In this paper, we explore the potential gains in using Sequential Monte Carlo (SMC) methods for extended target tracking (ETT) models based on Gaussian processes (GP). The existing random hypersurface based ETT models use Extended/Unscented Kalman filter for inference, which may lead to poor performance under high uncertainty. Particle filters (PFs) are known to provide robust performance in the cases where the non-linear Kalman filtering solutions fail. We design a Rao-Blackwellised particle filter (RBPF) where we exploit the conditional linear Gaussian structure of the GP parameters. We illustrate the gain in the performance with simulations.