نوع مقاله : مقاله پژوهشی
عنوان مقاله English
نویسندگان English
Augmented reality algorithms use Simultaneous Localization and Mapping (SLAM) techniques to determine the position of virtual object points in each frame. In these methods, the intrinsic and extrinsic camera parameters are first estimated, and then the position of the virtual object in each frame is determined using these parameters. These methods require prior knowledge of the camera’s intrinsic parameters or assume that the intrinsic parameters remain constant throughout the video. In our previously proposed method, there is no need to estimate the camera parameters, and the position of the virtual object is directly determined. To achieve this, the proposed algorithm first calculates the fundamental matrix between keyframes and the current frame, and then determines the position of virtual points in each frame using the intersection of epipolar lines. The calculation of the fundamental matrix is prone to errors due to the presence of outliers. Using techniques such as RANSAC to remove outliers increases the computational time. In this paper, which focuses on improving the speed of the proposed algorithm, we eliminate outliers using data from the previous frame and present a fast method for estimating the fundamental matrix in each frame. The proposed method for computing the fundamental matrix made our algorithm up to five times faster than the previous one.
کلیدواژهها English
