Research On The Methods Of Image Panoramic Stitching

Panorama stitching technology has developed rapidly in recent years and has been applied to various fields.For example,in order to attract customers,hotel accommodations use panorama-splicing technology to provide room panoramas so that users can intuitively experience the accommodation environment.Some places where traffic is circulated in large numbers also monitor security more comprehensively through panoramic splicing technology.When people go out to travel and encounter favorite scenes,they can also record all the scenes they see.In addition,panoramic splicing technology also has a wide range of applications in panoramic maps,tourist attractions,VR film and television,VR photography,panoramic video and other fields.In order to achieve a more robust,fast,and natural panorama stitching effect,this article needs to solve two problems:(1)achieve a seamless and natural panorama stitching quality(2)balance stitching quality,stitching speed,memory occupancy and other factors,so that the method achieves a usable effect.Therefore,the main work of this paper is focused on the following two research points:(1)Seamless natural panorama stitching method based on super pixelAlthough panoramic mosaics have matured,there are still some difficult problems that have not been solved,such as blurring caused by parallax problems and ghosting caused by moving objects.In addition,in order to achieve the application effect,there are also high requirements on the time and performance of the splicing method.This article aims at the above issues and optimizes the panorama stitching method.Super pixel segmentation is first performed,feature detection is combined with super pixel segmentation results and ORB feature points,and a matching strategy is improved to improve the speed and quality of feature detection and matching.In addition,the sutures and the image fusion method are optimized,which makes the splicing method robust to parallax,blur,and exposure problems.(2)720 panorama stitching method based on fisheye imagesOn the basis of 360-degree panorama stitching,in order to use less images to complete the 720-degree panorama stitching including the upper and lower perspectives,this paper uses images captured by a fisheye lens to stitch images.The quality of the fisheye lens-imaging model directly affects the image expansion effect,and then determines the quality of the mosaic.Based on the traditional fisheye lens imaging projection model,this paper proposes a projection model with parameters,and refers to the Mercator projection principle to reduce the distortion of the left and right parts based on the latitude and longitude correction..Under the condition that the relationship between the actual angle of the fisheye lens and the latitude and longitude of the image is unknown,the model proposed in this paper can obtain a better fitting correction effect without requiring the user to manually correct it.In addition,this paper improves the 720-degree panorama stitching method,adopts the strategy of first detecting feature points and post-model mapping,solves the problem of feature point loss,and effectively improves the matching rate of feature points.Finally,modify the warp mapping changes,reduce the error of the second calculation,and finally get the mosaic results.After the experiment and analysis,the method of panorama stitching used in this paper can obtain high-efficiency and high-quality panoramic stitching effect.