A video target tracking method based on particle filter and the features of Affine moment invariants

As the diversity and complexity of targets, which is the key to gain images feature of the well descriptive and identifiability in image identification and target tracking problems. In this paper, adopting Hu moment invariants, Zernike moment invariants and Affine moment invariants respectively extract the moment eigenvalues in different deformation pattern of aircraft images, analyzing and comparing with stability and anti-noise, the experiment proved that the affine moment invariants has optimality. In order to reduce the calculation burden and improve velocity, a feature extraction method of affine moment invariants is put forward based on target edge. By constructing the framework of particle filter for tracking, which can realized with moving targets tracking based on the affine moment invariants features. Simulation experiments indicate that it can realize steady targets tracking and has a well tracking accuracy in targets existing rotation, zoom or scale visual angle change.