I would say to try to detect the road with a linear Hough transform, then extract the pixels that belong to the road. If it's done, you're halfway there.
Then, do a segmentation. You can use k-means or an adaptative thresholding to segment the "things" on the road. Do some erosion/dilatation to get rid of the noise. Now, you should have the silhouettes of the vehicles.
Finally, use a Kalman filter to track the movement of the objects.
Hope this helps.
Asked: 2016-01-26 00:54:18 -0600
Seen: 330 times
Last updated: Jan 26 '16
B&W Image brightness- using a Trackbar
Fish detection. Is haartraining a good approach ?
Area of a single pixel object in OpenCV
build problems for android_binary_package - Eclipse Indigo, Ubuntu 12.04
OpenCV DescriptorMatcher matches
Which values for window size and number of pyramids are reasonable for calcOpticalFlowPyrLK?
Can't compile .cu file when including opencv.hpp
Using OpenCV's stitching module, strange error when compositing images
Please, put in some effort. We are not a
i need a solution, give me now
kind of forum. Take a look at the FAQ to fully understand my reaction ... This seems again like a school assignment and a student picking the shortest way to a solution...