search

kundralaci / JAMResearch

0 stars 0 forks source link

Vehicle Detection and Tracking Using the Block Matching Algorithm #63

Closed scorpeeon closed 10 years ago

scorpeeon commented 10 years ago

VehicleDetectionAndTrackingUsingTheBlockMatchingAlgorithm.pdf

scorpeeon commented 10 years ago

Rather old, but interesting paper. Detects and tracks vehicles using the Block Matching Algorithm which is the motion estimation algorithm employed in the MPEG compression standard. BMA has fixed block sizes (6x6 pixel, 8x8 pixel, etc), this makes it efficient, but might decrease precision. BMA is very effective in discriminating between close objects (partially occluding vehicles, crowded traffic scenes).

Due to noise, BMA generates many wrong vectors over static blocks located on the background of the road lane. Because of this, some filtering, cleaning, smoothing is needed.

Maybe we could use this to help the tracking, for separating close objects.