抽象的

Applying FAST & FREAK Algorithms in Selected Object Tracking

S.Khachikian, M.Emad

Object tracking is defined as a method applied in tracking and recognizing the state of a moving object selected by the user or found based on a specific feature in different image frames. For this purpose different features of the selected object like: corners, colour, geometric shape, dimensions etc. are extracted accordingly, and in the next frame, the new location of the object is recognized which makes the tracking of the moving path possible. Several methods are proposed among which extracting and using the key points of the image is the one mostly applied. Here, the FAST is applied and the binary descriptors and indexes of each key point found thereof are introduced by FREAK, inspired by human eye, applied in comparing and recognizing the objects in each frame. Based on this the accuracy and speed of the recognition increase with less memory space needed for implementation. The performance evaluation of this newly proposed method is made through the data set introduced by Mikalajczyk and Schmid. The obtained results indicate a 99% precision on images not subjected to transformation.