Robust Hand-Drawn Square-Roi Contour Detector Based on Adaptive Thresholding

Abstract

Hand-drawn square-ROI detector was developed as one of the vital components in Real-Time Pre-Placed Markerless Square-ROI (RPMS) recognition technique. It aims to; 1. To verify hand-drawn Square-ROI (Region of Interest) as a square, and 2. To create a robust and flexible square-ROI detector technique which can be applied in uneven lighting condition. In this paper, we aim to detect only the desired ROI and handle the uneven lighting condition which is one of the primary disturbance sources that may generate false results. This may lead to error in registration in Augmented Reality application due to inability to correctly define a marker. As a solution, our technique applies adaptive thresholding in order to address this issue and to create a robust and flexible technique. To verify our proposed technique, two kinds of square is used in the testing and evaluation phase. In this experiment, two influencing factors; viewing distance, and detection accuracy were used to validate our aim. The results of the experiments show that the proposed technique efficiently detects and defines the desired square-ROI and also robust to illumination changes.