Algorithms Research On Virtual Tour And Moving Object Detect In Fisheye Image

The perspective of fish-lens is wide, which can be180°even220°. Since the devise of the fisheye lens, the range of the application of it is more and more widely and the two important aspects of which are virtual tour and video surveillance. Currently, virtual tour has an important application in tourism, construction, and aerospace. Using a single fisheye image can achieve virtual tour semi-space, without using stitching technology. Fisheye lens is also applied to video surveillance to achieve the "blind spot" monitoring. This paper is aimed to study them.The implementation of virtual tour used a single fisheye image has three steps, first contour extraction, second re-projection and the last is the simulation of the camera. A lot of research on fisheye image contour extraction are detailed, first introduced the traditional methods of it, which are area method and scanning method, and their limitations are pointed out. Then the method based on region growing is improved, and it works better verified by experiments. And next re-projection based on fisheye image and the simulation of camera is detailed, at last achieved the simulation of virtual tour based on a single fisheye image.Moving object detection is an important aspect of video surveillance, which this paper studies a lot. Firstly, the optical flow method, frame difference method and background subtraction method for moving object detection are presented briefly. Secondly the algorithms of moving object detection based on background model are detailed, and make the method that based on the Gaussian mixture model as the basic method of moving object detection of this paper. But the method that based on the Gaussian mixture model has its shortcomings: running less effectively and no suppression of the shadows. Then using three frame difference method to improve the efficiency of the Gaussian mixture model method. Finally using shadow detection algorithm based on color feature processed detection results. And it works better proved by experiments.In the process of moving object detection toward to the video captured by fish-eye lens, it is necessary to correct fisheye image in window for the reason that there is a big fisheye distortion of the image. And the correct for fish-eye image in window is the significant association with virtual tour, and the paper achieves it at last.