To feed future world population, cost effective cultivation methods, water resource utilization, breeding new crop variety etc. are required. Creating Big Data in Agriculture and analyzing it are poetically useful for. Uncordially, national/international standards are not enough defied in Agriculture. Therefore, creating big data is difficult. In this session, we would like exchange experiences of member, work in this field, and discuss how to create big data in Agriculture trough APAN.
Time: 11:00 - 12:30
- Big data needs to be open and interoperable for being useful - the case for GODAN - 2016-08_APAN_GODAN.pdf
Johannes Keizer, GODAN & FAO
[Johannes.Keizer AT fao.org]
Science undergoes paradigm changes. Once science questions arose from empirical observations. Then it shifted to questions derived from models and hypothesis. Now datasets themselves are becoming the source of new scientific questions. Traditionally effort in a research project was 80% data production and 20% data analysis, but now these two proportions are inverted. Therefore access to datasets from observations and experiments becomes more and more a prerequisite for successful and effective research.
Once big data was a topic found only in space astronomy or high energy physics. Through remote sensing, satellites and drones, big data has found its way into Earth Observation. New technologies for analyzing genoms and relating genetics to phaenotypes are rapidly creating big data in biology.
Data need to be FAIR, findable, accessible, interoperable and reusable. This is the challenge. A set of agreements and common standards is necessary to produce FAIR data.
GODAN (Global Open Data in Agriculture and Nutrition -http://www.godan.net ) is an initiative that has been set up to promote open data in the area of Agriculture and Nutrition. The presentation will analyze the meaning of open data and the different shades of openness that we encounter in reality. It will outline what GODAN does to develop the capacity of institutions that want to make their data open.
Johannes Keizer is GODAN Secretariat, Strategic Partnerships Lead, Partnerships, Advocacy and Capacity Development Division (OPC), Food and Agriculture Organization of the United Nations (FAO) Team Leader. He is working for FAO since 1998 and improving AGRIS, AGROVOC, etc. He is leading standards and Open Data relating Agriculture in the world.
- Development of High-Precision 3D Measurement On Agriculture Using Multiple UAVs (Repeat of research workshop) - APAN42_MuhammadHaris-Agriculture_Workgroups.pdf
Muhammad Haris, Seita Sukisaki, Ryo Shimomura, Zhang Heming, Li Hongyang, Hajime Nobuhara, University of Tsukuba
[nobuhara AT iit.tsukuba.ac.jp]
Imaging system for high-precision 3D map on agriculture using UAVs was developed. The system were based on safe and easy UAVs with a ground station application which designed to be the interface between a human operator and the UAVs to carry out mission planning, flight command activation, and real-time flight monitoring. Based on the navigation data, and the way-points generated by the ground station, the UAVs could be automatically navigated to the desired waypoints and hover around each waypoint to collect field image data. By taking only low-resolution image, the proposed system is able to reduce the payload and increase the flight time of the UAVs. The input images then transform into higher-resolution image using reference images, taken by field server or ground-based device, via super-resolution techniques which is able to reduce blurring, blocking, and ringing artifacts especially in edge areas. Finally, we construct high-precision 3D map which proven having error of a millimeter order of magnitude. Our experiment result show that the input low-resolution can be transform into high-resolution image and effective to construct high-precision 3D map. The result indicate that the proposed system provides a reliable method of sensing agricultural field with high-precision 3D map.
Muhammad Haris received the B.Sc. in Computer Science from the University of Indonesia, Depok, Indonesia, in 2009. He is currently a doctoral student in Department of Intelligent Interaction and Technologies, University of Tsukuba, Japan. His research interests include image processing, super-resolution, image enhancement, and 3D reconstruction especially for aerial images.
- EasyPCC: Tools and dataset for high-throughput measurement of canopy coverage to natural outdoor environment for multiple crops - APAN42WeiGUOshare.pdf
Wei Guo, Bangyou Zheng, Tokihiro Fukatsu,Scott Chapman, Seishi Ninomiya, University of Tokyo, CSIRO, NATO
[guowei.net AT hotmail.com]
Image analysis techniques now are considered as a powerful tool for use in plant phenomics. However, the most useful phenotypic information about crops that are planted in fields is still being obtained by manual sampling that is not ideal because it destroys plants in the process, and it is extremely labor intensive and thus time consuming. The reason so much time is required is that for a given object/region under field conditions, images acquired with digital photography include a wide variety of light intensities, so their analysis involves careful, individual treatment that demands specialized knowledge of observers. Understanding crop biology under field conditions is extremely important, yet it has been impractical for the above reasons, because most phenotyping studies require dealing with very large crop populations. To address those problems, we proposed an effective and efficient tool to calculate plant canopy coverage rate from images taken under diverse light conditions in the field. Moreover, the source code package( Matlab, R, Python) and the annotated image datasets of different crops taken in field condition, which new algorithms and tools can be evaluated will be open freely for research use.
Wei Guo now works as a Project Assistant Professor at The University of Tokyo, Japan.
His Current Areas of Research are Field-based high-Throughput phenotyping using advanced imaging techniques (drone remote sensing and ground based observation), image analysis and machine learning approaches.
- Ground monitoring with multiple Field Server system in AgriBigData project - APAN42-AgWG-Fukatsu.pdf
Tokihiro Fukatsu, NARO
[fukatsu AT affrc.go.jp]
It is important for smart agriculture to collect field data. Field Server system with a high-resolution camera module can collect a large amount of image data for constructing agricultural big data (AgriBigData) to extract useful information. However, there are some bottlenecks to manage the system in practical fields. Now we have deployed the system including a new robotic Field Server in AgriBigData project field. Through the experience of the experiment, we clear up next issues to be solved for constructing a stable data collection in agricultural field.
Tokihiro Fukatsu has been working at National Agriculture and Food Research Organization. He received a doctorate in advanced agricultural technology & sciences. His main research is agricultural field monitoring with sensor network, image data, and robot system. He has developed Field Server, image monitoring system and robotic Field Server. Wearable system, agent system and robotics & mechatronics is also his interest research.