REAL-TIME MULTI-CLASS OPERATOR ACTION RECOGNITION USING RANDOM FOREST MODEL

摘要

PROBLEM TO BE SOLVED: To provide an action recognition system for recognizing an action of an operator by classifying an image of the operator by using a random forest model.SOLUTION: A plurality of predictions are generated by using a random forest model. Each prediction is generated by one of a plurality of decision trees and includes a predicted operator action and a reliability score. The plurality of predictions are reorganized into a plurality of groups and the plurality of groups are respectively associated with one of the operator actions. The reliability scores are totalled within respective groups and total scores associated with respective groups are determined. The operator action associated with the highest total score is selected.