Research On Fast Stitching Algorithm For Aerial Panoramic Images Of Fields Based On SIFT Feature Points

With the continuous development and popularity of UAV remote sensing technology,it is widely used because of its advantages of high speed,flexibility,ease of use,and high image resolution.However,the acquisition range of UAV aerial photography is limited by the flight height and camera field of view,and it is difficult for a single image taken to cover the whole target area.Therefore,it is necessary to use image stitching technology to perform panoramic stitching processing on UAV aerial photography sequence images to provide high-resolution panoramic images of large-area farmland for straw coverage and geographic information detection.In this paper,we focus on the fast stitching of field aerial photography panoramic images and study a UAV field aerial photography image stitching algorithm,the main research contents are as follows:(1)The feature detection process of SIFT algorithm is optimized,only the features in the overlapping region(20% in this paper)are extracted,the invalid feature points in the non-overlapping region are directly excluded,and the resolution of the image is reduced by downsampling processing,which again reduces the number of feature points while ensuring good stitching quality;compared with the traditional SIFT and SURF algorithms,the efficiency of the algorithm in this paper is greatly improved,the number of feature points Compared with the traditional SIFT and SURF algorithms,the efficiency of this paper is greatly improved,and the number of feature points is reduced by 98% and 94%,respectively,and the running time is also reduced by 94% and 74%,which achieves the balance of efficiency and accuracy and achieves the purpose of fast detection.(2)A new feature descriptor is introduced to first process the extracted feature points after downsampling before feature matching,and restore their coordinates and scales according to the same scale to ensure the integrity of image information and the accuracy of stitching;using new feature descriptors for matching improves the accuracy of feature detection with an average matching efficiency of about 64%,which is significantly higher than the traditional SIFT and SURF algorithms.Based on this,the RANSAC algorithm was used to continue optimizing the matching results,removing 15% to 20% of the false matches and improving the accuracy of the transformed model as well as the quality of the panoramic images.(3)The fusion strategy of optimal stitching line combined with a multi-resolution algorithm is adopted to complete the fusion of two aerial images of the field.Effective removal of shadows and stitching gaps,avoiding loss of image information,and preserving the integrity of image detail.Through the evaluation of subjective and objective indexes,the method in this paper achieves a good fusion effect and high-quality stitching in the stitching of images.(4)A layered stitching strategy is designed for the fast stitching of multiple high-resolution aerial images of fields.Compared with the traditional method,the information entropy and average gradient of the panoramic images generated by the fast stitching method in this paper have a small improvement,the contrast ratio is improved by about 20%,and the overall stitching accuracy and quality of the images are significantly improved.At the same time,from the point of view of stitching time consumption,the fast stitching method of aerial panoramic images in this paper,while improving the quality of stitched images,reduces the overall time consumption by more than 90% compared to traditional methods and meets the requirements of fast stitching detection.(5)Design and development of a software program based on the QT software platform that enables the stitching of straw cover detection images in the field,which can provide large area image information for straw cover detection and field geographic information monitoring.The stitching software integrates the improved algorithm of this paper,including the functions of parameter setting,image selection,feature detection,and panoramic stitching,and experimental testing of the software system shows good interactivity and usability.