2019/02/06 by Kaveh Fathian, Kasra Khosoussi, Fathian, Kaveh +7 · 1 citation
Computer Science · Social Sciences · #Advanced Computing and Algorithms #Computer Vision and Pattern Recognition (cs.CV) #FOS: Computer and information sciences #Face and Expression Recognition #Multiagent Systems (cs.MA) #Robotics (cs.RO) #Video Surveillance and Tracking Methods
paper · pdf · doi:10.48550/arxiv.1902.02256
openalex publication_date 2019/02/06 · openalex created_date 2022/07/29 · openalex updated_date 2026/08/01
Many robotics applications require alignment and fusion of observations\nobtained at multiple views to form a global model of the environment. Multi-way\ndata association methods provide a mechanism to improve alignment accuracy of\npairwise associations and ensure their consistency. However, existing methods\nthat solve this computationally challenging problem are often too slow for\nreal-time applications. Furthermore, some of the existing techniques can\nviolate the cycle consistency principle, thus drastically reducing the fusion\naccuracy. This work presents the CLEAR (Consistent Lifting, Embedding, and\nAlignment Rectification) algorithm to address these issues. By leveraging\ninsights from the multi-way matching and spectral graph clustering literature,\nCLEAR provides cycle consistent and accurate solutions in a computationally\nefficient manner. Numerical experiments on both synthetic and real datasets are\ncarried out to demonstrate the scalability and superior performance of our\nalgorithm in real-world problems. This algorithmic framework can provide\nsignificant improvement in the accuracy and efficiency of existing discrete\nassignment problems, which traditionally use pairwise (but potentially\ninconsistent) correspondences. An implementation of CLEAR is made publicly\navailable online.\n