Design Of A Panoramic Image Acquisition Platform Based On Multiple Cameras

Panoramic image expresses surrounding environment through paintings, photos, videos and 3D model in the form of wide-angle as much as possible. Compared with general images or 3D scenes, panoramic image provides much more virtual information from the physical or virtual space around the view point immediately, meantime, it brings the viewer a better visual experience. Panoramic image acquisition technology is playing an important role in the field of intelligent vehicles, intelligent transportation, and virtual reality. It is also a research focus in computer vision field. Therefore, it is necessary to study it systematically.Visual perception ability is a very important and efficient ability of intelligent vehicle environment perception, and it is also an essential ability. It can be used to detect traffic lights, traffic signs, parking lines and other basic traffic information which is the fundamental information for the control of intelligent vehicles. This thesis designed a panoramic image acquisition platform based on multiple cameras by analyzing the visual information demands of intelligent vehicles and studying the basic knowledge of panoramic image acquisition technology. The goal is to create a visual information access platform for intelligent vehicles and give them the ability of visual perception. According to the imaging model of the camera which was selected and the related knowledge of camera calibration, an accurate mathematical model of the platform was proposed, the error source of this model was analyzed, too. Meanwhile, the basic steps and key technologies of image mosaic are studied and a new method of image cylindrical projection model calculating which was proposed based on the experimental results of binocular stereo calibration. This model can calculate radius of the precise cylinder model and the radian of each image. Finally, based on the mathematical model of the accurate panoramic image acquisition platform and the image cylindrical projection model based on binocular stereo calibration, a new method for obtaining cylindrical panoramic images was proposed. The experimental results show that the mathematical model and algorithm proposed in this thesis can be used to implement the acquisition of panoramic image and the results are satisfactory, the platform can complete the task of panoramic image acquisition.