Modelling Vague Shape Dynamic Phenomena from Sensor Network data using a Decentralized Fuzzy Rule-Based Approach

Abstract

GIScience 2016 Short Paper Proceedings Modelling Vague Shape Dynamic Phenomena from Sensor Network data using a Decentralized Fuzzy Rule-Based Approach Roger Cesarie Ntankouo Njila 1 , Mir Abolfazl Mostafavi 1 , Jean Brodeur 1 Centre de Recherche en Geomatique, 2314 Pavillon Casault; Universite Laval, Quebec, Canada, G1K 7P4 Email: roger-cesarie.ntankouo-njila.1@ulaval.ca, Mir-Abolfazl.Mostafavi@scg.ulaval.ca, recherchegeosemantic@videotron.ca Abstract Modelling dynamic phenomena of vague shape from sensor data is still a challenging problem for many applications. In this paper, we propose a decentralized fuzzy rule-based approach based on fuzzy object model to build a more realistic spatiotemporal representation for such phenomena. This approach has been successfully implemented in a simulation case of bushfire monitoring, showing advantages for spatial decision making in a disaster management context. Keywords: sensors, sensor data, fuzzy objects, disaster management, spatial decision support 1. Introduction Extracting geospatial information from geosensor data can help to better understand a complex phenomenon for real time decision making process (Sadeq et al. 2013). Several approaches are used for the extraction of geospatial information from sensor data. Many of these approaches are developed based on the assumption that monitored phenomena are of crisp shape with well- defined boundaries. However, many dynamic phenomena have vague spatial boundaries, and their accurate detection and extraction from sensor data is a challenging problem. In this paper we propose a decentralized fuzzy rule-based approach to address this problem. In the proposed method, sensors detect vague shape phenomena using a fuzzy logic reasoning approach and collaborate with their neighboring sensors to infer vague spatial extent of the phenomena and its dynamics. We adopt Crisp-Fuzzy objects (Pauly and Schneider 2008) as a more realistic model for large scale and vague shape dynamic phenomena. This paper is organized as follows. After a brief background presented in section 2, section 3 describes the proposed approach implemented in section 4 for a bushfire simulation case showing it applicability for real time spatial decision process in disaster management. Section 5 presents conclusions and future works. 2. Background Spatial computing can be undertaken in a sensor system following a centralized (all data are sent to process center), a decentralized (located at sensor site or other site) or a hybrid approach (Chong and Kumar 2003). Crisp vector objects are extracted in the existing approaches using statistics or filters (Chintalapudi and Govindan 2003) or qualitative reasoning (Guan and Duckham 2009). Real-world phenomena are inherently uncertain (Carniel et al. 2015). Crisp-Fuzzy objects model (Pauly and Schneider 2008) is an interesting candidate for the representation of such phenomenon. In this model the geometry of object is composed of a kernel and a conjecture part, the kernel part belongs definitely and always to the vague object but one can’t say with certainty whether conjecture part considered as the broad boundary belongs to the vague object.