Vehicle make and model recognition using symmetrical SURF

摘要: SURF (Speeded Up Robust Features) is a robust and useful feature detector for various vision-based applications but lacks the ability to detect symmetrical objects. This paper proposes new descriptor enrich power of all possible matching pairs through mirroring transformation. A vehicle make-and-model recognition (MMR) application then adopted prove practicability feasibility method. To vehicles from road, proposed first applied determine ROI each road without using any motion features. scheme provides two advantages; there no need background subtraction it extremely efficient real-time applications. Two MMR challenges, i.e., multiplicity ambiguity problems, are addressed. The problem stems one model often having different shapes on road. results companies sharing similar shapes. address these grid division separate into several grids; weak classifiers that trained grids integrated build strong ensemble classifier. Because rich representation grid-based method high accuracy detection, classifier can accurately recognize vehicle.