Reached On Oblique Aerial Image Feature Matching Algorithm Based On Improved ASIFT

The sources of geospatial data are increasingly multi-sourced and refined.The emergence and development of oblique aerial photogrammetry has brought a new perspective to the geospatial information industry,enabling geospatial information to be more comprehensive,refined and intuitively expressed..With the development and open source of drone technology,the data acquisition method of the ground trust industry tends to be more refined.In this three land adjustments,many base maps and places that are difficult to reach by humans are drafted in the form of drones.And homework.Drones are portable,lightweight,and easily deform and tilt the image during image acquisition.Even the image is distorted.In addition to the tilt photogrammetry technology,the need for tilted images itself makes it necessary for each algorithm to process the image during the image processing stage.The "tilt" has a certain degree of invariance.Based on this,the main research contents and conclusions of this paper are as follows:(1)Improvement based on ASIFT algorithm.Based on previous research,this article uses the ASIFT algorithm to extract and extract the Affine algorithm.The Affine algorithm simulates the longitude and latitude angles of the camera while measuring the horizontal and vertical offset of the image.The vertical offset is tilt,Calculate the image tiltfrom the latitude angle.After the Affine algorithm is obtained,it is combined with the traditional classic SURF,ORB,AKAZE,BRISK and Harris operators to further improve the algorithm.The experimental results show that the improved algorithm feature extraction is significantly better than before the improvement.Among them,the worst BRISK algorithm is improved by 8.5 times before the improvement,and the feature points extracted by the SURF algorithm are 146.5 times that of the original algorithm.(2)Use the nearest neighbor search method(FLANN)for the extracted feature points,use the KD-TREE algorithm to determine the norm distance and Hamming distance to complete the rough matching of the image feature points,and use the RANSAC-based homography matrix matching again.The algorithm realizes the exact matching between the threshold and the feature points of the image.(3)When the Harris algorithm is improved,after experiment comparison,the algorithm is used to describe the characteristics,and then the scale,affine,color difference,and noise experiments are compared with the original algorithm.The A-Harris algorithm has a certain degree of Excellent.(4)After improving the algorithm,get the improved algorithm,and add A before the algorithm name.Through statistical matching and analysis of experiments with overlapping images on different routes andadjacent images of each lens on the same route,the experiments show that the images on the same route have an excellent matching effect and are short in time.Poor and time consuming.The improved algorithm and the original algorithm have the same degree of excellence in their respective systems.The A-ORB algorithm has the absolute advantage in time.The ASIFT algorithm and the A-BRISK algorithm have the same advantages in feature extraction.For the same degree of matching,A-BRISK algorithm is more time-consuming and relatively consumes more memory,which is likely to cause memory overflow.The A-AKAZE algorithm has intermediateness between time-consuming and matching.The A-Harris algorithm takes slightly longer.A-Harris and A-The matching rate of SURF is relatively stable.SURF is an improved acceleration algorithm based on SIFT.In the improved algorithm of this paper,A-SURF algorithm still has the time advantage,and the matching effect is not inferior to ASIFT algorithm.