Research On Panoramic Image Synthesis Based On Binocular Fisheye Lens

Panoramic vision is a very hot technical concept at present,not only widely used in the field of virtual reality,but also can affect all aspects of our lives.For example,Google Street View refreshes people's perception of the map through panoramic vision.In addition,panoramic vision also plays an increasingly important role in areas such as industrial simulation and medical imaging.The prerequisite for achieving full-scale vision is to obtain panoramic images that contain all the surrounding scene information.Such images generally need to be photographed through expensive professional equipment,and can be obtained after complicated post-processing,which is not conducive to the use and dissemination of ordinary people.Therefore,it is a hot topic to obtain panoramic images more conveniently in the field of panoramic vision.Based on this reality,we put forward an improved method for the panorama generation of binocular fisheye lens in this paper.Fisheye lens is characterized by an extra large wide angle,and the field of view captured by some lenses can even reach 180°.This provides the possibility of effectively reducing the number of original images in the synthesis of panoramic images.However,the distortion of the original fisheye image is too large to stitch directly.We need to process the original fisheye image to be able to complete the synthesis of the panorama image.Correction methods generally include camera calibration techniques and projection model analysis.Camera calibration method is the classic way to deal with lens distortion,often applied in the adjustment process in the professional camera.Based on the imaging principle of the fisheye lens,the projection model method completes the mapping of the fisheye image to the plane coordinates and is the basis of the subsequent panoramic image synthesis algorithm.Binocular fisheye image panorama stitching is completed on the basis of single image correction.Firstly,a reasonable coincident area is cut from the corrected image as a part of the splicing.Then use feature point registration and gradual fusion to splicing this part.Finally,a fast fusion method based on weighted model is proposed.To obtain a complete fast stitching algorithm of binocular fisheye image is not the end of the study.Since this method improves the splicing efficiency at the expense of the splicing effect,an optimization method for improving the splicing effect should be found.According to the imaging characteristics of the image to be stitched area,combined with the feature point algorithm in the traditional image registration field,using the optimization method based on the relative error of the feature point can eliminate the weight phenomenon of the stitched part to a certain extent.According to the angle change between the fisheye lens and the concept of the attitude angle in the spatial analytic geometry,an optimization method based on relative attitude change between the lenses can effectively improve the problem of misalignment of the object due to the lens angle deviation.Finally,according to the principle that parallax will be generated between the two cameras,the use of an optimization method based on depth of field information can improve the phenomenon of picture quality degradation caused by parallax.The rapid synthesis method and optimization method of binocular fisheye images can improve the existing panoramic image synthesis method to a certain extent,and achieve the purpose of obtaining panoramic images more conveniently and more quickly.