This research includes two parts: (1) background image generation for vehicle detection, (2) vehicle 3-dimensional (3D) shape recovery and vehicle tracking. In the first part, "background subtraction" approach is used to detect vehicles in the images. The problem of background image generation is modeled as a mixture of Gaussian distributions and our goal is to separate the background data from other components in the image. A median model is presented as the background image generation method. The second part proposes a size-difference method to recover the vehicle 3D parameters. Vehicle tracking at typical street blocks and intersections is done based on the combination of vehicle features, such as the 3D parameters and pixel intensity statistics.
Robust background image generation and vehicle 3D detection and tracking
2004-01-01
449303 byte
Conference paper
Electronic Resource
English
Robust background tracking for infrared image sequences
IET Digital Library Archive | 2006
|Robust Tracking Using Foreground-Background Texture Discrimination
British Library Online Contents | 2006
|Robust object tracking with background-weighted local kernels
British Library Online Contents | 2008
|Robust Vehicle Detection and Tracking for Traffic Surveillance
British Library Conference Proceedings | 1999
|