Coupler tracking algorithm for a wheeled train uncoupling robot based on gaussian particle filter combined with improved particle swarm optimization

The uncoupling work in the station is continuing. It is sometimes difficult to ensure the working efficiency when it is completed by people or when there is a slight vibration, which results in the fault of uncoupling. The wheeled train uncoupling robot with four degrees-of-freedom has been developed to replace human to uncouple the freight cars in the marshalling yard. To finish the uncoupling operation, the robot needs to track the coupler. The chief challenge of this visual tacking can be attributed to the difficulty in handling the appearance variability of the target coupler, both the coupler and the background, which change over time. The intrinsic appearance variability includes position variation, shape and size variations, due to different freight trains, while the extrinsic lighting condition and weather inevitably cause large appearance variation. The improved particle swarm optimization (IPSO) and the Gaussian particle filter are applied to develop a visual tracking method for freight train couplers. The model representation combining eigenspace and compressive sampling can reflect the appearance variation of the target, thereby facilitating the visual tracking task. The experiments demonstrate that the proposed method can track the freight train coupler accurately and efficiently.