Human -aided construction of vision models for robotic bin picking

Yuichi Motai, Purdue University

Abstract

This thesis presents a human-computer interaction (HCI) framework for building vision models of 3D objects from their 2D images. Our framework is based on two guiding principles of HCI: (1) Provide the human with as much visual assistance as possible to help the human make a correct input; and (2) Verify each input provided by the human for its consistency with the inputs previously provided. For example, when stereo correspondence information is elicited from a human, his/her job is facilitated by superimposing epipolar lines on the images. Although that reduces the possibility of error in the human marked correspondences, such errors are not entirely eliminated because there can be multiple candidate points close together for complex objects. Each input provided by the human is therefore checked against the previous inputs by invoking situation-specific constraints. While a basic framework for human-assisted model construction uses fixed camera viewpoints for collecting the images, this dissertation also shows how this constraint can be relaxed and how a human can be allowed to suggest the next-best viewpoint to use after some images have been captured from fixed viewpoints. Allowing a human to suggest a viewpoint raises new issues in hand-eye calibration that this thesis also solves. Another contribution of this thesis is in providing a human with a menu of low-level operator for interactive feature extraction. The human can examine the output of the different operator and choose the one perceptually that is most appropriate to acquire the vision model.

Degree

Ph.D.

Advisors

Kak, Purdue University.

Subject Area

Computer science|Electrical engineering

COinS