In this project, I built a computer vision pipeline for detecting lane lines and creating averaged and extrapolated boundary lines. The pipeline is as follows:
- Convert frame to grayscale
- Create masks for yellow and white pixels
- Apply a Gaussian smoothing
- Apply a Canny edge detection
- Create an additional mask to focus on the "region of interest" in front of the vehicle
- Convert the points(i.e. pixels) in XY space to a line in Hough space
- Where the lines in Hough space intersect (i.e. a point) a line exists in XY space
- Using the extrema of the lines generated, create two averaged lines
- Create two averaged lines across frames for a smooth video playback
- Draw the lines to each frame
Raw lines after conversion from Hough space
Averaged lines for smooth playback