Detection and segmentation of occluded vehicles based on symmetry analysis

Abstract

Occluded vehicles detection and segmentation are the critical parts of intelligent transportation system for vehicle tracking or traffic flow analysis. Firstly, difference of Gaussian filter is used to reduce the high-frequency noise with enhancing the vehicle's structural features, followed by extracting the spatial symmetric axes by mirror coding LBP operators. Then, we use the number and location of symmetric axes to detect the vehicle occlusion. At last, the global optimal segmentation is obtained under the constraints of symmetry axes combined with contour concavity analysis. Experimental results demonstrate that by comparing with classical contour based methods, our approach can get higher accuracy on both occluded vehicle detection and segmentation especially in the influence of complex background noise.