GPU-Based Detection of Stopping Vehicles

Tharindu Gamage, Jayathu G. Samarawickrama, A. A. Pasqual


Vehicle surveillance is an area which is having a considerable attention in the area of safety and security. In the context of vehicle surveillance, detecting vehicles that come to a stop is considered as one of the major requirements. In this work we address the problem of detecting stopping vehicles. Motion Templates, Optical Flow and Speeded-Up Robust Features (SURF) are used in the detection algorithm. The stopping vehicle detection algorithm effectively implemented using Graphics Processing Units (GPUs) with its power of using thousands of parallel threads. We obtained 96% accuracy level in detecting the vehicles those come to a stop and an improvement of average execution time of a frame from 0.11s to 0.087s using GPUs.

Citation Info :

Online only - International Conference on Advances in ICT for Emerging Regions (ICTer), Colombo, Sri Lanka, Dec 13-14, 2012.

Subscribe to ICTer News