An improved motion capture system for multiple wheeled mobile robots based on KCF and GMM

With the vigorous development of multi-agent collaborative technology, autonomous position for multiple wheeled mobile robots is becoming more and more important. Motion capture system is a kind of available high-precision position measurement system, where the objects are visually tracked based on kernelized correlation filter (KCF) and then navigated. However, KCF usually behaves badly in the presence of multi-scale and it cannot correct errors itself during the tracking process. To overcome this disadvantage, this paper proposes an improved motion capture system for multiple wheeled mobile robots based on KCF and Gaussian mixture model (GMM). The GMM is used to re-detect and correct the tracking model in KCF tracking process, which enables the tracking algorithm robust to the scale variation of tracking objects. Experiments show that the proposed method has better long-term tracking performance in the case of scale variation for motion capture system.