The current implementation takes the lines from the HoughLinesP and grades the into left (positive slope) and right (negative slope). This causes some errors with sharp turns. Keeping with an OpenCV pipeline for the time being try to fix cornering issues. (RL and imitation learning could probably handle this easily but it would be good to explore the OpenCV space first).
The lane lines in this case are : [[[-14, 100, 98, 50]], [[246, 100, -157, 50]]]
The calculated steering angle: 25
The current implementation takes the lines from the
HoughLinesP
and grades the into left (positive slope) and right (negative slope). This causes some errors with sharp turns. Keeping with an OpenCV pipeline for the time being try to fix cornering issues. (RL and imitation learning could probably handle this easily but it would be good to explore the OpenCV space first).The lane lines in this case are : [[[-14, 100, 98, 50]], [[246, 100, -157, 50]]] The calculated steering angle: 25
Doc for research: