Created by [ Rowan Dempster], last modified by [ Anita Hu] on Jan 14, 2020
The main purpose of the road line detection team is to visually and sense reflective road markings on the ground in order for the vehicle to detect crosswalks, stop lines, and lane lines (which need to be classified as uncrossable/crossable). These detections are used to build a map of the vehicle's environment in order for the car to make an informed decision on how it should proceed in the environment, therefore the points we give must be in 3D (in meters, not image pixel units). The two main methods that are currently looked into is through a neural network semantic segmentation approach and CV-based approach. You may also need to write C++ code to implement ROS Kinetic nodes that will run in real-time for your algorithms.
Curve-fitting:
http://www.wseas.us/e-library/conferences/2010/Merida/CIMMACS/CIMMACS-35.pdf
http://vision.stanford.edu/teaching/cs231a_autumn1112/lecture/lecture4_edges_lines_cs231a_marked.pdf
Resources:
https://github.com/daniel-s-ingram/self_driving_cars_specialization/blob/master/3_visual_perception_for_self_driving_cars/Environment_Perception_For_Self-Driving_Cars.ipynb
https://github.com/HsucheChiang/Advanced_Lane_Detection
BEV transform:
https://git.uwaterloo.ca/WATonomous/perception-year-2/tree/develop/roadline_detection
Document generated by Confluence on Dec 10, 2021 04:02