Object tracking using a new statistical multivariate hotelling's T2 approach
Anuja Kumar Acharya, Biswajit Sahoo, K. P. Swain · 2014
In this paper we proposed a new statistical multivariate method for tracking an object in a video. This method is based on the Hottelling T2test which is designed to provide a global significance test for the difference between two region or two group with simultaneously measured multiple dependent or independent variables. An object to be tracked can be found by comparing its multivariate mean in the successive frame of the video. The T2value give the measurement of the difference of two mean vector. In this approach the object window containing the matrix of intensity value is transformed into a set of feature vector. These set of features is compared using multivariate T2test in the successive frame for the significant matching of the object in its nearest locality. It is observed that higher the T2value more is the chance of mismatching and lower the T2value more is the chance of matching the multi attribute. Simulation result shows that the proposed method is capable of accurately detecting the non rigid, moving object in stationary as well as non stationary camera with noisy and occlusion environment.