A Hash algorithm-based large-scale image matching method

Abstract

The invention discloses a Hash algorithm-based large-scale image matching method. The method comprises the steps of 1) extracting all feature points of an original image and a target image; 2) extracting a feature descriptor at each feature point, and distributing a direction value to the feature point to generate a feature vector; 3) mapping all the feature vectors to a hash table by using a hashalgorithm, enabling each feature vector to correspond to one hash feature code, and respectively distributing the hash feature codes to a plurality of bucket groups by using a mapping function; And 4) selecting the feature vector of one feature point from the original image as an original feature, and matching the original image with the target image by using the feature vector in the hash tableto obtain matched homonymous points. According to the method, quick and accurate matching of large-scale images can be achieved, the matching efficiency of the large-scale images is effectively improved, and meanwhile the geometric accuracy of large-area image matching can be guaranteed.

Description

technical field [0001] The invention relates to the field of image matching, in particular to an image matching method based on a hash algorithm. Background technique [0002] Image matching is to find one or more transformation relationships in the transformed space, so that two or more images of the same scene from different times, different sensors or different perspectives are consistent in spatial position. The essence is to find points with the same name between two or more images. [0003] The well-known and widely used SIFT algorithm in the field of computer vision detects a feature point that is invariant to scale and rotation changes by establishing a scale space for the image and a direction vector for the feature point. This algorithm can obtain the position and neighborhood information of feature points very well, so that each feature point has a high degree of uniqueness. However, due to factors such as relatively high resolution of aerial images, large overl...

AI Technical Summary

Problems solved by technology

However, due to factors such as relatively high resolution of aerial images, large overlap rate, and large number of images, as well as direct use of feature descriptors for distance measurement and exhaustive matching strategies, the SIFT algorithm is traversing and searching for matching points in each image. The amount of calculation is huge, the speed is greatly reduced, and it cannot meet the current requirements of aerial drone photography

Images