Signal Processing 167 (2020) 107278 Contents lists available at ScienceDirect Signal Processing journal homepage: www.elsevier.com/locate/sigpro Cooperative sensor fusion in centralized sensor networks using Cauchy–Schwarz divergence Amirali K. Gostar a,∗ , Tharindu Rathnayake a , Ruwan Tennakoon a , Alireza Bab-Hadiashar a , Giorgio Battistelli b , Luigi Chisci b , Reza Hoseinnezhad a a School of Engineering, RMIT University, Victoria, Australia b Department of Information Engineering, University of Florence, Florence, Italy a r t i c l e i n f o Article history: Received 18 February 2019 Revised 29 August 2019 Accepted 3 September 2019 Available online 3 September 2019 Keywords: Random finite sets Multi-target tracking LMB filter Probability density function a b s t r a c t This paper presents a new solution for statistical fusion of multi-sensor information acquired from dif- ferent fields of view, in a centralized sensor network. The focus is on applications that involve tracking unknown number of objects with time-varying states. Our solution is a track-to-track fusion method in which the information contents of posteriors are combined. Existing information-theoretic solutions for track-to-track fusion in sensor networks are commonly devised based on minimizing the average infor- mation divergence from the local posteriors to the fused one. A common approach is to use Generalized Covariance Intersection rule for sensor fusion. This approach works best when all the sensors detect the same object(s), and performs poorly when fields-of-view are different. We suggest Cauchy–Schwarz diver- gence to be used for measuring information divergence. We demonstrate that employing Cauchy–Schwarz divergence leads to fusion rules that are generally more tolerant to imperfect consensus. We show that the proposed fusion rule for multiple Poisson posteriors is the weighted arithmetic mean of the Pois- son densities. Furthermore, we derive the fusion rule for labeled multi Bernoulli filter by approximating the labeled multi Bernoulli density to its first order moment. Numerical experiments show the superior performance of our solution compared to Kullback–Leibler averaging method. © 2019 Elsevier B.V. All rights reserved. 1. Introduction Statistical information fusion is a fundamental component of sensor networks and their applications, especially when the ap- plication involves tracking of the existence and states of randomly varying number of objects. Such applications are widespread, rang- ing from multi-target tracking in defense to civilian surveillance, and the tracking task is usually performed by a multi-object stochastic filter [1–8]. For wireless sensor networks that are composed of a finite set of sensor devices geographically distributed in a large environ- ment, multi-sensor fusion solutions are expected to return tracking outcomes that cover the whole environment. In such applications, the sensors usually cover different areas (probably with overlaps) and their information complement each other to collectively cover a large area. ∗ Corresponding author. E-mail addresses: amirali.khodadadian@rmit.edu.au (A.K. Gostar), tharindu.rathnayake@rmit.edu.au (T. Rathnayake), ruwan.tennakoon@rmit.edu.au (R. Tennakoon), abh@rmit.edu.au (A. Bab-Hadiashar), giorgio.battistelli@unifi.it (G. Battistelli), luigi.chisci@unifi.it (L. Chisci), rezah@rmit.edu.au (R. Hoseinnezhad). In terms of the formation of sensor nodes and the way they communicate, sensor networks are divided into two general types: distributed and centralized. Distributed sensor networks benefit from their scalability, ease of integrity with respect to other services, and reduced resource requirements for information management. On the other hand, centralized schemes allow more efficient energy management, simplified network coverage analy- sis, and better application design (placement of nodes, application awareness, and so on) [9]. This paper concentrates on multi-object tracking in a sensor network with two particular characteristics: (i) The sensor network is centralized, and (ii) each node is equipped with some local pro- cessing power, enough to compute the multi-object statistics based on its own observations and communicate those to the central node. With recent advances in affordable energy-efficient processing modules, sensing platforms are commonly equipped with local processors capable of running numerically tractable multi-object tracking routines. During the past fifty years, numerous multi- object tracking algorithms have been developed to address the challenges of false alarms, miss-detections and randomly varying number of objects. The methods include variations of Kalman fil- https://doi.org/10.1016/j.sigpro.2019.107278 0165-1684/© 2019 Elsevier B.V. All rights reserved.