Three-dimensional predictive object tracking system
Many applications require tracking of complex 3D objects. These include visual navigating of robotic arms on specific target objects, Augmented Reality systems that require real-time registration of the object to be augmented, and head tracking systems that sophisticated interfaces can use. Computer Vision offers solutions that are cheap, practical and non-invasive. This thesis proposes a new method for creating a high quality 3D Models as a novel method for 3D predictive object tracking. The thesis contributes in two important ways to the research area of environmental models acquisition. First we introduced a new pre-processing step in obtaining a 3D model. This step is called Distance Optimization. During this step we calculate the best range for scanning a 3D object, using multiple cameras and single projector, based on the intrinsic and extrinsic parameters of the system. This process helps us to always locate the best scanning range for any kind of camera which benefits us in many ways such as scanning time and 3D quality. The second contribution to this research area, was to introduce a novel real time feedback process for improving the quality of the 3D model obtained by the system. During this step we combined two different 3D modeling system Stereo Matching and Structured Light, and used a novel wavelet fusion technique called Regional Based Wavelet Fusion, to create a more accurate depth map using our system. To extract more details from our 3D images, we also used a sharpening step along with a point-cloud merging algorithm. This new approach for creating a high quality 3D model will enable us to use our technique in a real-time 3-dimensional predictive object tracking system. The contributions presented in this thesis have been fully implemented and empirically evaluated by comparing to some of the newest approaches in this research area.