Measuring particle kinematics using template matching algorithms

Kurzfassung

To describe the kinematics of many particle systems one needs to know the particle positions and velocities at all times. If the time evolution of the system is recorded by an optical device, the positions and velocities have to be estimated out of optical images. Template matching algorithms represent a viable tool to locate particles in optical images. A crucial factor on the performance of this method is the choice of the similarity measure. Recently it was shown that particle detection via template matching, using the correlation coefficient (CC) as the similarity measure, leads to good results in the presence of strong signal independent noise. Here, we extend the noise model by a signal dependent part, which is present in many imaging scenarios. We further introduce the mutual information (MI) as a nonlinear similarity measure and compare the performance of the MI and the CC under varying noise strengths and types. It turns out that the mutual information leads to superior results in the case of signal dependent noise. Besides the positions also the velocities of the particles have to be estimated. This usually involves the problem of identifying the same particle in consecutive images. We propose an alternative approach applicable to imaging scenarios where the particles appear elongated in the image due to their velocity. We design an anisotropic template to fit the elongation of the particles and estimate their velocity and direction of motion. With this method we propose a new technique to simultaneously estimate not only the position, but also the velocity and direction of motion of a particle out of a single image.