Automatic detection and tracking of planar markers in a stereo-vision system

Purpose: The present paper deals with the development of self-identifying planar markers which are automatically detected, tracked and reconstructed in 3D in a stereo-vision system. Method: Firstly, for every acquired image, the connected domains are segmented and classified by applying an adaptive thresholding algorithm. Once visual markers are extracted, based on specific inner geometric rules, they are classified by using a planar homography and then matched with a marker library. Two commercial low-cost webcams, arranged in stereoscopic configuration, were adopted. Cameras are calibrated by using the Zhang’s planar calibration procedure. A fast and efficient implementation is proposed based on the automatic marker detection. Finally, 3D triangulation is performed on the pairs of captured images. Result: The proposed algorithms were compared with an open-source package for camera calibration. Results pointed-out a faster calibration procedure not influenced by manual user interaction, and a less sensitivity on the number of calibration points and images. The implemented GUI drives the user to easily perform camera calibration and marker tracking. Discussion & Conclusion: The adopted marker layout allows to recognise coded markers even in presence of high (till 70°) projective angles. The proposed scheme for marker detection permits to easily edit and extend the marker library. More investigation is needed to improve the algorithm efficiency to make it more robust against illumination conditions and marker occlusions.