Research On The Generation Of Panoramic Image Based On Fisheye Lens

Compared with traditional vision system, panoramic image technology could present the information around the camera completely, which has important significance on fields such as medical, industrial, and military. And the fisheye lens is one of the most effective ways to establish a panoramic vision system, which is equipped with the advantages such as wide angle, compact structure and high resolution. As a result, this article will make thorough research to the existing panoramic vision technology based on the fisheye lens. The major work and innovations of this paper are as follows:(1) Make a study on the panoramic vision system. This paper classifies the panoramic vision system and mainly elaborates the panoramic vision systems based on the fisheye lens, which consists of the physical structure, optical characteristics and the model of fisheye camera, laying a foundation for camera calibration and fisheye image correction.(2) Propose effective fisheye image unwrapping algorithms. In the process of expanding a fisheye image to a panoramic image, this article proposes an effective rotating and unwrapping algorithm with the angle of 30 to overcome the redundancy and drawbacks of unwrapping angle in existing algorithms. Beside, in order to avoid rotating image, we propose a new panoramic image unwrapping algorithm base on the polar coordinates. The results show that the polar unwrapping algorithm is simple and effective while good unwrapping results are obtained.(3) Propose a new fisheye image correction algorithm based on the cross ratio invariance and division model. To improve the result of fisheye image correction, this article improve the fisheye image calibration method based on the cross ratio invariance and propose a new fisheye image correction technology based on the division model. To fully reflect the effectiveness of this method, data analysis and data visualization are carried out. Compared with the polynomial model, this method has lower computation complexity and obtains better results in the process of calibration.(4) Two different kinds of scale correction algorithms are proposed to deal with the scenery imbalance in the unwrapping panoramic image. The first method directly stretches the unwrapping fisheye image non-linearly according to the division model that calibrated in advance; According to the characteristic that the checkerboards’ intervals equal to each other in radial, the second method adopts the improved optical transformation function and modifies the unwrapping panoramic image by iterative computation. The experiment results show that the proposed scale correction algorithms will reduce the pixel information loss and gain uniform scale in the panoramic image.