电力监控全景图像拼接及接缝优化技术研究与实现

北京理工大学珠海学院2020届本科生毕业论文Research and Realization of Power Monitoring Panoramic ImageStitching and Seam Optimization TechnologyAbstractPower monitoring panoramic image splicing theory,technology and algorithm practice isthe current research hotspot in the field of image processing,it can well solve the currentindustrial field of large area monitoring needs of pain points and difficulties:in reality,due tothe limitations of imaging equipment,geographical location,viewpoint constraints and otherproblems lead to monitoring cannot directly capture the monitoring image of the large area,can only be manually observed manually from multiple locations of different display screens.Due to the large number of stations and scattered locations,the need for power monitoringpanoramic images,this study is mainly for the power industry's comprehensive understandingof the environment,multi-angle image acquisition of the environment around the transmissionline,and the use of image stitching algorithm to complete the image stitching,optimize theseams,make the seam transition smooth.The so-called power monitoring panorama image stitching is the relationship betweentwo or more partially overlapping photos in the scene matching each other.By sampling thefusion between images with a certain connection can obtain a new high-definition imagecontaining multiple image information,wide angle,complete,which is conducive toimproving the shortcomings of panoramic image acquisition equipment(such as wide-anglelens or fisheye lens)image edge distortion,to solve the problem of large shadows andunnatural image connection generated during the process of stitching and fusion area.In this paper,SIFT-based scale-invariant feature transformation matching algorithm isinvestigated in depth based on the power field images obtained by two surveillance camerasfrom different locations and perspectives.The traditional stitching technique is to select areference image,all other images are aligned with the reference image,fused into a panorama,the method due to the existence of accumulated errors will cause distortion and distortion ofthe stitched image,and the seam optimization between adjacent images is not ideal.The SIFTalgorithm mainly performs descriptor recognition locally,and the overall allowances betweenimages,except for the recognition point,are significantly different,suitable for processingimages generated by power monitoring.MATLAB software modeling for power monitoringimage depth identification,envelope image pre-processing,feature point extraction,detection,description and matching,image fusion and noise reduction to generate power monitoringpanorama,and mechanical stitching of the panorama after the seam (optical optimization)andother steps to provide solutions for power monitoring panorama view processingKeywords:Panoramic Image;Seam Optimization;SIFT Algorithm;Information visualization