def exploratory_pipeline():
'''
The pipeline to calculate the lane lines in an exploratory mode that produces
plots of the various stages along the way.
Notes:
- Calibration shall be performed at least once before using the pipeline
- This pipeline works with for a single image
'''
# Read in an image
image = mpimg.imread(test_image)
# Undistort the image before processing and plot an example
mtx, dist = Camera.getCalibrationData()
undist = Camera.undistort(image, mtx, dist)
# Process the undistorted image with thresholding to get a binary mask
binary_mask = Thresholding.lane_detection_mask(undist, KERNEL)
# Plot the undistorted and binary mask images
Plotting.plotResult(undist, binary_mask)
# Perspective transform
warped, M, M_inv = Image_processing.perspectiveTransform(binary_mask)
Plotting.plotResult(binary_mask, warped)
# Get the position of the left and right lanes
leftx_base, rightx_base = Image_processing.histogramPeaks(warped,
plot=True)
# Get the lane line equations using the sliding window
left_fit, right_fit = Image_processing.slidingWindowInit(warped, plot=True)
# Calculate the sliding window of a successive image, using the lane line
# equations that are already known from the previous image analysis
left_fit, right_fit = Image_processing.slidingWindowFollowing(warped,
left_fit,
right_fit,
plot=True)
# Calculate the curvature of the left and right lanes
left_curv, right_curv, curv_string = Image_processing.curvature(
warped, left_fit, right_fit)
# Calculate the offset of the car from the center of the lane line
offset_m, offset_string = Image_processing.offsetFromCenter(
undist, leftx_base, rightx_base)
# Overlay the lane area to the undistorted image
Image_processing.laneArea(warped,
undist,
M_inv,
left_fit,
right_fit,
offset_string,
plot=True)
def pipeline(image):
'''
The lane detection pipeline
Notes:
- Calibration shall be performed at least once before using the pipeline
- This pipeline worlks with a video
'''
global left_lane
global right_lane
global initiate_sw
global initiate_sw_counter
global initiate_sw_limit
# Undistort the image before processing and plot an example
mtx, dist = Camera.getCalibrationData()
undist = Camera.undistort(image, mtx, dist)
# Process the undistorted image with thresholding to get a binary mask
binary_mask = Thresholding.lane_detection_mask(undist, KERNEL)
# Perspective transform
warped, M, M_inv = Image_processing.perspectiveTransform(binary_mask)
# Get the position of the left and right lanes
leftx_base, rightx_base = Image_processing.histogramPeaks(warped)
# Apply the appropriate slidinig window technique
# Note:
# - If there are more than a few bad frames, a fresh sliding window full
# search is forced to take place
if initiate_sw:
# Get the lane line equations using the sliding window
left_fit, right_fit = Image_processing.slidingWindowInit(warped)
# Reset the sliding window control variables
initiate_sw = False
initiate_sw_counter = 0
else:
# Calculate the sliding window of a successive image, using the lane line
# equations that are already known from the previous image analysis
left_fit, right_fit = Image_processing.slidingWindowFollowing(warped, \
left_lane.current_fit, \
right_lane.current_fit)
# Update the sliding window control variables
initiate_sw_counter = initiate_sw_counter + 1
if initiate_sw_counter >= 5:
initiate_sw = True
# Sanity checks and update globals accordingly
if Image_processing.sanityChecks(warped, left_fit, right_fit):
# If sanity ok then update the globals with the new line equations
left_lane.current_fit = left_fit
right_lane.current_fit = right_fit
else:
# If sanity fails use the previously set globals for calculations
left_fit = left_lane.current_fit
right_fit = right_lane.current_fit
# And start a fresh sliding window search
initiate_sw = True
# Calculate the curvature of the left and right lanes
left_curv, right_curv, curv_string = Image_processing.curvature(
warped, left_fit, right_fit)
# Calculate the offset of the car from the center of the lane line
offset_m, offset_string = Image_processing.offsetFromCenter(
undist, leftx_base, rightx_base)
# Overlay the lane area to the undistorted image
result = Image_processing.laneArea(warped, undist, M_inv, left_fit,
right_fit, offset_string)
# Return processed image
return result