This program detects and classifies lanes and Determines which direction the car has to turn
Approach: Masking, HSL color filtering and Canny we first mask the given image to filter out noise as similar to problem 2, we then further filter the image using cvtColor to convert the image from BGR to HSL which allows us to identify the yellow and white lanes for better lane detection.
Warping image we warp to a birds eye view using getPerspectiveTransform (to find the homography matrix) we do this to allow detection of direction to be easier as we can determine direction based on the max and min points present in the warped image.
Sliding window as we now have a warped image of the lanes and the edge points we can start constructing the curve which lets us calculate the direction of motion and the radius of curvature. By checking for adjacent area close to the edge points detected we now map the curve far better than other conventional image recognition methods.
We then can construct a polygon using the extracted points which is later used for superimposing the original image with the detected path.
Finally we can combine all previous mentioned images to understand the working of this pipeline better.
📦Lane-detection
┣ 📂media
┃ ┣ 📜problem_2.mp4
┃ ┗ 📜problem_3.mp4
┣ 📂output
┃ ┣ 📜problem_2_output.avi
┃ ┣ 📜problem_3_output.avi
┃ ┗ 📜problem_3_output.gif
┣ 📜README.md
┣ 📜histogram_equalization.py
┣ 📜straight_lines.py
┗ 📜turn_prediction.py