A LONG-TERM FIELD MONITORING SYSTEM WITH FIELD SERVERS AT A GRAPE FARM

Abstract

Monitoring image data at a field site provides important and useful information concerning plant and environmental conditions for agricultural users. We propose an image monitoring system that makes it possible to work in harsh areas, collect large amounts of data in real time, and analyze the data according to users’ requests. The system is constructed for use via the Internet with Web-based sensor nodes, an agent program, and Web analysis applications. Web-based sensor nodes such as Field Servers allow the monitoring of various types of sensor data, including image data, via the Internet. The agent program performs complicated operations on the sensor nodes from a remote site via the Internet; it can also analyze monitoring data easily using Web applications. To evaluate the effectiveness of the image monitoring system, we executed a long-term field experiment at a grape farm in Obuse-town, Nagano, Japan. Four types of sensor data concerning environmental condition and 2 types of image data concerning plant detail and field overview were collected and stored at a remote site via the Internet. The image data were analyzed using a filtering approach operated by the agent program and Web applications which were used for calculations in a background subtraction model. The proposed system has worked stably since May 2006 and provided results in which approximately 84% of unnecessary image data were filtered. Due to its construction with a distributed architecture based on the Internet, this system has the advantage that non-expert users can easily obtain the monitoring data on the Web and image processing can be provided for users without changing the system architecture.