Abstract
It is very important for robot to find objects, especially for a home service robot which works in human living environment. To find the environment contains many kinds of objects, we need the object recognition method using large dataset. Random Forest, developed by Leo Breiman[1], is an ensemble classifier composed of many decision trees . The method outputs the class that is the mode of the classes output by individual trees. Each trees can be calculated independently. So, the algorithm runs efficiently on large databases. Thousands of input variables can be handled without variable deletion. In this paper, we present a method for the object recognition by random forest using a large dataset of RGB-D images. And we evaluate its acuuracy and computationnal cost.
