Advanced_Lane_Line_Detection

For detection of lane lines first we calibrate the camera

The camera is calibrated using chessboard images taken from the same camera present in the Camera_CAL directory.Camera calibration is done to find the distortions in the image, the image is then undistorted after finding the distortion matrix.

• First the image is converted to gray scale
• After gray scale conversion the dsitrotion matrices are found using calibrateCameara() function
• The undistortion matrix found using the calibrateCamera function is used to undistort the frames

The Lane Detection Pipeline is as follows:

• Camera calibration is done to find the distortion matrix
• The frame is read from the recorded video
• The frame is then undistorted
• The frames are now blurred to soften up the edges
• Now colour grading and binary thresholding is done for extraction of data using different colour spaces like HLS,grayscale and then edge detection using sobel x and sobel y , hence using bitwise or and and merging the binary data
• Now a specific region of interest is defined on the frame in terms of pixel values(x,y)//predefined
• The data inside the region of interest is only considered for lane line detection
• Now the region of interest is converted into birds eye view and flipped which helps us find the curvature of road and also makes the binary data incoming
• Now within the region of interest we have the binary values of the edges, the distance between two lane markers is know to use and is constant also, the markers should contiue incoming following a pattern
• Hence, considering the previous identifiers we use the sliding windows algorithm to continue the flow of lane lines in our region of interest, this helps us in finding both the lane lines on the road and marking them continuously.
• The sliding windows are defined in a pixel region of a specific width with our search going from the center of the lane lines to identify the lane lines
• Here we use a histogram, a polynomial function is formed by this detection
• Using the polynomial we find the radius of curvature and the lane offset from the center
• After all of this is done the detected region is masked on the original frame and on the frame the curvature and center offset data is added
• Now the frame is displayed Output of the code in reallife dataset with real situations on road: https://drive.google.com/file/d/1cnCpZcO7MLx6u-gNxVcD2jFN99WGMtvb/view?usp=sharing