SMART FIELD MONITORING WITH FIELD SERVERS BASED ON ICT Tokihiro Fukatsu 1 1 National Agriculture and Food Research Organization, Institute of Agricultural Machinery, Ibaraki, Japan e-mail: fukatsu@affrc.go.jp ABSTRACT In modern agriculture, it is important to collect various kinds of agricultural information such as crop conditions, environmental data, and farm operational records. To collect agricultural information effectively, I proposed a Webbased sensor network system for agricultural use and developed Field Servers with a Wi-Fi/4G module for broadband networking, a main computer unit with a Web server for interactive monitoring, and various sensors and cameras for performing detailed measurements. Field Servers can collect not only field environment data by sensors but also crop condition data and farm operation data by cameras. To reduce the cost and effort for Field Servers, I have developed Open Field Servers that are redesigned to create a useful and widely used field monitoring system for common users. To manage Field Server system effectively without effort, I have also developed support tools and system for data analysis, maintenance, and trouble shooting. I have deployed many Field Servers in different locations of agricultural fields, and conducted various kinds of experiments of agricultural applications such as environment monitoring for growth prediction, insect counting for pest damage forecasting, vegetation cover rate calculation for crop growth understanding, and others. Field Server system is expected to be applied in various fields, and I introduced new challenges to utilize the ICT-based field monitoring system in agricultural big data and movable field monitoring. Keywords: Field Monitoring, Sensor Network, Image Analysis, ICT, Open Hardware INTRODUCTION In modern agriculture, it is required not only to increase agricultural productivity but also to promote efficient management, perform eco-friendly farming, and avoid the risk of climate change. To solve these issues, it is important to collect various kinds of agricultural information such as crop conditions, environmental data, and farm operational records. One of the key technologies in field monitoring is a sensor network system (Akyildiz et al. 2002) which collects wide range of data easily by many sensor nodes with a small sensor unit and a wireless communication unit. To collect agricultural information effectively with the technology of the sensor network system, I proposed a Web-based sensor network system for agricultural use (Fukatsu and Hirafuji 2005) and developed Field Servers (Fig. 1) in 2001. Field Servers have a Wi-Fi/4G module for broadband networking, a main computer unit with a Web server for interactive monitoring, and various sensors and cameras for performing detailed measurements. Field Servers are controlled by a smart management program at a remote site named Agent System which collects monitoring data, analyzes the collected data, and provides useful information to users automatically (Fukatsu et al. 2006). Users can collect agricultural information easily via the Internet only deploying Field Servers. Field Servers and related ICT monitoring system were developed and commercialized in the late 2000s. I deployed many Field Servers in various kinds of agricultural fields and conducted some experiments of agricultural applications such as environment monitoring for growth prediction, insect counting for pest damage forecasting, vegetation cover rate calculation for crop growth understanding, and others (Fukatsu and Hirafuji 2014). Through the experiments, I have obtained some knowledge to perform field monitoring effectively and cleared new challenges to utilize Field Servers. In this paper, I describe an ICT-based field monitoring approach using Field Servers to facilitate practical agricultural applications, and some ideas to apply the monitoring system to further research. 1
Fig. 1. Field Server system. FIELD SERVER Farmers want to know useful information to support their works, and field monitoring system is required to collect various kinds of data according to their situations. For example, it is important to monitor field and crop-growth conditions in real time so that farmers can effectively judge and schedule farm operations such as irrigation control, pesticide application, and harvesting. Thanks to agricultural researchers, many simulation models such as growth prediction, disease and pest damage forecasting, and crop yield estimation are now available (Tanaka et al. 2011). With the models and field environmental data such as air temperature, rainfall, and solar radiation, farmers can predict crop conditions in their target fields. It is also important to record farming operations. Data on what operations are performed, where, when, how, and by whom, are used to manage labor, improve farming practices, and optimize field use. To meet these requirements, it is important for field monitoring system to collect various kinds of sensor data including image data in real time. Recently, several measurement devices are commercialized, but they are expensive, collect only environmental information such as weather conditions, and require much effort to modify according to users situations. On the other hand, the main computer of the Field Servers has some analog-digital converters, some digital ports such as I2C, a pulse counter, a DDS circuit, and others to connect various sensors easily. Some digital single lens reflex cameras can be connected to Field Servers like a network camera so that we can easily perform high-resolution image monitoring. Therefore, Field Servers can monitor not only the field environment but also crop conditions and farm operations with image data. Moreover, to reduce the cost of Field Servers and the effort of users, I have developed a modified Field Server named Open Field Server. The Open Field Server is redesigned for common users so that they can easily obtain, customize, deploy, operate, and utilize it without specialized experience to create a useful and widely used field monitoring system. It is constructed by open-source technology and easily-obtainable parts at low cost. All of its contents such as its design documents, assembly manuals, operation tips and tutorials are openly available on a portal Web site (Fig. 2) so that many users can easily get and use it. Internet free services such as Twitter and Flickr are actively used in the data collection system. Every user can freely use, modify and share technology, tools, information and experiences to effectively support the monitoring performed by all users. To manage field monitoring system effectively, we must consider not only hardware devices including Open Field Servers but also software tools for data analysis, maintenance, and trouble shooting. In the Field Server system, the Agent System has function to analyze collected data with external applications on the Web. That s because many types of image analysis methods have been developed for agricultural image monitoring, but it is inefficient to set all image analysis methods in the Agent System. So, the Agent System sends collected image data to the image analysis services such as binarization, detection, and filtering on the Web, and get the analyzed results via the Internet (Fig. 3). The support tools are also important to reduce management effort. For a long-term monitoring with many measurement devices, it is important to detect wrong devices immediately without effort. Our system has warning function which sends e-mail to users automatically when some devices become wrong. The main troubles of data loss are hardware troubles such as the power cable dropping out, local-site network troubles 2
depending on a change in the network environment and heavy traffic on the 3G/4G network service, and software trouble caused by auto-updating of the PC or temporary stoppage of external Web services. Fig. 2. Portal Web site of the Open Field Server. Fig. 3. Image analysis with external Web services. AGRICULTURAL APPLICATION I have deployed many Field Servers in different locations of agricultural fields, and have conducted various kinds of experiments of agricultural applications with the collected data. Here, I describe some features and points of the applications using Field Servers from these representative experimental cases (Fig. 4). First, I predicted the date of heading and blooming stage of wheat with a growth model (Kanzaki 2009). The wheat growth model needs a young panicle length as an initial parameter and an accumulated daily mean temperature. Thanks to the accurate models and measurement data from Field Servers, I got good results of the estimation day error within one day. Second, I conducted the vegetation cover rate calculations and the flower detection of rice in paddy fields based on the collected image data. The image data collected in agricultural field is easily influenced by the noise of environmental changes, which interfere with accurate analysis of image data. To solve this problem, we need a robust image analysis method to extract target crops from the image data taken under natural light conditions where the images can contain shadowed and lighted parts with reflected parts of plants (Guo et al. 2015). Using Field Server system with a high-resolution camera and a robust image-analysis method, I successfully got accurate results. 3
Third, I conducted an insect-counting application with the Field Server system. It is also difficult to collect clear and stable image data at field sites. In this system, I have achieved an effective insect-counting system with a synthetic attractant on a non-sticky white board and simple image analysis method based on the background difference technique (Fukatsu et al. 2012). I can estimate the number of target insects in the image data with high accuracy, and this method worked well to reduce the labor needed for counting and easily handled a large mass of image data. Fourth, I conducted an automatic farming record system with Field Server image data for a long term. Image data of farm scenery that are simply collected at fixed points during periodic intervals provide useful information of farm operations. I have developed a method that automatically performs event detection of farm operations with agricultural machinery by analyzing image data. With the specialized data viewer for Field Servers, users can also browse multi-year image data of farm information at one time so that they can easily and visibly compare the timing of farm operations and growing conditions of plants with the data of a reference year (Toda et al. 2013). Fig. 4. Agricultural application using Field Server. DISCUSSION AND FUTURE WORK ICT-based field monitoring system is becoming popular and many kinds of agricultural applications have been developed. These technologies are expected to be applied in various fields. Recently, the use of artificial intelligence and big data analysis is becoming increasingly popular as a key method of solving agricultural problems. Data mining based on machine learning and a nonlinear model provides an opportunity to make breakthroughs in cropping, breeding, and agricultural sciences using agricultural big data. However, it is very important but difficult to collect large amounts of crop growth information. To construct agricultural big data, Field Server system is indispensable to effectively collect training data of sufficient quality and quantity. I am currently working on a fundamental research project on agricultural big data in Japan to address the above challenges. In this project, I attempt to solve the problem of data collection with Field Servers. Movable monitoring robots are also expected to the next stage of field monitoring. For example, users expressed a desire to collect data both more closely and widely in their crop fields, but deploying many Field Servers would be too costly and need much effort. Field Servers deployed within or close to fields also involve the risk of disturbing farming operations. To address these issues, I also propose a mobile robotic Field Server with a locomotion unit for moving in target fields and a manipulator unit for flexibly measuring targets. The robotic Field Server is expected to be used for periodical sampling and long-term monitoring in agricultural big data. 4
REFERENCES Akyildiz, I. F., W. Su, Y. Sankarasubramaniam, and E. Cayirci. 2002. Wireless Sensor Networks: a Survey. Computer Networks 38:393-422. Fukatsu, T. and M. Hirafuji. 2005. Field Monitoring Using Sensor-Nodes with a Web Server. Journal of Robotics and Mechatronics 17(2):164-172. Fukatsu, T., M. Hirafuji, and T. Kiura. 2006. An Agent System for Operating Web-based Sensor Nodes via the Internet. Journal of Robotics and Mechatronics 18(2):186-194. Fukatsu, T., T. Watanabe, H. Hu, H. Yoichi, and M. Hirafuji. 2012. Field monitoring support system for the occurrence of Leptocorisa chinensis Dallas (Hemiptera: Alydidae) using synthetic attractants, Field Servers, and image analysis. Computers and Electronics in Agriculture 80(2012):8-16. Fukatsu, T. and M. Hirafuji. 2014. Web-based sensor network system Field Servers for practical agricultural applications. Proc. of the 2014 International Workshop on Web Intelligence and Smart Sensing doi:10.1145/2637064.2637088. Guo, W., T. Fukatsu, and S. Ninomiya. 2015 Automated characterization of flowering dynamics in rice using fieldacquired time-series RGB images. Plant Methods 11(7) doi:10.1186/s13007-015-0047-9. Kanzaki, M. 2009. Estimation of heading time by young panicle length of wheat. Japanese Journal of Crop Science 78(2):74-75. (In Japanese). Tanaka, K., T. Kiura, M. Sugimura, S. Ninomiya, and M. Mizoguchi. 2011. Tool for Predicting the Possibility of Rice Cultivation Using SIMRIW. Agricultural Information Research 20(1):1-12. (In Japanese). Toda, S., K. Kobayashi, Y. Saito, T. Fukatsu, T. Kiura, and M. Hirafuji. 2013. Know-Live: A Farm Information Web Disclosure System with Subjective Information. Agricultural Information Research 22(1):12-23. (In Japanese). 5