Lane detection in a most-shadowed image with some light

Home / Uncategorized / Lane detection in a most-shadowed image with some light

Question:
Currently I’m working on finding lane lines in an image with OpenCV and C++.

However, I encountered a big problem utilizing a good threshold on the following picture.

original image

I have tried using some thresholding methods below, but none of them worked.

sobel x filter with min threshold 10 and max threshold 60
sobel y filter with min threshold 10 and max threshold 60
R channel in BGR color space with min threshold 30 and max threshold 60
S channel in HLS color space and U channel in YUV color space. No matter what threshold I took in these 2 color spaces, it looks like disaster.

Take BGR-R as example, the lane lines are clear but the bright part is also obvious.

I have tried Histogram equalization to balance white and black part in image before thresholding, but it enhanced the contrast, which makes the result even worse.

I don’t know whether I can categorize this problem as shadow-removal, because most part in the image is covered by shadow.

I wish someone can teach me what pre-process steps I could take or any possible methods which can find the line.

Thanks in advance!


Answer:

Read more

Leave a Reply

Your email address will not be published. Required fields are marked *