Object Recognition based on RGB pictures is a widely practiced field. There have been many works regarding image processing, machine vision and image or object retrievals.
Abstract
Object Recognition based on RGB pictures is a widely practiced field. There have been many works regarding image processing, machine vision and image or object retrievals. Depth cameras, which turned popular the last few years, with the introduction of the Kinect camera for the Xbox console, offers the use of an object’s spatial geometry, and introduces new and interesting methods for object retrieval and classification. In this project we test and assess the Kinect’s capabilities for the purpose of object classification, trying to capture and classify objects belonging to the classes of Bottles, Bowls and Boxes. We have built a data set of 3D point clouds, designed a set of features based on geometric and statistical attributes of the objects in question and trained a classifier able to differentiate between the three classes, based on the features we calculated.
Flowchart
Results
original image
depth photography
