Abrupt motion tracking of Ochotona curzoniae via improved motion model based on particle filter

Abstract

Abstract The movement of Ochotona curzoniae is random, unpredictable, and often accompany by abrupt motion, Ochotona curzoniae has protective coloration, and possesses the characteristic of having low contrast with its typical background. Particle filter algorithm has been widely used to solve tracking problems, they can solve nonlinear and non-Gaussian problems. The motion model of the particle filter tracking algorithm ignores the acceleration factor, the observation values of position could not represent the future position. Due to this problem, this paper introduces acceleration variance of the “current” statistical model to motion model. In addition, this study fuse color feature with gradient feature as an observation model to overcome interference factors like a complex background, and low contrast. The experimental results show that the improved motion model and observation model has a better performance for abrupt motion, and could accurately track Ochotona curzoniae.