Just Some Intro

Water meter can be found in every household and industry. There are an estimated of 3.6 million household in Johor, Malaysia in 2016 [1]. Therefore, there are 3.6 million of water meters needed to be taken care of SAJ Ranhill Sdn Bhd, a Johor state government-linked company which responsible for water supply in Johor. The conventional way of collecting data of water meter from residential area is by sending technicians to record those reading. These method is very costly and time consuming [2]. Other than that, technician who record the reading into notebooks or electronic PDAs and store the data into database for further processing, might record a wrong data from a meter due to human errors. Other several factors, such as weather and location of water meter can cause problems to the technician. Technician might be unable to complete his job due to rainy days. Malaysia is a tropical climate country in which there is no dry season [3], therefore, weather factor contributes a high percentage of work failure for SAJ technician.

Previously, there exists many Automated Meter Reading methods proposed to assist data gathering. For this project, machine vision is applied on the water meter. By applying this method, the water meter reading can be captured from a camera, process it and then send to our database storage. Therefore, the problem of high human labor cost and time consuming can be solved.

Implementation

Most of the materials were bought from a hardware store in order to get an affordable price. After acquiring all the materials, the assembly of the proposed device commences. The first part of the mechanism of the system was building the upper part of the casing. The aluminum is cut according to the measuring design. The aluminum is cut using an aluminum cutting machine. This part of the project was completed in Kangkar Pulai. the finished design for the upper and lower part is shown in Figure 4 and figure 5 respectively. For the Perspex, a glinder is used based on the measuring design. The Perspex is used to allow sunlight to penetrate the device hence providing more brightness for the purpose of image processing.

Figure 4: The steel band is use to clamp between the pipe meter and the casing.

After the casing was completed, the pipe meter is then attached to the box using a steel band. A glue gun is then used to clamp the parts. A magnet bar is used to attach the upper and lower casing. The magnet bar was glued tightly in the difference pole between upper and lower casing. Figure 5 shows the magnet bar attach on lower and upper casing.

Figure 5: Shows 4 magnet bars attaching on lower part of the casing.

Finally, the stand of the pipe was made up from the PVC pipe where they were cut and joined by using an L shape connector. The height of the pipe is about 1 meter. To attach between the stand height and the base we just using the PVC glue to make it stronger. Figure 5 shows the structure of stand and pipe casing.

Raspberry Pi 3 is a tiny, affordable and portable PC as shown in Figure 6. Its operating system is Raspbian which is based on Debian operating system. Its high CPU 1.2 Ghz and 1 Gb of Ram is the reason why we are using it for image processing. It has Wifi module inside, so IOT enabled project can be done using Raspberry Pi 3. We also can insert Raspberry Pi Camera into the slot provided.

Figure 6: Raspberry Pi 3

First, the camera takes a picture and pre-process it. During the pre-processing, it includes image segmentation, noise filtering, gray-scale conversion, so that the detection and isolation of digits is possible. After isolating the character, now we need to train the character by using supervised learning. Supervise learning is a machine learning task of inferring a function from labeled training data. The training data consist of a set of training example. In completion, our system is able to display the corresponding character on the screen.

The software used is RealVNC. The main function of this software is to control the raspberry pi using a laptop from far away. However, both raspberry pi and laptop are required to be connected to the internet in order for the software to work. In RealVNC, there are two type of software which are VNC Server and VNC Viewer. The VNC Server need to be installed to the device that need to be controlled while the VNC Viewer is for the device that wants to control. In this case, the VNC Server will be installed into the raspberry pi while the SAJ worker will need to install the VNC Viewer. Both VNC Server and Viewer must be connected to the same VNC ID. It is one of the security system for the software so that outsiders cannot access the raspberry pi. The structure of the stand pipe and the final completed structure is shown in figure 7 and 8.

Figure 7: The structure of stand and pipe casing.

Figure 8: shows the completed structure of the smart monitoring pipe system.

The raspberry pi 3, camera module and other electronic part are placed in the casing of water meter. Figure 9 shows the image of water meter box build up to store the electronic component, hence the components are not damaged from vandalism and harsh weather. The top box is built with glass for the light to enter the casing to give lighting to camera as shown in figure 10.

Figure 9: The casing for water meter.

Figure 10: Top view for the casing.

Once the system starts to operate, the raspberry pi will be turned on. When the raspberry pi is turned on, it will automatically connect to the internet that it recognized. After that, operator will need to turn on their laptop, connect to the internet and open the software VNC Viewer. From the VNC viewer, the operator can have full control of the raspberry pi. Therefore, the operator will turn on the camera so that they can see the camera displaying the reading on the water meter. After the reading are retrieved by the worker, they will do the bill calculation and update it into the website as shown in figure 10. This website can only be accessed by the house owner of the water meter and the operator in charged. However, only the operator can edit the website. On the website, the operator will fill in the meter reading, bill, and the date the reading was taken. After the system has updated it, the customer can open the web to see how much they need to pay and make the payment by clicking the button “here” to automatically redirect them to the SAJ e-bill website. This is conveniently presented in Figure 11.

Figure 11: The website that the end user can access.

Figure 12: The website that the end user can access.

Results (and Discussion)

During the progress for the image processing, several errors are encountered. The image processing from the Raspberry Pi camera is supposedly detect the number at the water meter and display the output which is the number on the meter. Then, it has been decided that to use a different approach in order to get the reading which is by using the RealVNC. By using the RealVNC, the raspberry pi can be controlled from anywhere by using a laptop as long as both laptop and raspberry pi were connected to the internet. By doing so, the camera can be accessed manually through the software. Therefore, the reading of the water meter can be taken, as shown in Figure 13.

Figure 13: Detected character on Raspberry Pi 3

During the first test, the image displayed by the camera was not so clear. It was identified that the raspberry pi camera was too close the water meter. Even though we had already tested the distance of the camera and the meter before to get a clear vision, but when testing of the camera along with the meter casing was done, the image was a little bit blur. This may due to the shadow caused by the casing and the raspberry pi when the camera is put above the meter. As a solution, a slight adjustment of the casing was performed so that the distance of the camera and the meter can be increased and reduce the shadow covering the meter.

During the final field testing, the camera display was perfect when the device were put under sunlight. However, during the testing inside the presentation room, the light inside the room was insufficient to get a clear display of the camera. As a solution, a torchlight was put in the casing hence a much clearer view can be obtained. In addition, the position and angle of the light is also important so that the light will not dazzled the camera since the light is too near to the camera.

Conclusion

This project utilizes Conceive – Design – Implement – Operate (CDIO) type of framework. Before doing any designing, the team had started a survey by doing a questionnaire in order to gain more information about the end user and what are the problems they are facing. After hearing their thoughts, then the team had started to brainstorming on what problem faced by the end user that should be focus on. After it has been decided, the team then starts to find solutions to solve the problem and started designing the devices.

After a few alterations made from the original plan of the prototype, the objective of the design were successfully achieved. The main objective of taking the water reading from anywhere has been successfully done. The alterations were made in order to suite the real world field environment so that the device can work more efficiently.