A statistical approach to circle tracking

By applying an edge detection algorithm the silhouettes of objects can be efficiently detected in an image or tracked through multiple frames. Traditional methods of object detection ignore information about an object being tracked. This information can be used to reduce processing time and increase accuracy of object detection. This paper proposes a method that uses this information to provide a tracking algorithms for circles in images with low processing time. It does this by creating a probability distribution function which it integrates to calculate an estimated object position. The processing time and accuracy of the algorithm is then tested against comparable methods, such as the Randomised Hough Transform. For the parameters given it is more accurate than the Randomised Hough Transform at about a quarter the processing time.