We developed a method of human detection in a paddy field in order to safely operate a robot combine harvester. Images with and without human in a paddy field were used to develop an image processing algorithm. Previously we developed a texture feature method to detect human and its location in an image. However, the result was not so accurate, particularly in the background of the image or amongst lodging rice plants. Therefore, a new color feature method was developed to improve the precision in this study. When both the color feature and texture feature methods are used in combination, the precision was improved significantly.

The behavior of a tractor running on a slope can be expressed by four degrees of freedom, namely, vertical, pitching, rolling, and forward. In this study, we simulated the behavior of a tractor equipped with a swing mechanism on the front axle via dynamic equations. The effect of supporting conditions on tractor behavior was studied. Furthermore, tractor behavior on a slope in the event of a fall accident was simulated. The influence of change in the center of mass on anteroposterior fall was considered. The results showed that the tractor behavior depends on reaction forces of each wheels which no equipped swing mechanism, and the behavior of tractor equipped swing mechanism depends on rear axle angle. In addition, it was found that the position of the center of mass must be located on wheelbase at a distance of 32-65% from the rear axle to prevent fall accidents; the most suitable value is 49%.

A subject was employed for mowing slopes. An electromyogram (EMG) of the shoulders, forearms, and thighs was recorded. Work strain of the shoulders, forearms, and thighs was estimated from the integrated EMG data and mowing area. Difference between the influence of work strain of a brush cutter and that of a mower was investigated. And the machines requiring less work strain were revealed. The results indicated that during cutting, work strain using the mower was 47%, 66%, and 58% less than work strain using the brush cutter for the shoulders, forearms, and thighs, respectively. During lost time, work strain using the brush cutter was 53%, 42%, and 13% less than work strain using the mower for the shoulders, forearms, and thighs, respectively. Work strain using the mower for the overall duration of mowing slopes was 30%, 56%, and 50% less than work strain using the brush cutter for the shoulders, forearms, and thighs, respectively. Several slopes having different widths and lengths were assumed. It was observed that machines requiring less work strain varied depending on the width and length of the slopes.

Self-propelled small farm implements (rotavator, weeder, ridger, etc.) can play important roles in small-scale farming in developing countries, and their efficacy, or quality of operation should be commensurate with power consumption. Nonetheless, measurement of the power of small machines is difficult due to limited space for mounting torque-sensing devices. In this study, a simple and direct sensing of engine reaction torque was devised for evaluating the power consumption of a walking-type power weeder in paddy fields without the need for installing slip rings. The fitting between the housing of engine and the main body of the weeder was modified allowing the engine to rotate freely around its own center shaft, and a cantilever load cell was fixed between them to convert the force into engine reaction torque. Dynamic calibration was carried out in the laboratory to determine the relation between the engine reaction torque measured and the rotor shaft torque applied stepwise; the relation was linear with a coefficient of determination (R-square) of 0.98. Random and incessant setting of the rotor shaft torque resulted in a root-mean-squared relative error of 4.2% of estimation based on the calibrated relation. By multiplying the rotational speed with the rotor shaft torque, the power requirement of the weeder was accurately and continuously recorded in actual wet field conditions.

In this study, we attempted to reduce handle vibration in a brush cutter by unbalancing the centrifugal clutch weight. The individual contributions to handle vibration of the up-down and left-right vibration at the clutch cover were first analyzed. The results showed the largest contribution was from the left-right vibration. We therefore tried to reduce the left-right vibration by unbalancing the clutch weight. A small weight was placed on one side of two clutch shoes to reduce the input force along the left-right direction. The first-order left-right vibration at the clutch cover was found to be strongly reduced, with only a slight increase in the up-down vibration. The results indicated that handle vibration could be reduced by about 10dB using the unbalanced clutch.