A real-time fiducial marker based on vision-based techniques to visualize the coexistense of the real and synthetic _3d breast cancer model in the identical real space

• Main Authors: Abdullah Bade, Siti Hasnah Tanalol, Rechard Lee, Ho, Wei Yong
• Format: Research Report
• Language: English
• Published: Universiti Malaysia Sabah 2014
• Online Access: https://eprints.ums.edu.my/id/eprint/24581/1/A%20real-time%20fiducial%20marker%20based%20on%20vision-based%20techniques%20to%20visualize%20the%20coexistense.pdf; https://eprints.ums.edu.my/id/eprint/24581/

Various studies have been conducted to explore the potential of Augmented Reality (AR) technology in mainstream application such as in medicine, visualization, maintenance, path planning, and education. With the rapid development in computing technology, AR technology can be used to aid the surgeon and the patient to have in depth visualization about their sickness in which human senses are not able to detect. An AR system combines the field of Computer Vision (CV) areas such as marker, markertess tracking and feature detection in order to fully utilize the AR potential. Although there exist methods for patient diagnosis through anatomic landmark or fiducial marker as the region of interest (ROI), this method, however, required an additional surgery in order to insert the fiducial marker to the patient It is not only time-consuming, but also invasive and might cause trauma to the patient. Thus, innovations in the current techniques for the insertion of the fiducial marker may improve its accuracy and reduce its risks, and enhancing comfort for the patient. Therefore, this project will investigate the feasibility of real­time markertess square-ROI recognition (RPMS) based on the integration of contour-corner approach as the fundamental component in registering the virtual imagery with its real object without the needs to use conventional marker. To enhance the conventional contour and corner approach, a smoothing and adaptive thresholding are performed to the captured input stream and then use subpixel corner detection to obtain better and accurate corner points. The RPMS technique starts with first getting an input from the real environment through a web camera. It will then, process the input, finds and detects strong interest points from the manually drawn Square-ROI. From the ROI, features such as number of corners and vertices will be extracted and later used to determine a marker. For testing purposes, two sets of experiments have been conducted to evaluate the RPMS technique. The first test evaluate the RPMS performance accuracy in identifying the hand-drawn ROI on an A4 size paper as a marker, followed by the used of a mannequin in the second experiment. For each experiment, the evaluation is repeated independently with eight different sizes of ROI, ranging from 3 x 3 to 10 x 10 cm with 1.0 mm border line's thickness. initially, in the experiments, the visual sensor (webcam) is positioned at 60 cm from the hand-drawn square-ROI in order to determine the best square-ROl's size and the optimal viewing distance. From the experiments, the best execution times obtained are 0.39 ms (A4 paper) and 0.81 ms (mannequin) with 6 x 6 cm as the best square-ROI size. It is found that for the size of 6 x 6 cm, the optimal viewing distance is from 7 cm to 23 an. In these experiments, it shows that the RPMS technique takes 10.09 ms to detect comers and 1.38 ms to detect the square-ROI. These indicate that, the RPMS technique is efficient, accurate and robust within the experiments environment and could be portable to any desired target area or domain.