Autors: Draganov, I. R., Mironov, R. P.
Title: Moving Objects Detection in Video by Various Background Modelling Algorithms and Score Fusion
Keywords: Background subtraction; Low rank recovery; Matrix completion

Abstract: The paper presents results from testing ten of the fastest background modelling algorithms applied for detecting moving objects in video. The algorithms are Fast Principal Component Pursuit (Fast PCP), Grassmann Average (GA), Grassmann Median (GM), Go Decomposition (GoDec), Greedy Semi-Soft Go Decomposition (GreGoDec), Low-Rank Matrix Completion by Riemannian Optimization (LRGeomCG), Robust Orthonormal Subspace Learning (ROSL), Non-Negative Matrix Factorization via Nesterovs Optimal Gradient Method (NeNMF), Deep Semi Non-negative Matrix Factorization (Deep-Semi-NMF) and Tucker Decomposition by Alternating Least Squares (Tucker-ALS). Two new algorithms employing score fusion from Fast PCP and ROSL, which yielded alone the highest Detection Rate, Precision and F-measure, are proposed. The first algorithm has higher Detection Rate from all the others and the second—the highest Precision. Both are considered applicable in various practical scenarios when seeking either higher reliability

References

    Issue

    Smart Innovation, Systems and Technologies, vol. 309, pp. 347 - 359, 2022, Singapore, Springer Nature, DOI 10.1007/978-981-19-3444-5_30, ISBN:978-981193443-8, ISSN:21903018,

    Copyright Springer Nature

    Вид: публикация в международен форум, публикация в издание с импакт фактор, публикация в реферирано издание, индексирана в Scopus