Shape and Color Object Tracking for Real-Time Robotic Navigation

TL;DR

This paper introduces a real-time method for detecting and tracking single-colored balls using camera calibration, color classification, and circle estimation, suitable for robotic navigation amidst occlusions and clutter.

Contribution

The paper proposes a novel real-time approach combining offline calibration and online color segmentation for effective ball tracking in robotic navigation.

Findings

Good balance between real-time performance and robustness

Effective in occlusion and background clutter scenarios

Accurate circle parameter estimation

Abstract

This paper presents a real-time approach for single-colored ball detection and tracking. The approach consists of two main phases. In a first offline calibration phase, the intrinsic parameters of the camera and the radial distortion are estimated, and a classification of colors is learned from a sample image of colored balls. The second phase consists of four main steps: (1) color segmentation of the input image into several regions based on the offline classification, (2) robust estimation of the circle parameters (3) refinement of the circle parameters, and (4) ball tracking. The experimental results showed that the approach presents a good compromise between suitability for real-time navigation and robustness to occlusions, background congestion and colors interference in the scene.