A Novel Method for Resolving Vehicle Occlusion in a Monocular Traffic-Image Sequence

Abstract

This paper presents a novel method for resolving the occlusion of vehicles seen in a sequence of traffic images taken from a single roadside mounted camera. Its concept is built upon a previously proposed vehicle-segmentation method, which is able to extract the vehicle shape out of the background accurately without the effect of shadows and other visual artifacts. Based on the segmented shape and that the shape can be represented by a simple cubical model, we propose a two-step method: first, detect the curvature of the shape contour to generate a data set of the vehicles occluded and, second, decompose it into individual vehicle models using a vanishing point in three dimensions and the set of curvature points of the composite model. The proposed method has been tested on a number of monocular traffic-image sequences and found that it detects the presence of occlusion correctly and resolves most of the occlusion cases involving two vehicles. It only fails when the occlusion was very severe. Further analysis of vehicle dimension also shows that the average estimation accuracy for vehicle width, length, and height are 94.78%, 94.09%, and 95.44%, respectively.

Color-based Texture Image Segmentation for Vehicle Detection

“…This method requires very low-resolution images to work correctly. Earlier research efforts for solving vehicle occlusions on monocular traffic-image sequences (Pang et al (2004)) and for detecting and tracking vehicles by segmenting scenes with severe occlusions (Gentile et al (2004)) have been published. However, based on their experimental results, Zhang et al (2008) stated that "quantitative evaluation and comparison demonstrate that the proposed method outperforms state-of-the-art methods".…”

Range and Size Estimation Based on a Coordinate Transformation Model for Driving Assistance Systems

“…Let the projective height of P 1 P 2 be h i12 (v a ) when P 1 projects onto v a , and the height be h i12 (v b ) when projecting onto v b . Then, by substituting the obtained h i12 (v a ) and h i12 (v b ) into (22), C 1 and H 1 can be obtained as expressed in (23), (24).…”

“…Lin et al [23] computed the number of people in crowded scenes by detecting features of human heads. Pang et al [24] analyzed vehicle projections with geometry and divided their occlusions in the images to provide essential information to the traffic surveillance systems. Broggi et al [25] utilized inverse perspective mapping to transfer images of the front driving lanes into a bird's view of parallel lanes to detect and identify vehicles with a bounding box.…”

A model-based vehicle segmentation method for tracking

“…However, they all require a specific detection zone (or region of interest) where vehicles must be identified separately before occlusion happens. Pang et al [13] work on foreground image to detect occlusion and separate merged vehicles; however, their method is sensitive to foreground noise, and can't handle change of vehicle orientation.…”