Vision-based tasks and dynamic contours

Abstract

The major contribution of this work is the development of constant curvature dynamic contours. This new formulation was fast enough to track objects falling in the air and accurate enough to be used for visual servoing. The classical active contour model has two basic internal forces: tension and curvature. These forces are included to provide cohesion, equal control point spacing, and locally smooth shapes. These classical internal forces have undesirable attributes that are in conflict with the original design goals. Tension evenly spaces the control points, but also causes the models to collapse in weak image gradients. Curvature produces locally smooth curvature, but it does so by forcing the model toward a straight line. This thesis returns to the original active contour model motivations to reformulate these internal forces such that these desired properties are achieved without the introduction of unwanted model behavior. A new spacing force and a new constant change in curvature force are introduced and their performance characteristics are discussed. Experimental results that demonstrate the efficacy and performance of the proposed reformulations are included. These models are then applied to visual servoing, tracking, and grasping tasks. Several problems in robotics and computer vision do not have global optimal solutions. However, elegant and efficient algorithms can be designed to provide suboptimal solutions to some problems. Robotic grasping is one of these. We investigate constraints, procedures, and heuristics that allow partial working solution to robotic grasping. The proposed work attempts to find procedures and algorithms that allow an eye-in-hand system to grasp unknown (possibly moving) objects in a cluttered environment using the contour of the target. In order to complete this task, we use contour extraction methods and depth from motion techniques. Our grasping framework is implemented using a control system for the Puma 560 that employs custom-made hardware components. The proposed solutions are modular in order to be applicable to other manipulators and grippers.