Assembling A Whole Picture Of The Vehicle Against The Still Background

With the development of economy,the sales volume of automobiles in China is increasing year by year.In 2016,the total sales of automobiles reached 28.02 million,an increase of 9.74% year-on-year.However,as the number of vehicles increases,the number of traffic accidents on roads also increases year by year,and the transportation of contraband and other illegal and criminal activities carried out through vehicles also increase.Therefore,it is necessary to preserve the characteristics of automobiles on highways.Information provides evidence for combating criminal activities.In order to preserve the characteristics of cars of different sizes on the highway,a solution is proposed in this paper: a camera is set up at the entrance or exit of the highway toll station to continuously collect car pictures,and finally the image mosaic technology is used to generate the panorama of the car,save it to your computer.However,the difficulty in generating a car panorama through this method is that the image sequence to be stitched contains the same background and the object to be stitched is a sports car located on the background,that is,the target to be stitched is dynamically changed.The background is the same.Therefore,this dissertation focuses on the research of dynamic car image mosaic in the static background,and explores how to apply the image mosaic algorithm in static scene to moving vehicles in static background.The main research content includes the following aspects:1)A set of auto image stitching system is designed and implemented to solve the splicing problem caused by the indefinable position of the stitched target and background and the position of the overlapping area when stitching the moving target in a static background.2)For the problem that the target to be stitched and the background cannot be separated,this paper proposes two schemes based on the feature that the displacement of the moving target in the image to be stitched is greater than the background displacement.One is a filtering method based on the matching of background points of feature points,and the other is One is a background block filtering scheme based on template matching.3)According to the filtering method of background points based on feature point matching,a novel image mosaic algorithm based on feature point detection is proposed.The algorithm mainly includes steps of feature point extraction and matching,matching point purification,image registration and fusion.Through experiments,it is verified that the algorithm can generate car panoramas well when the image features are obvious.4)Aiming at the problem that the auto image stitching algorithm based on feature point detection cannot be successfully stitched when the image feature is not obvious,an image stitching algorithm based on image gray is proposed.The algorithm mainly includes image preprocessing,template matching,and Rough matching and fine-match vehicle displacement calculation,multi-resolution mosaicing and other steps,wherein a background block filtering scheme based on template matching is used in the rough matching stage to filter out the background blocks in the image,and the correlation measure in template matching Methods,an adaptive threshold distance metric algorithm is proposed.5)The experiment compares the splicing effect and performance of car image splicing algorithm based on feature point detection and car image splicing algorithm based on image gradation.