Home  >  Project  >  Details

Sensor based Farm Work Assistance


We are engaged in a project for support the agriculture using information technology. Our group takes notice of the farmers and developing the method to acquire the agricultural work information. Through the use of our sensing technology, we can get the information without imposing a strain on the farmers. The aim of our research is to model the experience of the proficient farmers by analyzing the chore information with the information about subject plant and environment.


The research to support the agriculture by using information technology is variously done. In the past, the aim of the support is to computerize the information of the documents such as the schedule and work sheet. Recently, the technical support system is being developed in order to improve the agriculture and food self-sufficiency.

The main matter of the technical support is to impart the agricultural skill efficiently. To realize it, it is necessary that the system records and visualizes the step of the task. The important thing is where the farmer is working. We are researching about it.

GPS cannot accurate position of the farmer because of the barrier in a greenhouse. We can measure the position of the farmers without particular equipment by using camera. It is a reasonable solution because camera is also used to observe the status in the greenhouse.

The measurement of the farmer is basically the problem that two dimensional measurement in the agricultural field. Therefore, we first perform the mapping of the image plane and farm field plane, and then we estimate the position in the farm field using the mapping. In view of this, the accuracy of the estimated position depends on the position on the image. The detection method needs to robust to the illumination variation because the greenhouse is not completely interior and the illumination variation is large. Meanwhile, the agricultural works are generally done in various pose and farmers use a wide variety of tools. Therefore, we use the illumination invariant background modeling based object detection method which requires no information about shape of the subject.