Design and Development Low Cost Coral Monitoring System for Shallow Water based on Internet of Underwater Things

Abstract — Coral monitoring has become major focus to prevent coral bleaching, so that various methods have been developed by researchers to analyze coral reefs’ health. This paper proposed a low cost coral monitoring system based for shallow water based on IoUT architecture. The proposed system consists of buoy as base component, controller unit based on Single Board Computer equipped with 5V power source, underwater camera and communication unit to send data into cloud server. This system is able to extend Wi-Fi signal from underwater camera by using coaxial cable and transmit image into cloud server. From experiment, we measured that reliability delay using coaxial cable to capture and save image is 5,6 seconds. Our performance experiment on cloud server indicate that distributed server need 186,6 seconds to save retrieved images while single server need 53,6 seconds. Moreover, we also have developed bleaching analysis using color detection and online representation model of our retrieved coral images.Keywords — iout; coral bleaching; underwater monitoring; low cost system

I. Introduction

In ocean ecosystem, coral reefs take many important roles to support underwater life. It provides feeding, spawning and nursery ground for underwater existences. Moreover, coral reefs also provide coast protection from damaging wave, help nutrient recycling, assist carbon and nitrogen fixing, and habitation for approximately 25% of diverse biotas. As the result, the livelihoods of 500 million people and income worth over $30 billion are at stake [1].

There are many challenges to keep coral reefs healthy, such as global climate change, sea temperature elevation, solar radiation and coral diseases. One of the most influential factor is coral bleaching. Coral bleaching is defined as expulsion of coral’s pigment algae, causing coral reef become pale and allows the skeleton to become visible through transparent tissue [2]. When it happens, coral reefs will starving. If bleaching is prolonged, it causes death to coral. A prevention act is needed to avoid a damage to coral reefs.

Coral monitoring has become a major focus to prevent bleaching. There are numerous approach to monitor coral to prevent from bleaching. Collecting sample manually, conducting surveys, and map it into color index are no longer considered as effective [3], [4], [5]. Study of [5] showed that lack of coordination, inefective cost and scaling issue still remains as challenges. Using remote sensing to perform monitoring program potentially addresses many of those caveats. Another approaches are also proposed by researchers, such as image analysis [3], multispectral remote sensing [4] and thermal stress analysis [6]. However, those approaches are relatively costly because not every researchers and scientists have adequate access to satellite imaging data.

In this paper we propose a design of underwater monitoring system, using low cost, low power and portable sensor to monitor coral reefs on shallow water. By using Internet of Underwater Things (IoUT) architecture, we deliver a prototype that has portability, modularity, nearly real-time data transmission, and integration into Big Data architecture. In particular, the contribution of this paper is the development of low-cost networked embedded system combined with underwater camera and Big Data architecture to provide continuously coral reefs monitoring system.

The remaining of this paper is organized as follows: Section II reviews previous works from other researchers. Section III presents the system design of our device. Section IV presents the experimental implementation of the hardware and software modules and finally the paper concludes with Section V.

II. Previous Works

Various methods have been actively developed by researchers to analyze and monitor coral reefs’ health which many of them are using remote sensing technique. Many aspects with use of remote sensing are able to evaluate global environment affecting coral bleaching, such as using thermal stress to analyze coral bleaching [6] and multispectral thermal sensors [4]. The advance of this method is likely to continue because it’s considered as promising method regarding its capability to respect the synthesis of multiple data products [5]. Similar approach using digital image processing is also proposed in [3] which is able to quantify whiteness level of coral by image data into qualitative data. These mentioned methods are not requiring portability feature because they process available data from satellite and digital camera. But these research also denote that using image processing is capable to assess underwater environment.

While the use of remote sensing is extensively used, another researchers did study to built Underwater Wireless Sensor Networks (UWSN) to put sensor nodes and retrieve image data, seawater temperature and GPS tracking [7]. The stationary underwater sensor nodes contain embedded computer with a wireless LAN and GPS sensor in every node of sensors. But considering the information capacity, they didn’t use low frequency as the wireless communication and separate communication and sensors function.

Technological revolution of Internet of Things (IoT) also bring new approach to explore water areas, which is known as Internet of Underwater Things (IoUT). It creates machine-to-machine network and links real life with physical device. IoUT also has some difficulties that must be tackled to create solid architecture such as different communication technologies, network density, tracking technologies, etc. However, we use proposed architecture presented in [8] into our low cost system because it has simple and clear layers diversity.

The use of transmission method must be considered as important to transmit data from underwater camera to the cloud server. This remains as challenge since radio waves do not propagate well underwater. Therefore, most communication of underwater rely on acoustic links, altough it has narrow bandwidth [8]. Tsuyoshi et al. use watertight LAN cable to transmit data from underwater camera to floating embedded computer [7]. An experiment from Anguita et al. represent that UWSN can be used to transport data with the throughput result of 100 kbps [9]. That experiment also indicate that it works not beyond 1,9 meters. So it cannot be used in our system. To transfer image from Raspberry Pi to cloud server, result from [10] represent that it’s possible to use mobile network to connect into internet. Our low cost monitoring system is also use this approach to send data into our cloud server.

III. System Design

Fig. 1 below illustrates an overview of our proposed low-cost monitoring system. It contains four main subsystems: buoy mechanics, underwater camera to take coral picture periodically, a control unit to run image capture program, and communication unit to transmit data into cloud server.

Fig. 1 System overview

A. Buoy mechanics

The buoy is the floating component where remaining subsystems are placed. This buoy is intended to keep control unit, power source and communication unit safe from water. It has wheel-shaped form and made of fiberglass to keep air trapped. It also has a curvature in the middle of buoy to put Raspberry Pi in it. There is also a small hole in the bottom of the buoy to put underwater rod to stick underwater camera in it.

Fig. 2 Buoy model

This buoy is not designed as a fixed model. Therefore, we didn’t attach it to the ocean floor to let it movable as seen in Fig. 2.

B. Underwater camera

Underwater camera is a device attached to underwater rod that used to capture coral’s condition. The rod has 3 meters in length because we targeted shallow water area. The device we use here is GoPro Hero 4 with Wi-Fi turned on to communicate with Raspberry Pi. This kind of communication lead to loss of Wi-Fi signal, because water will absorb 2.4 GHz frequency produced by GoPro. To overcome this limitation, many researchers use alternative approach such as using Underwater Acoustic Networks (UAN) or watertight LAN cable. But by using coaxial cable type RG 174/u, we’re able to extend Wi-Fi signal to reach Raspberry Pi by glue coaxial cable in end in underwater camera case and in control unit case.

We also made some adjustment on GoPro default setting regarding burst mode shoot. Instead of using 30 consecutive shoots, we set it into 3 sequence shoots to reduce GoPro’s inner post processing time.

B. Control unit

Control unit play the most important role because it handles many parts of the system: a) calling GoPro API to send capture command remotely b) run cron jobs scheduler to capture image automatically c) providing temporary image storage for captured image. The control unit we used is Raspberry Pi 3 model B version that has built in Wi-Fi adapter and powered by 5V DC 10.000 mAH power bank. In the USB port, we attach Huawei E1750 as a communication device to connect to internet. For ease of use we connect HDMI port into touchscreen monitor to monitor program output.

In the software environment, we use Raspbian OS Jessie version as the SBC operating system. For application development, we use Node.js v4, Python 2.7, and Bash shell. The capture application is natively written in JavaScript run by Node.js. We utilized a function from goproh4 library [11] to make HTTP API call to underwater camera, download the image to the Raspberry Pi, and transmit it to the Big Data server immediately as shown to the algorithm 1. This system is also using Python to listen physical button to start capture application. To operate continuously, we use cron job scheduler to execute application every particular time.

D. Communication unit

Communication unit will provide internet connection and allow control unit to send image data into cloud server. The device of communication unit is GSM Modem Huawei E1750 which able to transmit image with throughput around 3,1 Mbps over 3G mobile network.

Control unit will execute command through wvdial driver, which will send command to broadband GSM Modem to connect into internet. The transmitted data will be sent to our Big Data cloud server, which will receive it and save it to be analyzed later.

IV. Implementation and Experimental

In this section we will describe the implementation of the software and hardware development as well as the performance of experimentiation.

A. Implementation

After develop whole components individually, we assembling them into integrated system as seen in Fig. 3. This system mainly consists of two sections, underwater part and floating part. In the underwater part there’s underwater camera sticked to underwater rod. For the floating part we built box case to put Raspberry Pi, broadband GSM modem, 10.000 mAH power bank and touchscreen monitor.

Those sets were assembled into buoy before we detached it to the shallow ocean area. We also have written a Bash script to run wvdial automatically when GSM modem attached into Raspberry Pi.

Fig. 3 Communication unit components

Fig. 3. represents component structure of our low cost system. The dimension measurement of case box without its base are 36cm x 23cm x 6cm as seen in Fig. 4., while the buoy has radius 90cm and height of 87cm. When we attach 3m underwater rod in the middle of buoy, it has dimension approximately 90cm x 90cm x 310cm.

The cost of our low-cost underwater surveillance system is relatively cheap comparing to another monitoring approach. As described in Table 1 below, total cost of implementation is Rp 11.991.000 or about 900 USD.

Fig. 4 Case box dimension model

Table 1 Cost calculation

B. Experimental

We have commited an experiment to deploy our low cost water monitoring system in Lenggoksono bay, Malang. In this experiment we’re performing benchmark on Big Data server comparing to conventional (single) server, capture transmission delay and simple image analysis.

Fig. 5 Onsite experiment

Our first measurement was calculating image capture and saving time from Raspberry Pi to underwater camera using underwater rod with length of 3 meters. As designed before, we extend the Wi-Fi signal by using coaxial cable type RG 174/u. The experiment took 9 consecutive captures using time Unix command with our Node.js application and download captured image immediately.

Fig. 6 Delay of image fetching

Fig. 6 above shows that using coaxial cable to extends Wi-Fi signal resulting delay 5,6 seconds in average. For comparison, by using normal Wi-Fi in the air, it took 6,8 seconds in average to do same thing. In this experiment we use 4000px x 3000px resolution and the size of image may vary (about 4–6 Mb each).

We also measure the time of image transmission over 3G network to our cloud server. Our goal is comparing response time when we send data to single server than distributed Hadoop server. Computers we used as server were Dell Precision T1700 with Intel Xeon E3–1241 v3 processor, 8GB RAM and 1 Tb SATA hard drive. We simulate conventional single server by using one computer, and distributed server by using 1 master and 3 slave Hadoop nodes.

Fig. 7 Single and distributed server delay comparison

Results indicated in Fig. 7 was single server’s responses time was overall faster. While they need average time 53,6 seconds, Hadoop distributed server need 186,6 seconds to save our image. This is because Hadoop server must replicate received image into 3 slave nodes than save it into single server. Table 2 also indicate that scripts execution time in single server are nearly 0 second comparing to Hadoop distributed server that need 130 seconds. However, distributed architecture is still preferable to provide high availability, ability to process unstructured data and high scale data processing in the future [12].

Table 2 Script execution time

Further step is analyzing bleached coral by using color detection. We normalize retrieved image by using histogram autolevel equalization, applying RGB threshold to isolate bleaching color, and find bleaching percentage as seen on Fig 8. Our color thresholds are R >= 220, G >= 125 and B >= 0. To count whiteness percentage, we use equation (1) as written below

(1)

where Pxb is bleaching pixel and Px is image pixel. After applying the algorithm, we can detect 11,2% and 2,1% bleaching color from two image samples.

Fig 8. Image bleaching analysis

For the representation model, we have also have developed an online dashboard to presenting our retrieved images (Fig. 9). This model loads our coral images from distributed servers and put them into interactive maps. We intend to make our platform open so that everyone can learn and and get knowledge from it.

Fig. 9 Online representation model

V. Conclusion and Future Plan

In this paper we have developed a low cost coral monitoring system that consists of buoy mechanics as the base component, controller unit to control underwater camera, underwater camera to capture coral condition and GSM modem to transmit data to cloud server. The dimension of our system is 90cm x 90cm x 310cm with portable model so it can moved easily.

We also have overcome signal loss problem in underwater Wi-Fi camera with coaxial cable. The capture command was successfully executed and our program was able to download captured image with average time 5,6 seconds. In the server side, Hadoop distributed server took 186,6 seconds in average to retrieve and save image to 3 slave nodes, comparing to 53,6 seconds on single server. However, Hadoop distributed server provide high availability, ability to process unstructured data and high scale data processing.

In the future, we have a plan to utilize our system with 360 degree camera to capture wider area of corals and build more rigid Big Data architecture. Moreover, we also want to implement image processing analysis and classify bleaching level of retrieved image.

References

[1] Underwater Earth, “Global Coral Bleaching,” [Online]. Available: http://www.globalcoralbleaching.org/. [Accessed 7 June 2016].

[2] T. J. Goreau and R. L. Hayes, “Coral Bleaching and Ocean “Hot Spots”,” Ambio, vol. 3, no. 23, pp. 176–180, 1994.

[3] M. Chow, R. H. Tsang, E. K. Lam and P. A. Jr, “Quantifying the degree of coral bleaching using digital photographic technique,” Journal of Experimental Marine Biology and Ecology, vol. 479, pp. 60–68, 2016.

[4] J. Hedley, C. Roelfsema, B. Koetz and S. Phinn, “Capability of the Sentinel 2 mission for tropical coral reef mapping and coral bleaching detection,” Remote Sensing of Environment, vol. 120, pp. 145–155, 2016.

[5] J. D. Hedley, C. M. Roelfsema, I. Chollett, A. R. Harborne, A. R. Harborne, S. J. Weeks, W. J. Skirving, A. E. Strong, C. M. Eakin, T. R. L. Christensen, V. Ticzon, S. Bejarano and a. P. J. Mumby, “Remote Sensing of Coral Reefs for Monitoring and Management: A Review,” Remote Sensing, vol. 8, no. 59, 2016.

[6] S. F. Heron, G. Liu, J. L. Rauenzahn, T. R. L. Christensen, W. J. Skirving, T. F. R. Burgess, C. M. Eakin and J. A. Morgan, “Improvements to and continuity of operational global thermal stress monitoring for coral bleaching,” Journal of Operational Oceanography, vol. 7, no. 2, pp. 3–11, 2014.

[7] T. Suzuki, K. Kato, E. Makihara, T. Kobayashi, H. Kono, K. Sawai, K. Kawabata, F. Takemura, N. Isomura and H. Yamashiro, “Development of Underwater Monitoring Wireless Sensor Network to Support Coral Reef Observation,” International Journal of Distributed Sensor Networks, vol. 2014, 2014.

[8] M. C. Domingo, “An overview of the internet of underwater things,” Journal of Network and Computer Applications, vol. 35, no. 6, p. 1879–1890, 2012.

[9] D. Anguita, D. Brizzolara and G. Parodi, “Building an Underwater Wireless Sensor Network Based on Optical: Communication: Research Challenges and Current Results,” in The Third International Conference on Sensor Technologies and Applications, Athens, Glyfada, 2009.

[10] C. A. Pérez, H. N. Hellín, F. S. Valles, A. I. García and R. Torres, “Low Cost Buoy for Monitoring Recreational Areas,” in Sixth International Workshop on Marine Technology, Barcelona, 2015.

[11] C. Nils, V. Seregin, J. Gladchun and B. Chalstrom, goproh4: Node.js GoPro Hero 4 library, Zenodo, 2016.

[12] T. White, Hadoop: The Definitive Guide, California: O’Reilly, 2012.

[13] Y. Sun and T. Melodia, “The Internet Underwater: An IP-compatible Protocol Stack for Commercial Undersea Modems,” in Eighth ACM International Conference on Underwater Networks and Systems, New York.