Stabilizing Projected Points inAugmented Reality Applications
Independent thesis Advanced level (degree of Master (Two Years)), 20 credits / 30 HE creditsStudent thesis
Positioning projected points in a mobile Augmented Reality application based only on the orientation sensor of the device can cause instability due to variations in the sensor data. A method is proposed to use the image data from the camera feed of the device in order to stabilize the points. Point correspondences are found by matching features in subsequent frames, and a robust method is then used to determine a homography from these, which allows us to map between the two frames and thus correct for camera rotation.The implementation was evaluated on an Android smartphone and results indicate that this method can achieve close to real-time performance, taking on average 80 ms to produce a homography from a busyscene. The determined homographies can correct for camera rotations and produce only small errors in the positions. The choice of parameters is important to the results, and it should be noted that the method does not work for significant movement or low-detail situations.
Place, publisher, year, edition, pages
IdentifiersURN: urn:nbn:se:kth:diva-155913OAI: oai:DiVA.org:kth-155913DiVA: diva2:763359