SMART WAT+

Abstract ………………………………………………………..Page 2 Problem to be Addressed……………………………………...Page 2 Analysis…………………………………………………………Page 3 Technical Solution……………………………………………...Page 5 Demonstration…………………………………………………..Page 10 Conclusion……………………………………………………....Page 10 Reference………………………………………………………...Page 11 Appendix A……………………………………………………...Page 12 Appendix B………………………………………………………Page 13 Appendix C………………………………………………………Page 14

Page 1 of 22 1. Abstract

As the number of people with access to electricity around the world rises, electricity meters have grown in significance. Today, standard electromechanical energy meters enabling the monitoring of electricity usage are still a market leader in terms of absolute install base. The technology, established more than a century ago, is very limited in terms of data communication, forcing the majority of energy supply companies to perform manual meter readings or energy bill estimation. These methods either lower company profit margins or increase energy bills for the consumer.

To address these problems, two major technologies have appeared in the commercial market, namely automatic meter reading (AMR) meters and meters using advanced metering infrastructure (AMI). Both wirelessly transmit data to the energy supplier; however have various limitations, the most significant being the cost. This is mainly due to the many other new functions these technologies provide, that are useless to the energy supplier simply wishing to reduce the cost of data collection. Due to the high costs, the absolute penetration of these technologies is low in most countries.

SmartWat+ aims to provide companies and related consumers with a solution that aims to satisfy the wireless data transmission requirement. This is done using a simple and cheap add-on that can be fixed onto existing electromechanical energy meters. The add-on is programmed to take a photo of the analogue display of numbers on the old meters at a given time interval, process the image and then wirelessly transmit the converted data to the company. A low-cost camera, a microcontroller equipped with image recognition software, and a Global System for Mobile (GSM) transmitter are responsible for this action. The cost of the add-on device is significantly lower than those using AMR and AMI technologies in most countries, solving the problem described.

Considering the number of currently operating electromechanical meters around the world, there is great potential for such a product. Even as major developing countries roll out AMI technology, new potential markets in Africa will emerge as electrification rates in the continent increase. This process, however, is relatively slow, in the range of multiple decades.

The following feasibility report on the new energy meter add-on details the research and results concerning the potential need, and hence market for this product. In addition an overview of the technology needed for such a product to be built is presented. Finally, a conclusion is drawn based on the cost and performance estimations of the product, where these are compared to the original design specifications derived from the initial problem presented above.

2. Problem to be addressed

The most widely used technology[2] indicating the electricity usage of a customer is the electromechanical watt-hour induction meter. It contains a disk rotating at a speed proportional to the power passing through it. The numerical value indicating the power usage rotates along with the disk, resulting in an analogue display, which must be read manually. Meter readings are costly, subject to error, slow, hence unpractical in an industry where a fast, reliable update to user’s consumption is invaluable to energy companies.

Due to the significant advancement of wireless communication standards, coupled with worldwide wireless coverage, new electricity meter technologies have emerged with wireless data transmission capabilities. Nevertheless, their additional functions and complexity result in a high cost, limiting their install base in most countries. The problem therefore lies in finding a means to provide electricity meters with wireless capabilities to the majority of electricity users at a low cost and time frame.

Page 2 of 22 3 Analysis

The proposed solution is an electricity meter add-on, which adds data transmission functionality to the already installed electromechanical electricity meters. In order to develop suitable design specifications for the product, research into possible markets (countries) has been undertaken. The feasibility of implementing such a product is highly dependent on the country under consideration. Identifying the correct market, the competitors in those markets (modern electricity meters), in addition to their prices in these markets is of pivotal importance. These analyses are presented in the following sections.

3.1 Appropriate Markets

A country where the implementation of an energy meter add-on is feasible has to meet several criteria for economic and technical purposes. These include: high electromechanical meter install base (in terms of percentage and volume), no governmental plans to introduce modern electricity meter technology on a national scale in the next 10 years[2], relatively high competitor prices, and reliable country-wide communications infrastructure. Research into 51 countries representing the world’s most significant energy meter markets has been undertaken using data from energy research companies[2]. Data from countries meeting all of these criteria are shown in Graph 1.

Graph 1: Install Base of Electromechnaical Meters by

160 Country 140

Millions 120 All Electricity Meters 100

80

60 Electromechanical Install Base Electricity Meters 40

20

0 United States Germany Japan Romania

Country

There are numerous country-specific reasons why modern electricity meter technologies have not been undertaken. In the United States, energy meter markets are in private hands, and due to the recent housing crisis and lower than expected revenues; companies have not invested in the technology[2]. In Romania, the situation is similar, as recently privatised former national companies cannot afford to invest in such technology[2]. In some highly developed countries, consulting firms have not recommended nation-wide modern electricity meter investment, due to the high reliability and economy of installed electromechanical meters. This was the recommendation by Ernst & Young to Germany[1]. On the other hand, delays in plans to implement the technology have occurred in certain countries, namely Japan[2]. An add-on with the correct design specifications would be feasible to implement in these countries.

3.2 Competing Technologies

3.2.1 Automatic Meter Reading (AMR)

AMR technology enables the wireless transmission of electricity usage from the energy meter to the energy supplier. There are two ways in which it can be applied to electricity meters. Page 3 of 22

 Option 1: Adding a communications module to existing electromechanical meters.  Option 2: Implementing the technology in a new digital solid-state electricity meter.

Communication modules use optical sensors that detect the number of revolutions of the rotating disk of the electromechanical meter, while in the second option AMR electricity meters use digital current and voltage sensors to measure the power. Both options use simplex (one-way) communication between users and energy providers, and simply match the digital data with a user account.

Limitations

 Option 1 – Small market share, can cost more than replacing existing electromechanical meter with a new AMR meter  Option 2 – Costs still range between 46-80 EUR[2] in most countries, gives utility providers power to remotely deny users energy.

3.2.2 Advanced Meter Infrastructure (AMI)

This technology is a more advanced version of AMR. The key difference is that it involves a duplex (two-way) communication between the users and energy providers. This allows for more advanced functions such as electricity readings on demand, and energy forecasting using built-in software.

Limitations

Due to its advanced functions, energy meters using AMI technology are more expensive than AMR meters or AMR communication modules (price range typically between 80 to 150 EUR). In addition, sensitive user data can be collected leading to privacy issues including the vulnerability of that data if hackers gain access through the built-in software.

3.3 Development of Constraints

The most important design constraint involves the target price of the add-on, as the main purpose of the technology is to provide a sought after function at a lower price. To set a price target, the price of the competing technologies must be assessed in each of the proposed countries. This is shown in Graph 2.

Graph 2: Unit Price of Electricity Meter using AMR and AMI Technology

Romania 236

*Note: AMR is Japan 80 not widely available in Romania and AMI 68 Germany Country Germany AMR

United States 83 46

0 50 100 150 200 250 Unit Price (EUR) Page 4 of 22

From Graph 2, it is evident that the cost of competing electricity meter technology significantly varies depending on the country under consideration. Note that AMR is not present in Romania and Germany. The deductions are that a target price can be set at approximately 30 EUR to give high enough price margin. Considering countries rollout new meter technologies to 80%[2] of the consumer base in approximately 8 years [2] and add-ons are simpler to install, we can estimate 5 years to 80% of consumer base. Table 1 gives the benefits assuming a cost of 35 EUR, and savings calculated as a result of the cheapest AMI/AMR technology available in a specific country. We can see that at 35 EUR significant savings are made.

Table 1: Estimated Savings by using SmartWat+:

Total Install Yearly Rollout Estimated Unit Price Estimated Base (Millions) (Millions) of SmartWat+ add-on Savings (EUR (EUR) Millions)

United States 144.3 23.1 35 254.0 Germany 45.1 7.2 35 238.1 Japan 89.7 14.3 35 645.7 Romania 8.8 1.4 35 281.7

4. Technical solution

4.1 Design Specifications

Our product must provide communication capability to existing electromechanical smart meters. In order to do this, it must capture the data from the meter, process it, and send it to the energy supplier. In order to achieve this, and ensure a maximal number of competitor’s limitations are overcome, the following design specifications are proposed:

. Image Capture: A camera can be used to take a picture of an electromechanical meter display. . Processing: A microcontroller is needed to support wireless communication, and image processing. . Software: Image recognition software is needed so that the processor can identify numbers in the image and convert them to machine-encoded numbers. . Simplex Communication: A wireless transmitter is required so energy usage data can be sent to energy supplier. This must be simplex to address privacy concerns of AMI technologies mentioned earlier. . Reliability: The product must reliably detect and send the number on the electrometrical meter to the energy supplier. . Cost: From the previous section it was determined that the add-on should cost around EUR 35. . One-Function Device: The device must only be able to read the energy consumption of the user. This is done to reduce complexity and lower the device price, in addition to alleviating concerns explained earlier, of AMR/AMI technologies violating user privacy.

Page 5 of 22 4.2 Overview of Functionality

To implement the desired function of the electromechanical meter add-on the following approach was taken. First, a camera takes an image of the electrometrical meter display (analogue numbers representing energy consumption in kWh on meter). Second, the microcontroller loads the image into memory, identifies the images using the installed image recognition software, and finally stores the converted machine-encoded numbers in memory. These numbers are then sent through a wireless network using a transmitter. The energy supplier will then receive the electricity meter reading of the user. Diagram 1 shows this process, where arrows represent the direction of communication between modules.

Diagram 1: SmartWat+ Functionality

4.3 Module Analysis

The add-on can be broken down into four modules: camera, processing, software and communication. Each module represents an existing technology made up of one or multiple products on the market. Different possible products for each module will be compared against the design specifications, and the right one will be chosen for the design of the add-on. This ensures a higher quality, reliable end-product design.

4.3.1 Camera

The two most popular image sensors of digital cameras are CMOS (complementary metal-oxide semiconductor) and CCD (charge-coupled device) image sensors. The two technologies were compared against criterion important to the add-on in Table 2. Note that camera must be able to take images in low-light conditions. It is clear from Table 2 that CMOS is the preferred technology.

Table 2: [3] CCD vs. CMOS Technology

CCD CMOS Low-Light Noise High Low Low-Light Image Quality Low High Cost High Low

The two most relevant CMOS cameras were compared in Table 3. The camera must be chosen so that image recognition software can easily identify numbers within the image. This requires high sensitivity, accurate colours and clean edges in the image. From Table 3 it is evident that OV7670 is a better candidate compared to the FCB-MA 130, even if it has a lower resolution.

Page 6 of 22 Table 3: Comparison of two CMOS cameras [4][5]

Camera Name OV7670 FCB-MA 130 Resolution Variable (Highest: 640x480 pixels) 1280 x 720 pixels Extra Features  Automatic exposure adjustment for  More compact accurate white colour  Image stabilisation  Enhanced edges in image  High sensitivity Cost Low High

In order to make a connection to image processing stage, appropriate output format of the image should be chosen. There are three types of available output formats in the OV7670 CMOS camera, namely monochrome, RGB and YCbCr (these are not acronyms). YCbCr requires the most number of bits per pixel, whereas monochrome requires the least [2]. To reduce sizes of the image and processing time, we will choose monochrome output format for image processing stage.

4.3.2 Software – Optical Character Recognition

Optical Character Recognition (OCR) technology is necessary in order to convert the images of the electromechanical meter readings to machine-encoded numbers. Due to budget and licensing constraints, only open-source OCR software has been considered in this feasibility study – they are Tesseract, Ocrad and GOCR. A comparison of the performance of available OCR software on bitonal characters is shown in Table 4.

Table 4: Performance of Available OCR Software[6][7][8]

Tesseract Ocrad GOCR

Average Error Rate (%) 0.9 6.3 14.3

Average CPU Runtime (Seconds) 21.31 4.80 114.83

From Table 4 one can see that Tesseract’s average error rate when recognizing bitonal characters is much lower than that of Ocrad and GOCR. While its speed is not the fastest, it is not as slow as GOCR. The shortest expected time interval between meter readings is one week, so Tessaract is sufficiently fast. Since accuracy of meter reading is one of the most important design specifications of the add-on product, while speed is not identified as an important criterion, Tesseract is the most suitable optical character recognition software for the add-on.

In an attempt to reduce the size hence resolution of the image for greater efficiency, Tesseract was tested on an image containing numbers for multiple image resolutions. It was found that the lowest image resolution where all numbers are recognized accurately is 176 x 144 pixels.

4.3.3 Microcontroller

A microcontroller-unit (MCU) will be used to provide a platform that interlinks the camera and communication modules. When selecting the MCU two criterion need to be considered: memory, the ability to connect to the camera and communication module, and programming the microcontroller. Selection of a microcontroller is based on the correct memory requirements that minimize price, and the support required for the two external module connections. Memory

Page 7 of 22 There are two types of memory to be considered: the program and data memories. A reasonable amount of program memory is needed so that the MCU can handle the control of the camera module, image processing software and communication module. The program memory must be kept as small as possible to minimize the cost. Data memory refers to memory storing the input image and output binary numbers. The size of the image taken from our chosen camera is 24.75KB (corresponding to the 176x144 pixel resolution) (see Appendix B), where the output numbers to transmit has negligible size. Table 5 below gives the price range and amount of different data and program memories in a range of possible microcontrollers.

Table 5: Microcontroller Comparison [1], [2], [3]

Microcontroller Memory Bulk Price/EUR

Data/KB Program/KB PIC 18F6722 3.83 128 6.4 PIC 16F877A 0.360 14 4.6

ATMEGA 328-PU 2 32 1.53

Due to the large size of the image from the camera module, external RAM (Random Access Memory) is needed when the image is processed. The microcontrollers listed above all have additional pins supporting this additional. Table 6 below gives the price range for different external RAMs.

Table 6: External RAM Memory Comparison [4], [5], [6]

External Memory Memory/KB Bulk Price/EUR Microchip 23A256 32 0.99

AMIC Technology A623308A 64 2.32

Microchip 23A640 8 0.28

Connection to other Modules

To permit a connection to the communication module a microcontroller that supports serial communication needs to be selected. A UART (Universal Asynchronous Receiver Transmitter) port will provide the required pins to enable serial communication. All microcontrollers above were selected to support serial communication. No special support is needed to connect to the camera module.

Programming

An appropriate integrated development environment (IDE) can be used to write and load appropriate programmes onto the chosen microcontroller. In the simplest add-on design, the program should be able to send data (SMS – see section 4.3.4 - communication) containing the model number of the electromechanical meter, as well as its associated meter reading. Sending time intervals can also be defined. Appendix D includes the program used to send data from the microcontroller to a mobile phone using the communications module explained in the next section 4.3.4 of the report. Processor Selection

Page 8 of 22 The built-in memories of all of the considered microcontrollers are too small to load the input image; hence an external RAM is needed. With reference to Table 6, the suitable choice is the A623308A product as only 24.75KB of data needs to be stored. Since the microcontroller needs minimal program memory and data memory is taken care of, the ATMEGA 328-PU microcontroller can be chosen for the add-on.

4.3.4 Communication

The digital communication system used for the add-on is channel access method, allowing several network terminals to transmit over the same communication channel hence share the same bandwidth[1]. This will allow millions of add-ons to communicate over a given bandwidth, reducing the operational costs. There are two main cellular network standards using this technology: Global System for Mobile (GSM) and CDMA2000. A comparison of the two standards is presented in Table 7 below.

Table 7: GSM and CDMA Comparison

GSM CDMA Advantages . Coverage: All countries apart . Better coverage in rural areas from Japan & Korea . More modern technology . More efficient in terms of . Very Secure power consumption Disadvantages . Less secure, but can be . High market share only in North improved using encryption America

Since worldwide coverage and power consumption are very important, GSM is the preferred technology. Japan, however can no longer be considered as a viable market as seen in section 3 of the report. Although security is also important for user privacy, GSM security can be increased using encryption (see Table 7).

In order to physically send the data from the electromechanical meter to the energy supplier using the GSM network, 3 main components are needed. These components, forming the communication module of the add-on include: a GSM modem, a SIM cardholder and an antenna. The GSM modem provides access to a given mobile network through the SIM card, and sends the information to the energy supplier via the antenna. Table 8 below summarises main specifications and costs of the components chosen for the communication module.

Table 8: Essential Components for GSM Communication [9], [10], [11]

Component Detail Bulk cost/EUR Anaren  Supports 2.4 – 2.483GHz 2.95 66089-2406 antenna  GSM supported E.A.D  850 - 2100MHz 6.4 FQTN35144-MR-16 Antenna  GSM Supported Huawei  GSM protocol 18.7 MG323-B GSM modem  850 –1900MHz Molex  SIM socket 0.67 91236-0002 SIM card holder

When considering the antennas, the cost and bandwidth are the most important criterion. Nevertheless, bandwidth is more important as it has to function in the countries highlighted in section 4.3.4. These countries have variable frequency bands, some below 2.4GHz. This means Page 9 of 22 the first antenna from Anaren must be chosen. The SIM cardholder and GSM modem were selected based on price, where the selected modem had to match the frequency band of the antenna.

Operational costs of using GSM networks are out of the scope of this feasibility study due to the complexity of the analysis. Many factors influencing price cannot be evaluated such as typical contracts for bulk consumption.

5 Demonstration

A demonstration of the Tesseract image recognition software, as well the communication module was undertaken. The first demonstration involved downloading the open source software and testing it against an image of the previously mentioned 176 x 144 pixels. All the numbers were successfully recognized. The second demonstration tested the communication module of the add- on. A microcontroller, and all components of the module were purchased and tested. First, a program was written to instructing the microcontroller to send given numbers to a mobile phone number (See Appendix D for code). Second, a SIM card was inserted into its slot, and the SMS was sent to the mobile phone. A screenshot of the process is shown in image 1 below.

Image 1: Communication Demonstration

6 Conclusion

The Smartwat+ add-on can be constructed to fulfil all the design specifications specified in section 3 at a cost of 34.34 euros. The table below breaks down this final cost, gathered from the previous sections of the report.

Table 9: SmartWat+ Cost Component Cost/EUR Microcontroller 1.53 External Memory 0.99 Camera 5.99 Communication Subtotal 25.84 Grand Total 34.34

In section 3, a target price of 35 euros was set to ensure the feasibility of the product from a market perspective. The add-on and competing product price differences were evaluated and estimated savings were calculated. From the technical section two adjustments were found. Firstly the cost of the add-on is 0.66 euros lower than estimated, and secondly the add-on cannot be implemented in Japan due to it not supporting the GSM network. The re-calculated savings for the appropriate countries are shown in Table 9. Page 10 of 22

Table 10: Savings of SmartWat+ United States Germany Romania Savings (EUR Millions) 269.2 242.9 282.7

It is clear that if energy companies choose SmartWat+ over its competitors to give electromechanical meters communication abilities, savings in the order of hundreds of million euros can be made. The research undertaken in this feasibility report clearly shows that the product is feasible to implement in the countries listed in Table 9.

References

[1]: Ernst & Young, Cost-benefit analysis for the comprehensive use of smart metering, 2013, http://www.bmwi.de/English/Redaktion/Pdf/cost-benefit-analysis-for-the-comprehensive-use-of-smart- metering-systems,property=pdf,bereich=bmwi2012,sprache=en,rwb=true.pdf

[2]: ABS Energy Rearch, World Electricity Meter Report, [Article] 2010, Vol. 8, p1-358. 358p. 307 Charts, 63 Graphs, 2 Maps, 55231842

[3]: Carlson B.S, Comparison of modern CCD and CMOS image sensor technologies and systems for low resolution imaging, 2002 Proceddings of IEEE, 10.1109/ICSENS.2002.103701

[4]: Omnivision, OV7670 Datasheet, 08/07/2005 http://www.voti.nl/docs/OV7670.pdf

[5]: Sony, FCB-MA 130 datasheet [online] http://www.phase1tech.com/userFiles/product_files/FCB-MA130.pdf

[6]: R. Smith, "tesseract-ocr - An OCR Engine that was developed at HP Labs between 1985 and 1995... and now at Google.," Google, [Online].

[7]: A. Diaz, "GNU Ocrad Manual," 2 January 2014. [Online]. http://www.gnu.org/software/ocrad/manual/ocrad_manual.html [8]: R. Smith, "An Overview of the Tesseract OCR Engine," in Document Analysis and Recognition, 2007. ICDAR 2007. Ninth International Conference on, Parana, 2007.

[9]: Microchip, PIC18F8722 Family data sheet. http://ww1.microchip.com/downloads/en/DeviceDoc/39646c.pdf

[10]: Microchip, PIC16F87XA data sheet http://ww1.microchip.com/downloads/en/DeviceDoc/39582C.pdf

[11]: Atmel, Atmet 8-bit Microcontroller with 4/8/16/32Kbytes in system programmable flash, 2012 http://www.atmel.com/Images/doc8161.pdf

Page 11 of 22 Appendix A: Market Analysis

Table 1: Calculation of Electromechanical Install Base

Raw Data: Raw Data: Processed Data: Absolute Electricity Install Base of Install Base of Meter Install Base Electromechanical Electromechanical Electricity Meters (%) Electricity Meters (Units) United States 144320307 75 108240230 Germany 45110298 95 42854783 Japan 89677255 97 86986937 Romania 8760504 100 8760504

Table 2: Calculation of Automated Meter Reading Electricity Meter (AMR) Prices:

Raw Data: Demand Raw Data: Price for Processed Data: Unit (Units) Demanded Units Price (EUR) (EUR) United States 400000 18400000 46 Germany 0 N/A N/A Japan 10000 800000 80 Romania 0 N/A N/A

Table 3: Calculation of Advanced Meter Infrastructure Electricity Meter (AMI) Prices:

Raw Data: AMR Raw Data: Price for Processed Data: Unit Demand (Units) Demanded Units Price (EUR) (EUR) United States 16500000 1369500000 83 Germany 2000000 136000000 68 Japan 3500000 280000000 80 Romania 3000 708000 236

Table 4: Calculation of Expected Savings:

Raw Data: Estimated Estimated Data: Processed Processed Data Processed Total Install Data: Yearly Estimated Unit Data: Price Difference Data: Savings Base Estimated Price of Lowest between (EUR) (Units) Rollout SmartWat+ Competitor SmartWat+ and (Units) Add-on (EUR) Price per lowest price Unit (EUR) competitor (EUR) United States 144320307 23091249.12 35 46 11 190,502,805 Germany 45110298 7217647.68 35 68 33 226,273,255 Japan 89677255 14348360.8 35 80 45 626,305,949 Romania 8760504 1401680.64 35 236 201 281,737,809 Page 12 of 22 Appendix B

The estimated size (in Kilobytes) of photo can be computed given pixel size and bit-per-pixel:

Height – Image height in pixels (176 for image used for SmartWat+) Width – Image width in pixels (144 for image used for SmartWat+) Bit per pixel for Monochrome format – 8 bits per pixel (SmartWat+ uses monochrome output)

Page 13 of 22 Appendix C: Image Processing

Optical Character Recognition (OCR)

OCR is the conversion of images of texts into machine-encoded texts. For this project, OCR is necessary in order to convert the images of electricity meter readings to machine-encoded numbers. There is a wide selection of commercially available OCR software, which can be modified and adapted to the project. However, due to licensing issues, only OCR programs with a free software license will be considered (e.g. Apache, GNL)

Overview of Free OCR Software

 Tesseract Features: Tesseract is “probably the most accurate open source OCR engine available” Only reads an image in TIFF format by default Command-line program Written in C/C++ Can be configured to recognize only digits

Minimum Text Size: “Accuracy drops off below 10pt x 300dpi, rapidly below 8pt x 300dpi. A quick check is to count the pixels of the x-height of your characters. (X-height is the height of the lowercase x.) At 10pt x 300dpi x-heights are typically about 20 pixels, although this can vary dramatically from font to font. Below an x-height of 10 pixels, you have very little chance of accurate results, and below about 8 pixels, most of the text will be "noise removed".”

 GOCR Features: Converts scanned images of text into text files Command-line/user interface Written in C

Minimum Text Size: No information found online

 Ocrad Features: Only reads an image in pbm (bitmap), pgm (greyscale) and ppm (colour) formats Character recognition is based on an algorithm that recognizes characters by shape. It uses a strategy similar to binary search, thus making it very fast. Command-line program

Minimum Text Size: No Information found online

Page 14 of 22 Comparison of Accuracy and Runtime among Tesseract, GOCR and Ocrad

Tesseract, GOCR and Ocrad were tested on accuracy and runtime by Peter Selinger at Dalhousie University. The results of the tests revealed that in terms of runtime, Ocrad was fastest, followed by Tesseract, which was moderate, and GOCR was found to be very slow. In terms of accuracy, Tesseract performs significantly better than both Ocrad and GOCR. Ocrad was found to be equal to or more accurate than GOCR

Minimum Resolution Required for Accurate Tesseract OCR Recognition According to Tesseract documentation, the minimum size required for the software to accurately recognise characters is specified in terms of the pixel height of the lowercase character “x”, also called x-height. For an x-height of 10, accuracy starts to drop, and below 8 pixels, most characters will not be correctly recognised. The specifications above were verified by testing Tesseract’s ability to recognise integers from 1 to 9 with varying resolution. The font used was Arial, and the font size was set to 72pt. The results are shown in the table below:

Table 1 summarises results from images with different specifications

Accuracy of x-height Image Tesseract Number Image (JPEG) (pixels) Resolution Output Recognition (%)

52 960x720 0123456789 100 X

13 240x180 0123456789 100 x

Page 15 of 22 11 192x144 0123456789 100 x

10 176x144 0123456789 100 x

9 149x112 0123455789 90 x Error: ‘6’ incorrectly recognized as ‘5’

8 141x106 0123456189 90 x Error: ‘7’ incorrectly recognised as ‘1’

Page 16 of 22 6 110x82 mzussru 0

Error: All numbers were incorrectly recognized

Page 17 of 22

Appendix D: Demonstration

Overview

Figure 1: Wiring schematic for GSM communication implementation

The Tx pin, and Rx pins from the GSM shield – Transmit and Receive – are connected to pin 2 and 3 of the microcontroller respectively. Pin 2 and 3 of the microcontroller will be initialised to Tx and Rx via programming afterwards – refer to programming section below. The ground for the microcontroller board and GSM shield will be common, so that Tx and Rx pins work as expected. Each board will be supplied with appropriate voltages: 3.3V for that particular GSM shield, and 5V for the particular microcontroller board. Experiments with appropriate programmes have shown that microcontroller board indeed requires 5V instead of 3.3V, due to the fact that more power is required to process Tx and Rx pins.

Page 18 of 22

Programming

AT commands are insruction sets used to control modems through interaction with the devices - microcontroller, PC or others. Such controls include SMS services, data link, calling services and any other configuration pertaining to mobile devices. Below code shows how to use AT commands to send a SMS using GSM module.

Figure 2: Code developed to implement communication module

Page 19 of 22 http://www.sciencedirect.com/science/article/pii/S0168900203007927

A - Camera

References:

[4]: Microchip, 23A256/23K256 datasheet, 08/02/2011 http://ww1.microchip.com/downloads/en/DeviceDoc/22100D.pdf

[5]: AMIC technology, A623308A, 07/2006 http://www.farnell.com/datasheets/63505.pdf

[6]: Microchip, 23A640/23K640 datasheet, 08/02/2011 http://ww1.microchip.com/downloads/en/devicedoc/22126c.pdf

[7]: Kamil Sh Zigangirov, Theory of Code Division Multiple Access Communication, Chapter 1.2, 2004 http://ieeexplore.ieee.org/xpl/ebooks/bookPdfWithBanner.jsp?fileName=5237094.pdf&bkn=5237 086

[8]: MOBITEK System Sdn. Bhd, ‘What is GSM modem?’ , 2008 http://www.mobitek-system.com/Products/GSM_Modem.html

[9]: ANAREN – 66089 – 2406 ANTENNA, 66089, 24C, 2400MHz, http://www.anaren.com/sites/default/files/uploads/File/66089-xxxx-Antenna.pdf

[10]: Embedded antenna design – FQTN35144-MR-16- ANTENNA, http://www.semiconductorstore.com/pages/asp/DownloadDirect.asp?sid=1394249273334

[11]: Huawei, Huawei MG323-B GSM LCC module hardware guide, 02/07/2012 http://consumer.huawei.com/en/solutions/m2m- solutions/products/support/datasheets/detail/mg323-b-en.htm?id=14461

[12]: Molex, ChipSIM connectors and Cardholders, 16/09/2013 http://rhu103.sma-promail.com/SQLImages/kelmscott/Molex/PDF_Images/987650-7041.PDF

[13]: Omnivision, OV7670 Datasheet, 08/07/2005 http://www.voti.nl/docs/OV7670.pdf

[14]: J. Schulenburg, “GOCR-documentation”, 03/06/2002 http://aist1.pisem.net/gocr.html

[15]: A. Diaz, "GNU Ocrad Manual," 2 January 2014. [Online]. http://www.gnu.org/software/ocrad/manual/ocrad_manual.html [16]: R. Smith, "An Overview of the Tesseract OCR Engine," in Document Analysis and Recognition, 2007. ICDAR 2007. Ninth International Conference on, Parana, 2007.

[17]: R. Smith, "tesseract-ocr - An OCR Engine that was developed at HP Labs between 1985 and 1995... and now at Google.," Google, [Online]. https://code.google.com/p/tesseract-ocr

Page 20 of 22 [18]: P. Selinger, "Review of Linux OCR Software," Department of Mathematics and Statistics, Dalhousie University, 27 August 2007. [Online]. http://www.mathstat.dal.ca/~selinger/ocr-test/.

[19]: R.F solutions, TMAS GSM Modem, October 2007 http://datasheet.eeworld.com.cn/pdf/RFSOLUTIONS/296843_CABLE-RS232.pdf

[20]: SIMcom, SIM900 the GSM/GPRF module for M2M applications, http://www.sos.sk/a_info/resource/c/sim900.pdf

[21]: SIMcom, SIM300 Hardware specification, 27/12/2005 http://probots.co.in/Manuals/SIM300.pdf

[22]: Siemens, TC35, TC37 Hardware interface description, 21/12/2001 http://www.robotshop.com/media/files/pdf/datasheet-gsm-tc35.pdf

[23]: GSM bands information by country, 12/02/2014 http://www.worldtimezone.com/gsm.html

[24]: Sourceforge, gimagereader - A graphical GTK fronted to tesseract-ocr 20/02/2014 http://sourceforge.net/projects/gimagereader/files/latest/download

[25]: Centre for sustainable energy, Reading your gas or electricity meter 2013 http://www.cse.org.uk/advice/advice-and-support/reading-your-gas-or-electricity-meter

[26]: Energy firms accused of adding £560m to household bills 13/01/2011 http://www.theguardian.com/money/2011/jan/13/household-energy-prices-rise-oil-soars

[27]: Google, tesseract-OCR FAQ 29/01/2014 https://code.google.com/p/tesseract-ocr/wiki/FAQ

[14]: J. Schulenburg, “GOCR-documentation”, 03/06/2002 http://aist1.pisem.net/gocr.html

Reference:

 1) “AUTOMATIC ELECTRIC METER READING SYSTEM: A COST-FEASIBLE ALTERNATIVE APPROACH IN METER READING FOR BANGLADESH PERSPECTIVE USING LOW-COST DIGITAL WATTMETER AND WIMAX TECHNOLOGY”, from a publication of “G-Science Implementation & publication”, by TANVIR AHMED, MD SUZAN MIAH, MD. MANIRUL ISLAM and MD. RAKIB UDDIN, on Sept. 2011

 2) “Smart metering system and intelligent meters”, from “Ernst & Young: Cost Page 21 of 22 benefit analysis for the comprehensive use of smart metering”

 3）“Technical and governance considerations for advanced metering Infrastructure/smart meters: Technology, security, uncertainty, costs, benefits, and risks”，from "Energy Policy”, by Mark P. McHenry

 4) from Mamoun Shahin, “Smart Meter Middle East” , by Metering.com, 14th Nov. 2008. URL: “http://www.metering.com/smart-metering-middle-east-13930/”

 5) “Smart meter data: Balancing consumer privacy concerns with legitimate applications”, from “Energy Policy”, by Eoghan McKenna , Ian Richardson, Murray Thomson

Page 22 of 22