Evaluation of marking techniques for a human-robot selective vineyard sprayer. Academic Article uri icon

abstract

  • A human-robot interface was designed for a target-specific pesticide robotic sprayer developed to reduce the amount of pesticides applied. This paper presents three techniques for marking targets in the remote interface. The assumption is that collaboration of a human operator with a robotic system can increase target detection rate and decrease false alarm rate. The human task is to mark the targets, defined as the grape clusters. Three techniques of target marking were developed and evaluated: (i), the operator marks the center of a constant diameter circle. The operator can mark multiple circles for each target. (ii), the operator marks a changeable size ellipse. The operator can mark multiple ellipses for each target. (iii), the operator free handedly marks the target contour. Experiments were conducted to evaluate the effect of the marking methods on the target detection HIT, FA (False Alarm) and MISS rate. 76 students participated in an experiment in which each student evaluates the three marking methods. Experimental results show that in order to maximize the target detection HIT rate, the best marking method is the constant diameter method which produced 94% HIT rate. The preferable marking method to minimize FA is marking the clusters by free hand.

publication date

  • January 1, 2012