Overview of a system AMR based in computational Vision and Wireless Sensor Network Jorge R. B. Garay Integrated Systems Laboratory University of Sao Paulo (USP) CEP 05508-970– São Paulo – Brazil {Jorge@pad.lsi.usp.br} Sergio T. Kofuji Integrated Systems Laboratory University of Sao Paulo (USP) CEP 05508-970– São Paulo – Brazil {kofuji@pad.lsi.usp.br} Tadayoshi Tiba Materials Research Laboratory – MRL, SP – Brazil {tiba@pad.lsi.usp.br} Abstract - This paper presents the proposal of integrating an Automatic Meter Reading System (AMR) with Computer Vision and Wireless Communication. The AMR system reads the numeral meter through a video micro-camera, process the collected image using an Optical Character Recognition (OCR) module embedded in the sensor node. The results are sent through ZigBee wireless links to the center of management, in which a mesh topology is used for communication. Keywords: AMR, Visual Sensor Networks, OCR, ZigBee, Mesh topology. I. INTRODUCTION This paper approaches both conceptual design and implementation, for a home automation system for reading water and power consumption meters; concepts and techniques for the evolution of the system, which might be extended to it, are also explored. Some of those techniques are presented at [8], the new method recognize the meter´s reading utilizing the computer vision technique, it is a kind of contact-less measurement; Bannak and Nascimento [1] propose the use of an electronic system integrated in a Smart Card for reading power meters; at [3],it is proposed the development of an heterogeneous system to determine the rule on problems with measurement; Hafner, Erig and Lopes S. [6] propose the use of reconfigurable logic to create a new quality measurer embedded with particular characteristics of processing and storing data; Gordan Štruklec and Vedran Bilas [5] develops an AMR for reading water meters with wireless communication (ZigBee), the system considers the use of sensors assembled to the meters, that processes and stores consumption information at the meters; possibly one of the most relevant developments is found at [8], where it is proposed an AMR based on a RF network , the communication protocol used is ZigBee. The author implements sensor modules embedded on the meters that does its reading and then transfers the collected info, which departs from a ZigBee channel to a CDMA one which is integrated to a TCP infra-structured network for the treatment of the collected data. Some aspects in common are easily identified among the diverse initiatives. The first of them is the creation of solutions on embedded hardware and software to optimize the quality of the meters readings. Other than that are the integrations of heterogeneous hardware, software and wireless systems, which allow for a wider understanding of the information shared through logical information models and visualization of the collected data in real time. The objective of this paper is to present an integrated system (software and hardware) which would allow the automation of the meter readings system and in which the tools and techniques in use could offer a contribution towards the search for solution of the existing problems on such a systems, specially on what it matters to the automation process for water and power data acquisition and fraud fight. A relevant aspect of this project, regarding the use of ZigBee communication and mesh topology, is that the processing of video images is done locally in real time (at the sensors) using an embedded OCR, which reduces the throughput of the network. In section 2 it is presented an overview of the systems components. In section 3 the process of imaging acquisition and treatment is described. On section 4 the ZigBee communication process is addressed and previous results of the utilized hardware are presented and finally at section 5 the conclusions are shown. II. SYSTEM OVERVIEW In the present time, the process of reading the meters are done manually by employees of the service provider company or by third companies. This procedure ends up by adding costs to the service which is offered to the consumers, while the proposed system, in which the video images are acquired from the meters, provides a remote monitoring, more reliable and without the additional costs of manual readings. The implementation of automation in the meter reading system is described in three layers, each one of them independent of its implementation, making way for a future optimized system´s performance. The layers can be observed in the diagram of figure 1. 978-1-4244-4388-8/09/$25.00 ©2009 IEEE