Generation Technology, Based On Graphics And Image Of The Virtual Driving View

In this dissertation, the vision synthesis technology has been studied for the remote driving of Unmanned Ground Vehicle (UGV). Used of Virtual Reality (VR) technology, a new vision synthesis technology has been proposed, which is based on the combination of real images and 3D graphics. The data structure and the rendering method of 3D vehicle model has been studied, and an new method of computing vertex normal vector has been proposed, which is based on the polygon neighborhood with similar vertex normal. The experiment results show that the new method provides more realistic display. To provide realistic vision on the environment outside of the vehicle, Image based Rendering (IBR) methods are studied, and the method based on epipolar geometry is especially discussed. With the observation that the original method may produce gaps in the predicted image, a new hybrid algorithm based on the combination of forward-mapping and backward-mapping, has been proposed to improve the rendering effect. Since the stereo correspondence is the key for the IBR algorithm, this dissertation compares the matching results among several typical measurement factor, and presents a new stereo matching algorithm based on the combination of box-filtering and image transformation, the experiment results show that the new algorithm perform with more robustness in the presence of luminance diversity between the two images.