Improved SIFT Algorithm Based For Image Matching

Unmanned aerial photography is used in many fields such as traffic surveillance,surveillance inspection,power inspection,agricultural plant protection.However,the image processing of drone aerial photography faces many problems such as high resolution of the captured image,complex background,susceptibility to weather that resulting in excessive image noise and high real-time requirements.Classical algorithms such as Harris,SIFT(Scale-invariant feature transform),and FAST cannot ideally meet the requirements of real-time and accuracy.Therefore,in-depth research on image preprocessing and image matching algorithms in the field of drones is needed.This article first preprocessest is to reduce noise and increase contrast of the original image;then to improve the SIFT algorithm use the improved algorithm to extract feature points of the image and use dimensionality reduction processing to improve the operation efficiency of the algorithm.Finally image matching is performed and multiple conditional constraints are used in the matching process to improve the matching accuracy rate.The main research contents are as follows:(1)For the drone aerial image is too noisy because the content such as jitter and light is relatively unclear.The combination of wavelet transform and Laplace operator in the image preprocessing process can effectively enhance the contrast of the image and remove the noise from the image.(2)For the traditional SIFT algorithm,the running time is too long and there is no way to meet the high real-time scenario.Therefore,an improved SIFT algorithm is proposed.In the process of constructing feature point descriptors,a circular neighborhood is constructed with key points as the center of the circle,divided into 10 directions,and a 50-dimensional feature descriptor is generated.Improved algorithm speed(3)Combine the two-dimensional entropy of the image,and compare the relationship between the two-dimensional entropy and the empirical threshold.reducing the key points required in the matching process can improve the speed and accuracy of the algorithm.(4)In the key point matching process,a multi-conditional constraint algorithm combining quasi-Euclidean distance and two-dimensional entropy of the image is proposed,which can effectively improve the accuracy of algorithm matching.