Marking Detection and Classification for Localization

Abstract

Our project aims to be able to create a bounding box around lane markings, such as crosswalks and lane dividers, and then classify them into their corresponding categories. To achieve this we use a subset of AI called deep neural networks where we train a model based on gathered data to be able to classify the markings. To classify an item we must first find them, to do this we utilize basic rules and coding packages from OpenCV (open computer vision). Once this is done we ensure that the output of our model is readable and understandable to be overlaid on the original image.