The application of corrected three-frame difference in vehicle tracking

This paper describes a corrected tracking algorithm which improves the precision and accuracy of Camshift algorithm on tracking vehicle objects. An improved three-frame difference was combined with Camshift algorithm to recognize the exact region of a moving vehicle. Firstly, in order to correct the error introduced by three-frame difference and get the accurate tracking window automatically, a simplified real-time vehicle template was establish through three-frame difference procedure, and then the data of the tracking window was used by Camshift algorithm to track the vehicle object. Our algorithm eliminates the spatial redundancy and time redundancy of inter-frame difference and three-frame difference, and optimizes the precision of vehicle identification as well as the accuracy of vehicle tracking. The algorithm is tested on PC and the data source is from an actual video. Experimental results prove that our algorithm increases vehicle tracking accuracy to 96.15%, compared with 33.33% of inter-frame difference.