In this work, we proposed and developed an approach for markerless object detection and registration in augmented reality. Our system enables the superimposition of videos or 3D graphics on natural objects in a real-time tracking process. The object of interest is detected within a sequence of images using the feature points and their invariant descriptors. The matching process between the test image and the training images is calculated using a 2D homography to generate the projective transformation for the video registration part. Further, the 3D graphic is overlaid into the scene by estimating the real camera pose and solving transformations relating the virtual and the real reference frames. The conducted experiments provided accurate and time effective results. Our approach detects and tracks markerless objects in real-time and enables video and 3D models registration for augmented reality experiences.
Read full abstract